1 Introduction

Artificial intelligence (AI) is the most advanced and complex human creation since its theoretical emergence in the early 1950s (Duan et al. 2019). In today’s technology-driven society, data is the new gold. Massive volumes of data are available for research and development, propelling the AI revolution (Dinh and Thai 2018). A variety of AI techniques take an interdisciplinary approach and may be used in a variety of sectors, including generalized medicine (Saba et al. 2012; Hamet and Tremblay 2017; Straw 2020), medical diagnosis (Suri 2008; Saba et al. 2022), and healthcare (Jiang et al. 2017; Yu et al. 2018; Davenport and Kalakota 2019). AI techniques, including computer-aided diagnosis (CAD) systems, have the potential to transform medical diagnosis and enhance patient outcomes (Castellino 2005; Giger and Suzuki 2008). By combining the knowledge of healthcare professionals with AI algorithms, CAD enables more accurate and efficient disease detection (Doi et al. 1999; Baaske et al. 2022). CAD systems heavily rely on diverse and comprehensive datasets to learn patterns and features from large amounts of data, enabling precise diagnoses and predictions (Abbasian Ardakani et al. 2021; Baaske et al. 2022). The availability of high-quality datasets is crucial for ensuring the reliability and effectiveness of CAD systems (Bancilhon et al. 1985). AI is simpler to use and better at detecting and predicting sickness, but a lack of data restricts its usage (Kohli et al. 2017; Chen et al. 2018). Using AI, doctors can better diagnose and forecast illness (Suri and Rangayyan 2006; Kuppili et al. 2017; Tandel et al. 2020). Epidemic highlights need to understand how COVID-19 data is acquired, transferred, and reported to forecast and prevent the spreading of covariant (Ienca and Vayena 2020). AI-based systems (a) learn, (b) interpret, and (c) draw conclusions from enormous data sets (Bengio and LeCun 2007). Machine learning (ML) algorithms perform better with data from a trustworthy, legitimate, secure, and trusted platform or data repository (Gupta et al. 2020a, b). In addition, paper-based medical records have been replaced with electronic ones, necessitating a secure method to transmit and gather data (Huang et al. 2019). Some major issues in healthcare are (a) data storage and security (b) healthcare system interoperability, (c) decentralizing digital healthcare, and (d) data transparency and trust (Churi et al. 2021). Since most healthcare data for AI model training comes from a closed ecosystem of siloed organizations, it may be inaccurate or biased. (Barhamgi and Bertino 2022). Key medical information stakeholders include doctors, researchers, healthcare organizations, and the government (Chen et al. 2019). To overcome these limitations, a fine balance between data protection, global health, public trust as well as robust human-AI interactions is essential (Naudé 2020). Best practices should be defined to ensure appropriate data collection and standards worldwide (Ienca and Vayena 2020). AI has the potential to revolutionize the healthcare industry if provided with secure data and transparency (Saba et al. 2019). Using encrypted and privacy-preserving technology, the Blockchain may help AI secure data and markets, decentralize computation, and coordinate untrusted devices. (Pandl et al. 2020).

The Blockchain is a distributed ledger that stores and exchanges data in a cryptographically safe, authenticated, and immutable way (Nofer et al. 2017). Blockchain is a connected network of blocks that preserves transactions permanently (Dinh et al. 2017; Zheng et al. 2017). This approach uses a document’s timestamp, which cannot be altered (King and Nadal 2012; Nofer et al. 2017). Previously, Blockchain was primarily employed as the underpinning technology for cryptocurrencies and Bitcoin (Nakamoto 2008; Narayanan et al. 2016). Blockchain has various application areas, ranging from supply chain (Francisco and Swanson 2018), education (Gräther et al. 2018; Bathula et al. 2022a, b), edge-computing services (Xiong et al. 2018), internet-of-things (IoT) (Reyna et al. 2018), finance (Treleaven et al. 2017), and the sharing of clinical or radiological data (Zhang et al. 2018; Tagliafico et al. 2022). The use of Blockchain in healthcare is gaining popularity in academic and non-academic fields (Ekblaw et al. 2016) and may be used to regulate access to electronic health information (Dagher et al. 2018).

In addition to the challenges highlighted in the introduction, the healthcare industry faces significant hurdles in ensuring data storage, security, and interoperability (Churi et al. 2021). Traditional paper-based medical records have been replaced with electronic systems, necessitating secure methods for data transmission and storage (Weeks 2013). Moreover, the closed ecosystem of siloed healthcare organizations contributes to data fragmentation and bias, limiting the effectiveness of AI models trained on incomplete or biased datasets (Barhamgi and Bertino 2022). The fusion of AI and blockchain technologies in healthcare offers promising solutions to overcome these challenges. By leveraging the decentralized and immutable nature of blockchain technology, AI systems can access secure, tamper-proof data repositories, thereby mitigating concerns regarding data integrity and privacy (Pandl et al. 2020). Blockchain-based solutions offer promising avenues for regulating access to electronic health information, facilitating secure data exchange, and ensuring patient privacy (Dagher et al. 2018).

The proposed review attempts to study how the fusion of AI and blockchain technologies can address existing challenges in healthcare data privacy, and security. By analyzing current research and emerging trends, the review seeks to provide insights into best practices and strategies for harnessing the transformative potential of AI and blockchain in healthcare. Figure 1 demonstrates the intersection of Blockchain and AI in healthcare. Data from hospitals, patients, clinical labs, and diagnostic centers is collected (top region). This data fuels AI and Blockchain applications (“AI Layer” and “Blockchain Layer”) with a focus on security.

Fig. 1
figure 1

The convergence of Blockchain and AI in healthcare: a systematic integration for revolutionizing data management. API application programming interface, EHR electronic health records, IPFS interplanetary file system

Full size image

The outcomes are utilized by physicians, healthcare providers, researchers, the public, and government agencies (bottom region). Integration is achieved through APIs (yellow boxes), while data storage is handled by the IPFS). When AI and Blockchain are combined, they may provide a secure, immutable, and decentralized system for sensitive data (Marwala and Xing 2018). Adopting Blockchain will entail AI’s interpretability, trust, and privacy issues. As a result, the integration of these two technologies seems inevitable (Dinh and Thai 2018; Zhang et al. 2021b).

1.1 Motivation

The proposed review’s major motivation is to explore the potential benefits of integrating Blockchain and AI in healthcare, emphasizing the importance of privacy and trust across multiple institutions (Zheng et al. 2019; Zerka et al. 2020). With the growing adoption of digital solutions and the need for secure and efficient healthcare delivery, there is a pressing demand to explore innovative approaches that leverage AI and blockchain. The study aims to address this demand by examining the applications, issues, and solutions related to the fusion of AI and blockchain in healthcare. By identifying key trends and insights, the study aims to enlighten healthcare practitioners, researchers, and the government about the transformative potential of AI-driven healthcare solutions supported by blockchain technology. Ultimately, the motivation lies in advancing the understanding of how this fusion of emerging technologies can revolutionize healthcare delivery and improve patient outcomes.

1.2 Contributions

  • Understanding and presenting a detailed review of the fusion of AI and blockchain in healthcare.

  • Identifying trends, challenges, and potential benefits associated with this integration in healthcare.

  • Identifying AI’s potential hazards in the healthcare system.

  • Highlights fusion of AI and blockchain technologies in securing EHRs and addressing COVID-19 data transmission.

  • The study covers key aspects including Blockchain processes, features applied in healthcare, various Blockchain types, challenges, and their solutions, as well as the architecture of Blockchain-based healthcare systems.

  • Underscores the significance of AI and blockchain dependencies for decentralized computing and scalable architectures.

  • The study emphasizes blockchain’s potential to enhance data security, privacy, and transmission in healthcare AI, while also adding to the knowledge on how blockchain can benefit the AI-driven healthcare sector.

  • Highlights advanced AI techniques like pruning and explainable AI combined with Blockchain for enhancing healthcare systems for the future.

Here is the review’s structure and content: Sect. 2 details the literature search strategy using the PRISMA model, providing statistical insights into the prevalence of AI and Blockchain integration in healthcare research. Section 3 introduces Blockchain engineering basics, while Sect. 4 offers a concise overview of Artificial Intelligence. Section 5 explores the synergies between AI and Blockchain, emphasizing their interdependence. Section 6 delves into the integration of Blockchain with AI in healthcare, presenting a tripod framework for analysis. In Sect. 7, a critical discussion examines challenges and opportunities, considering real-world applications, and Sect. 8 concludes the study.

2 Search strategy and its statistical distribution

This review is based on 402 research and review works. The section has been split into two segments: one discusses the citation selection employing the PRISMA model, while the other part analyzes the statistical distributions of the respective repositories.

2.1 PRISMA model

The PRISMA schema model was adapted for this study, and an extensive search was carried out on PubMed, IEEE, and Google scholar, resulting in the consideration of 402 relevant articles. Keywords used included “AI in healthcare,” “Blockchain and healthcare,” “deep learning in healthcare,” “federated learning and healthcare,” “Blockchain and deep learning,” “Blockchain and federated learning,” “Blockchain and deep reinforcement learning (DRL),” “AI and Blockchain integration,” “AI and Blockchain in healthcare supply chain management,” “AI and Blockchain in e-Health,” “AI and Blockchain in EHR,” “AI and Blockchain in clinical research and supply chain,” “AI without Blockchain,” “Blockchain without AI,” “AI integration with Blockchain,” and “Blockchain integration with AI”. In this study, we used bibliometric variables and the narrative literature review (NLR) approach (Jahan et al. 2016). We focused on comprehensive research articles with innovative techniques in healthcare, excluding irrelevant and non-analytical publications like conference abstracts, posters, and hypothetical discussions. The PRISMA model illustrating the incorporation of Blockchain and AI-related citations utilized in our study is displayed in Fig. 2. Here we meticulously detail the procedures undertaken to choose and analyze the 402 studies. Here’s an overview of our Article Selection Process:

  • Search Strategy: A comprehensive search was conducted using PubMed

  • , IEEE, and Google Scholar databases to identify relevant articles.

  • Initial Search Results: A total of 9239 articles were sourced from PubMed, 6321 articles from IEEE, and 2,36,703 articles from Google Scholar, resulting in a combined total of 2,52,263 articles.

  • Screening Process: The titles and abstracts of the remaining articles were screened against the inclusion and exclusion criteria, resulting in the exclusion of 2,13,288 records.

  • Duplicate Removal: To locate duplicate articles, we utilized Clarivate Analytics' EndNote software “Find Duplicates” function (Analytics 2018) and removed 31,723 records.

  • Article Eligibility: Following the initial screening, 7252 articles were deemed eligible for full-text review.

  • Full-Text Review: The full texts of these articles were reviewed in detail to assess their relevance and alignment with the research objectives. A total of 4127 articles with full text were evaluated for eligibility.

  • Insufficient Data Exclusion: After the full-text review, a total of 3731 articles were excluded due to insufficient data or lack of relevance to the research topic.

  • Final Selection: Out of the evaluations conducted, articles that met the eligibility criteria were selected for further analysis, contributing to a total of 402 included articles.

Fig. 2
figure 2

PRISMA selection model

Full size image

2.2 Statistical distributions and trends in AI-enabled blockchain articles in healthcare

2.2.1 Blockchain and AI-enabled articles are considered for healthcare studies

Out of 402 relevant articles considered for this research, there are a total of 79 publications that use both AI and Blockchain in healthcare, with a refined analysis centered on 51 implemented projects. The statistical distribution and discussion are based on these 51 publications. We can see from Fig. 3a that there has been a massive rise in the number of articles produced, mostly appearing between 2020 and 2023. In 2017, only a single study combined AI and Blockchain in healthcare. In 2018, there were two studies published about merging AI and Blockchain in healthcare, while four in 2019. The number of publications published climbed substantially, to 20 in 2022 and 24 in 2023. Figure 3b shows the number of articles implemented. Out of 79 articles on AI and Blockchain in healthcare, 51 are implemented, while 28 are not. Figure 3c displays the distribution of implemented and unimplemented articles from 2017 to 2023 about the use of AI and Blockchain in healthcare. In summary, there has been an increase in the fusion of blockchain, and AI, in the healthcare industry.

Fig. 3
figure 3

a Distribution of Published Articles Using AI and Blockchain in Healthcare Research. b A Comparison of Implemented and Non-Implemented Articles on Blockchain and AI Integration in Healthcare. c Annual Distribution of AI and Blockchain Integration in Healthcare Implemented and Non-Implemented Articles (2017–2023)

Full size image

2.2.2 Exploring the diverse landscape of AI-enabled blockchain articles: blockchain types

Examining the implemented integrated articles, blockchain types utilized by these healthcare articles are (i) public (ii), private (iii) consortium, and (iv) hybrid (Alhadhrami et al. 2017). Figure 4a gives an illustration of several articles that employed different forms of Blockchains specific to the healthcare sector. Figure 4b gives a more detailed analysis of year-wise publications utilizing the type of blockchain in implemented articles.

Fig. 4
figure 4

a Number of articles employing a type of blockchain in implemented articles. b Distribution of blockchain types employed in AI and Blockchain Integration in Healthcare articles (2017–2023)

Full size image

Out of all the 51 implemented articles, 20 publications used private Blockchain, 14 used public Blockchain, only two employed consortium Blockchain, and one utilized hybrid Blockchain. Public blockchains offer users convenience through their open, permissionless access and decentralized structure, thereby enhancing simplicity and user autonomy. A private blockchain allows for limited access and data sharing among authorized users. The consortium blockchain has been in use since 2021, and its popularity is growing rapidly. Because of the complex nature of the hybrid blockchain, it has rarely been used by researchers. Despite the growing popularity of Blockchain, many authors in their articles do not specify the type of Blockchain used.

2.2.3 Exploring the diverse landscape of AI-enabled blockchain articles: utilised AI techniques

Examining the implemented integrated articles, AI techniques utilized by these healthcare articles are discussed here. Figure 5a gives an illustration that employed different forms of AI techniques used in the 51 fusion articles specific to the healthcare industry. Figure 5b gives a more detailed analysis of year-wise publications utilizing the type of AI technique in implemented articles. The analysis of articles utilizing various AI techniques across the years reveals an evolving landscape. In the cumulative span from 2017 to 2023, out of fifty-one published articles a total of seven articles incorporated Machine Learning (ML), with a noticeable rise from 2019 onwards. Deep Learning (DL) demonstrated consistent adoption, with a total of eleven articles, highlighting its sustained relevance. Federated Learning (FL) gained prominence, contributing to eight articles, particularly peaking in 2022. Reinforcement Learning (RL) and the combination of ML + DL each played a role in three articles. Remarkably, the combination of DL + FL emerged as a prevalent choice, appearing in thirteen articles. The application of Explainable-AI (XAI) techniques surfaced in one article, reflecting a growing interest in transparent AI methodologies.

Fig. 5
figure 5

a Number of articles employing AI techniques in implemented articles. b Distribution of AI techniques employed in AI and Blockchain Integration in Healthcare articles (2017–2023). AI Artificial Intelligence, ML Machine learning, DL Deep learning, FL Federated learning, RL Reinforcement learning, XAI:Explainable AI

Full size image

2.2.4 Exploring the diverse landscape of AI-enabled blockchain articles: diverse applications

The merging of AI and blockchain technology has ushered in new avenues of exploration in healthcare. These advancements are being used in a spectrum of applications in the healthcare sector namely for early diagnosis, patient care, and data security. Figure 6 shows how Blockchain integration with AI is being used for various applications in healthcare from 2017 to 2023. Securing (EHR), electronic health data (EHD), and patients’ healthcare records (PHCR) as well as transmitting and updating COVID-19-related data securely among hospitals are the top two applications, accounting for 39% and 31% of all articles, respectively. The remaining 30% contribute to the ever-changing field of Blockchain and AI in healthcare by being employed in a variety of sectors such as clinical research, cardiovascular medicine, organ transplant systems, illness prediction, and specialized medical professions.

Fig. 6
figure 6

Various Applications with the fusion of AI and Blockchain in healthcare

Full size image

Table 1 includes references to numerous research publications on the of AI and blockchain in healthcare, and the suggested review in 2023. It contains information such as the reference and year, application focus, keywords used, period covered, number of citations, and checkboxes indicating the existence of AI (ML, DL, FL, RL, XAI) and blockchain (BC) technology. Alzheimer’s disease, COVID-19, e-health, cancer care, public health, EHR management systems, and healthcare insurance are among the topics covered in the review papers. Our proposed review comprehensively covers various facets of healthcare and heavily emphasizes both AI and blockchain, demonstrating a holistic approach to the integration of both technologies.

Table 1 Comparative analysis of existing AI-enabled Blockchain review articles in healthcare
Full size table

3 The engineering of blockchain-background

3.1 The blockchain processes

Blockchain is a technology that doesn't have to be controlled by anyone, and it only became popular after it was used for Bitcoin in 2008 (Narayanan et al. 2016; Nakamoto 2008). Blockchain’s key characteristics, such as anonymity, confidentiality, data integrity, and lack of the need for a third-party organization, are boosting interest in Blockchain (Feng et al. 2019; Islam et al. 2020). A Blockchain is a sequence of blocks that contain verifiable transaction data (Niranjanamurthy et al. 2019) that are linked together by a connection (hash value) to the preceding block, forming a chain-like structure (Zheng et al. 2017). The header of a block contains information about the date, a hash representation, the hash of the previous block, and the cryptographic nonce (Zheng et al. 2017). The first block of a blockchain is commonly known as the genesis block (Nakamoto 2008). The genesis block is represented with ‘0000’ in Fig. 7 as it does not have any previous hash. Data integrity and immutability are ensured by the transactions in each block and their references to the preceding block.

Fig. 7
figure 7

Blockchain block diagram

Full size image

The fundamental architecture of Blockchain is seen in Fig. 8, which consists of a connected chain of blocks. In each block, it has the attributes such as header, timestamp, nonce, data, and previous hash. Consensus methods are created to assess the trustworthiness of the blocks and decide which node will store the next block and how other nodes will confirm the newly added block. Some examples of consensus algorithms are Proof-of-work (PoW) (Nakamoto 2008), Proof-of-stake (PoS) (King and Nadal 2012), and Practical byzantine-fault tolerance (PBFT) (Castro and Liskov 1999; King and Nadal 2012). The Blockchain process steps specific to healthcare are explained in Fig. 9. It has a sender, receiver, and a distributed database to do transactions and create a block. It uses cryptographic hashing for the distribution and validation process. The transactions are committed to the Blockchain once the validation is done by the nodes.

Fig. 8
figure 8

The basic blockchain structure

Full size image
Fig. 9
figure 9

The blockchain process

Full size image

3.2 Blockchain-based security, big data privacy and availability

Considering the findings of a study on data (Halevy et al. 2009), if given enough data, even the most basic AI system could surpass most modern technology. For the goal of disease diagnosis and detection, big data may be analyzed using AI algorithms. Such data is used to create the models, for example, machine learning since more relevant data may lead to generalized training leading to more accurate predictions. Here, the key challenge is to determine how to make data interchangeably reliable, secure, and available (Alhazmi et al. 2022). Genuine information integrity may be compromised for several reasons (Hernandez and Zhang 2017). The scale of fraudulent actions has grown alongside the development of technology. As a result, one of the most important aspects of the process is not having proper security procedures. Even during the outbreak, it was difficult to manage Coronavirus patient data due to databases all over the world as the data was massive, and was manipulated along the way (Kabir and Marlow 2022). Moreover, the European Medicines Agency (EMA) has detailed how hackers acquired official communications concerning COVID-19 vaccine evaluations. We are aware that a single, deidentified dataset is necessary for predictive analytics in healthcare to improve model training (Lee et al. 2017). In healthcare, it is exceptionally challenging to collect massive volumes of data, keep it secure, and make it accessible on demand without compromising security. Fortunately, Blockchain technologies may provide a feasible solution to this problem (Cheng et al. 2020; Huang et al. 2020; Archana Bathula 2022). Blockchain has emerged as a potentially game-changing technology with various features among which the two most crucial aspects of it are (i) data security and (ii) privacy (Zhang et al. 2021a, b). The data structures created by Blockchain technology already have security safeguards built into them, which is a crucial component of the Blockchain’s security (Zheng et al. 2017).

Blockchain’s Distributed Ledger Technology (DLT) organizes data into blocks. In a cryptographic chain, each block is linked to the ones before it to prevent tampering, and hence obtain immutability (Zhang and Jacobsen 2018). A consensus mechanism and cryptography safeguard validate and approve each block transaction, ensuring its legitimacy and trustworthiness (Pan et al. 2021). The network’s nodes will serve as a repository for the data stored on the blockchain. It will be safe from riots, natural disasters, hacking attempts, and other threats that currently bring down and destroy centralized systems (Alhazmi et al. 2022) and from a single point of failure or any information manipulation or theft. In a distributed blockchain data storage network, a backup copy of the data is always accessible and may be easily retrieved whenever necessary, regardless of what happens to a node. Patients can set access limitations for their medical data using Blockchain, enabling researchers’ temporary access (Dagher et al. 2018; Liu et al. 2020a, b). Blockchain’s Zero-Knowledge Proof (ZKP) (Sun et al. 2021), is a reliable method in such a scenario, wherein one node (the prover) demonstrates to another node (the verifier) that it is aware of a certain piece of information without disclosing the information’s actual content (Yang and Li 2020). With the use of ZKP private and secure transactions may be accomplished on a public Blockchain (Rasheed et al. 2021).

3.3 Types of blockchains

There are three kinds of DLT systems and they are permissioned, permissionless, and hybrid networks, and their characteristics are based on the user’s activities and data availability (Nofer et al. 2017; Niranjanamurthy et al. 2019). The specific aims and needs of the network influence the choice of a blockchain network. Enterprise environments that prioritize privacy and control are best suited to permissioned Blockchains. Permissionless Blockchains offer high transparency, security, and trust. Hybrid Blockchains combine permissioned and permissionless Blockchains. The varieties of blockchains with their accessing capabilities are shown in Fig. 10.

Fig. 10
figure 10

Types of blockchains

Full size image

Public blockchains, exemplified by Bitcoin and Ethereum, stand as open and permissionless networks accessible to anyone (Nakamoto 2008). Renowned for their attributes of transparency, decentralization, and immutability, these networks prove ideal for applications demanding trust and security without reliance on a central authority. Bitcoin functions on a Proof of Work (PoW) model, where miners competitively solve puzzles to validate transactions (Velde 2013). Ethereum 2.0 introduces a Proof of Stake (PoS) blockchain, selecting validators based on their cryptocurrency holdings (Dannen 2017). “Endless OS” (EOS) utilizes Delegated Proof of Stake (DPoS), enabling token holders to choose delegates for transaction validation (Hu et al. 2021a, b). IOTA follows a unique approach with a Directed Acyclic Graph (DAG), where transactions confirm previous ones (Živić et al. 2020; Bathula et al. 2022a, b).

Hybrid blockchains blend features from both permissioned and permissionless models, offering a flexible solution where parts can be public while others remain private (Marar and Marar 2020). This approach is ideal for scenarios requiring a balance between transparency and privacy. An extension of hybrid blockchains involves integrating side chains, offering scalability and efficiency by handling specific tasks independently (Back et al. 2014).

Private Blockchains are authorized networks with limited participation and access. They are often used by organizations or consortiums to maintain control over the network and ensure data privacy (Dinh et al. 2017). They provide selective data exchange among authorized parties, quicker transaction processing, and improved scalability.

Consortium blockchains blend aspects of private and public blockchains, maintained cooperatively by selected members. They offer a more decentralized approach than private blockchains, with consensus methods varying (Dib et al. 2018). Each blockchain type has its own set of traits and applications, enabling organizations and individuals to choose the most suitable blockchain architecture based on their specific requirements for security, scalability, control, and data sharing.

3.4 Blockchain architecture

Blockchain software development is quite difficult and involves the consideration of many factors. There are five layers in the architecture of Blockchain to integrate Blockchain into healthcare applications. There are layers such as (i) Blockchain layer and decentralized ledger technologies (DLT) (ii) API (iii) Interface, and (iv) Application. Figure 11 describes how healthcare participants can safely exchange or gather data using Blockchain by providing the various interface layers of the Blockchain architecture.

Fig. 11
figure 11

The architecture of Blockchain-based healthcare system. API: Application programming interface, SDK: Software Development Kit, DLT: Distributed ledger technology

Full size image

Blockchain layer It is a platform for architectural design and a communication channel. It is called a base protocol layer and is accomplished via the use of technologies like Ethereum, Hyperledger, R3 Corda, Ripple, and Quorum (Androulaki et al. 2018; Benji and Sindhu 2019). The consensus protocol is an approach used for implementing this Blockchain layer (King and Nadal 2012; Hasselgren et al. 2021).

DLT layer A decentralized ledger technology Layer (DLT) is a network of users, or nodes, that share and manage a database of records. It is tampered-resistant, and encryption methods are used for the safe and secure storage of any data. It will store data that is unchangeable and tamper-proof. IPFS, Filecoin, Sia, Stroj, and Swarm are some instances of DLT layers (Orhan et al. 2021; Warnat-Herresthal et al. 2021).

Application programming interfaces and software development kit layer (API/SDK) It consists of libraries, whereas an SDK is a development kit that enables API usage. Notably, Blockchain API is a critical component of this architecture (Dinh and Thai 2018). REST API, WEB3 JS, RPC, Firebase, SOAP, OpenML, Unity, SigularityNET, and CoinMarketCap are examples of APIs that provide efficient mechanisms for connecting to software components or resources (Gropper 2016; Zhuang et al. 2019). SDK, on the other hand, provides vital building tools such as code libraries, compilers, and documentation to aid developers in software development. Cosmos SDK and Tatum SDK are two examples. Recognizing the importance of both API and SDK, they play a crucial role in allowing effective application integration (Kwon and Buchman 2019).

Interface and application layers The interface layer of a healthcare system allows user interaction and access to all features through desktop and mobile applications via this layer. Applications in the interface layer may also provide login and registration functions to authorize users (Zhang et al. 2021b). The top layer (application layer) enables (healthcare) applications to access and interact with data securely. It covers various healthcare applications like EHR, patient identification management, healthcare supply chain management, clinical trials, biomedical research, and COVID-19 response (Khatri et al. 2021).

3.5 Blockchain challenges and its solutions in healthcare

Some of the biggest problems in healthcare that blockchain may tackle include interoperability of electronic health records, patient care, and data privacy concerns (Khatri et al. 2021). Blockchain allows efficient data exchange while maintaining data integrity and patient privacy (Katuwal et al. 2018). Blockchain networks are now being used in many different industries, one of which is healthcare (Mettler 2016; Alhadhrami et al. 2017). The three desired features that Blockchain systems try to attain are consistency, decentralization, and scalability (CDS) (Islam et al. 2021). In Table 2, we have presented an overview of key problems in Blockchain-based healthcare applications and their potential solutions.

Table 2 Critical issues/challenges and their solutions in Blockchain-based healthcare applications
Full size table

4 Overview of artificial intelligence: a brief exploration

4.1 Artificial intelligence background and evolution

AI is a domain within computer science focused on the creation of intelligent machines. Originating in the 1950s, AI has transformed by leveraging advances in algorithms, processing powers, and data accessibility, driving the discipline to unprecedented heights and matching human intellect.

The very first aspect of health care is accumulating and evaluating data such as medical records and historical data (Duan et al. 2019). Data acquisition is the most extensively needed in AI for training models and ensuring digital automation (Yu et al. 2018). AI can successfully handle categorization by mapping nonlinearity between input fluctuations and illness severity (Davenport and Kalakota 2019, Skandha et al. 2020a, b, Tandel et al. 2020). AI methods are used for predicting a patient’s treatment outcomes based on their specific characteristics and the treatment context (Suo et al. 2018). It also tries to address the problem of sparse and small datasets in the medical field (Pereira et al. 2021). Data is essential to AI and ML algorithms, which perform better with reliable, secure, trustworthy, and credible data (Chen et al. 2018). AI in healthcare confronts several obstacles, including ethical use, algorithm biases, data privacy, and trust building. These challenges must be addressed to ensure patient welfare and safety while ensuring data privacy and security.

4.2 Popular AI learning techniques in healthcare

ML is the foundation of AI algorithms, which interpret data, learn from it, and make predictions. AI algorithms are classified into four types (i) supervised learning, (ii) unsupervised learning, (iii) reinforced Learning (iv) federated learning. Machine learning (ML) and deep learning (DL) are two types of AI-supervised learning techniques (Davenport and Kalakota 2019). Figure 12 provides a comprehensive taxonomy of learning techniques, illustrating the diverse methods contributing to AI subjects.

Fig. 12
figure 12

Mapping the AI Landscape: A Comprehensive Taxonomy of Learning Techniques. DL-Reg: Deep Learning Regression, NN-MO: Neural Network Multiple Object Regression, KNN: K- Nearest Neighbor, SVM: Support Vector Machine, DT: Decision Tree, RF: (Random Forest), XG: Extreme Gradient Boost, Convolution NN: Convolution Neural Network, RNN: Recurrent Neural Network, GMM: Gaussian Mixture Model, EM: Expectation and Maximization, DEC: Deep Embedded Clustering, JULE: Joint unsupervised learning Clustering, ACOL: Adversarial Learning Clustering, DCC: Deep Continuous Clustering, VaDE: Variational deep embedding Clustering, CCNN: Clustering CNN, RCNN: (Region-based Convolutional Neural Network), GAN: (Generative Adversarial Learning), MRCNN: Mask RCNN, SARSA: (State Action Reward State Action)

Full size image

4.2.1 Supervised learning

In this learning, each data point is assigned a label by subject-matter experts who label the data as a whole using two techniques regression and classification. Several methods have been devised to address diverse regression challenges, each with unique features and applications.

4.2.2 Regression

It is a supervised learning problem to predict a continuous numerical output. Regression techniques use the labeled cohort to predict the next data point using parametric-based techniques such as linear regression or logistic regression (Jamthikar et al. 2020; Teji et al. 2022).

Linear regression Determines a linear connection between input variables to predict continuous numerical results. Linear regression is used in healthcare to predict patient outcomes, such as calculating illness progression based on multiple parameters (Kan et al. 2019). In addition to that linear regression in healthcare is used for analyzing the patient’s history, especially in clinical trials.

Logistic regression Despite its name, logistic regression is employed for binary classification, evaluating the likelihood of an instance belonging to a given class (Panda 2022). It is used in healthcare to forecast whether a patient will have a given ailment or not.

Ridge regression A regression approach that uses regularization to avoid overfitting, which is especially beneficial when dealing with multicollinearity. Ridge regression is a tool used in healthcare to predict intricate interactions between several health indicators (Saqib 2021).

Lasso It encourages smaller models by picking important characteristics and avoids overfitting. It is used for medical feature selection, disease diagnosis, and prediction (van Egmond et al. 2021).

Artificial neural networks (ANN) Neural networks, which are influenced by the functioning of the human brain, learn intricate patterns through connected layers of nodes. Image recognition, predictive modeling, and pattern analysis are examples of applications in healthcare (Kumar and Kumar 2013).

Deep learning regularization (DL-Reg) It is a technique for reducing overfitting in deep neural networks by restricting model complexity. Important for generalization in medical applications such as image processing and patient prediction (Huang et al. 2017).

Multiple regression (multireg) It is a statistical approach that models the connection between a dependent variable and numerous independent variables (Allison 1999). Analyzing the impact of many factors on health outcomes, such as investigating the impact of patient characteristics on therapy response (Agarwal et al. 2022).

Neural network multiple object regression (NN-MO) A specific neural network approach for predicting several object-related variables in regression problems at the same time. In healthcare, NN-MO (Cui et al. 2018) can be used to predict various health-related outcomes at the same time, such as patient vitals, disease progression, or treatment reactions that occur.

4.2.3 Classification

Classification is another type of supervised learning problem in which the goal is to predict the category class or label of an input. In healthcare, classification aids in diagnosing diseases, recognizing anomalies, predicting risks, improving decision-making, and prompt interventions, ultimately enhancing patient outcomes (Peng et al. 2011; Castelli et al. 2018).

Support vector machine (SVM) SVM classifies data by determining the hyperplane in the space of high dimensions that best separates multiple classes. In healthcare, it is used to classify medical information or images, which aids in illness diagnosis and treatment planning (Srivastava et al. 2019; Jamthikar et al. 2021a, b; Konstantonis et al. 2022).

SVM is the most popular and widely used algorithm in health care for classification problems.

Naive Bayes It is a probabilistic technique based on Bayes' theorem that assumes independence among features and is often used for classification tasks (Webb et al. 2010). Employed in healthcare for activities such as email filtering in clinical communication and illness prediction based on numerous patient characteristics.

Convolutional neural network (CNN) It is a deep learning technique developed for image processing and pattern identification that uses convolutional layers to automatically learn hierarchical features (Chua and Roska 1993). Used in medical imaging to detect cancers or abnormalities in radiological images, improving diagnostic accuracy. The general alexnet-based CNN architecture is given in Fig. 13.

Fig. 13
figure 13

The general alexnet based CNN architecture (Courtesy of Dr. Luca Saba et al. 2021a, b, c, Italy)

Full size image

Decision tree It is a model in which judgments are made based on characteristics, leading to a conclusion or prediction. Medical decision-making applications include forecasting patient outcomes and choosing treatment options based on patient characteristics (Tandel et al. 2020; Skandha et al. 2022a, b). A decision tree in healthcare is used for finding the comorbidities and strong characteristics of the disease.

Random forest It is a collaborative instructional technique that builds a large number of decision trees and blends their outputs to increase overall forecast accuracy. In healthcare, for tasks such as illness risk prediction, the combined power of numerous decision trees is used to produce more robust results (Maniruzzaman et al. 2019a, b; Jamthikar et al. 2020).

Hybrid Hybrid models are the combination of two techniques of ML or DL, frequently using strengths from different algorithms to improve overall performance. In healthcare, hybrid models may incorporate diverse algorithms for tasks such as illness prediction, predictive accuracy optimization, and model resilience (Jena et al. 2021; Suri et al. 2021a, b, 2022a; Das et al. 2022; Skandha et al. 2022a, b). Figure 14 depicts an example hybrid architecture of Inception-ResNetV3.

Fig. 14
figure 14

The hybrid architecture of Inception-ResNetV3 (Courtesy of Dr. Siva Skandha et.al (Skandha et al. 2022a, b), India)

Full size image

XGBoost It is an improved gradient boosting technique that excels at managing complicated connections among data, frequently surpassing other boosting approaches. In healthcare, it is used for a variety of activities including forecasting patient outcomes and improving decision support systems (Jamthikar et al. 2021a, b; Jamthikar et al. 2021a, b).

AdaBoost It is an ensemble learning approach that combines weak learners to build a strong learner by adding weights to misclassified cases. Used to improve accuracy in tasks such as forecasting patient outcomes or detecting disease risk factors (APC 1871; Acharya et al. 2013a, b, c, d).

k-Nearest Neighbor (kNN) It is an easy method that employs the vast majority of class among its k-nearest neighbors in feature space to classify data items. In healthcare, it is used for tasks like patient similarity analysis and illness categorization based on comparable instances (Shouman et al. 2012). This classifier is also another popular classifier in healthcare, it is used as a statistical approach for clinical trials for classifying and characterizing the diseases.

Supervised capsule neural networks (CapsNets) pioneered by Sabour et al. (2017), have revolutionized computer vision tasks through supervised learning by introducing capsules within CNNs to capture specific features and enhancing feature representations (Paoletti et al. 2021). By focusing on intricate spatial relationships and providing interpretable features, CapsNets overcome the limitations of traditional CNNs (Mukhometzianov and Carrillo 2018). CapsNets are useful in diagnosing COVID-19 from medical images like X-rays and CT scans (Toraman et al. 2020; Ragab et al. 2022) identifying patterns and abnormalities, and predicting disease severity (Afshar et al. 2020; Tiwari and Jain 2021; Gupta et al. 2022). Their ability to interpret complex visual information aids in understanding and analyzing COVID-19 medical images (Farki et al. 2021; Monday et al. 2022). Moreover, CapsNets have potential in drug discovery, enabling the prediction of drug efficacy and identification of potential candidates against COVID-19 (Jin et al. 2023). Figure 15 illustrates the Sample Architecture of a Global Machine Learning Healthcare System.

Fig. 15
figure 15

Samle Architecture of a Global Machine Learning Healthcare System. (Courtesy of Dr. Luca Saba et al. 2021a, b, c), Italy)

Full size image

4.2.4 Supervised graph learning

Graph learning is a multidisciplinary technique that models and analyses complex interactions in data sets using graph-based structures. It uses clustering algorithms, graph embedding techniques, and Graph Neural Networks (GNNs) (Gori et al. 2005). GNNs have found practical applications in supervised tasks, especially node classification in social networks. Designed to process data organized in graph structures, GNNs excel in tasks like link pre-diction node classification, and community detection (Gu et al. 2021; Tandon et al. 2021). In the context of COVID-19, GNNs have been effectively applied to predict outcomes (MacLean 2021; Kumar et al. 2022a, b, c, d). By leveraging graph structures, GNNs can capture relevant relationships and dependencies, leading to improved predictive performance (Palanivinayagam et al. 2022). Graph learning is useful in COVID-19 scenarios for disease spread dynamics, healthcare resource allocation, and public health interventions. It also has applications in drug development, molecular biology, and social network research.

4.2.5 Unsupervised learning

Unsupervised learning approaches aid in the identification of patient groups with similar features, allowing for more individualized treatment strategies (Molinari et al. 2012). Unsupervised methods in AI utilize two techniques: clustering and segmentation, which combine related elements in a dataset and remove areas of interest before labeling the information (Araki et al. 2016).

4.2.5.1 Clustering

Clustering is a method that groups comparable data points based on certain qualities or attributes without previous labeling (Xie et al. 2016). It is used in healthcare to group patients who have similar medical characteristics, allowing for individualized treatment plans and personalized treatments (Strauss et al. 1973).

Deep embedded clustering (DEC) It is an unsupervised deep learning approach that blends deep neural networks with clustering to learn high-level representations of data while also performing clustering. DEC can be used for patient stratification, classifying people according to hidden patterns in their health data, perhaps leading to more personalized therapy (Wu et al. 2021).

4.2.5.2 Joint unsupervised learning (JULE)

JULE is an unsupervised method that simultaneously learns a feature representation and a clustering assignment (Moriya et al. 2018). It is used in the healthcare sector to identify patient groups with similar characteristics and to uncover hidden structures in complex medical data.

Agglomerative clustering with online learning (ACOL) It is an agglomerative clustering method that dynamically adjusts to incoming data by using online learning (Wu et al. 2020). To ensure that the clustering model adapts to new data, ACOL is useful in healthcare contexts where data is continuously updated, such as patient records.

Deep continuous clustering (DCC) It is an unsupervised learning approach that combines clustering and deep learning for continuous data (Shah and Koltun 2018). DCC may be used to analyze data from continuous patient monitoring to find trends and patterns in physiological parameters that can be used to spot abnormalities early.

Variational deep embedding (VaDE) It is a generative model that combines Gaussian mixture models with variational autoencoders (Ji et al. 2021). It can assist with tasks like anomaly detection and disease cluster identification by helping to capture complicated relationships in patient data.

Capsule-based clustering neural network (CCNN) It is a clustering technique designed for unsupervised learning tasks. Using comparable characteristics or structures, CCNN can help in automatically classifying images for study in the medical field (Zhang and Wu 2018).

K-means clustering (KMean) It is a standard clustering approach that divides data into k groups according to the mean values of the feature sets. It may be applied to healthcare tasks such as grouping patients according to clinical characteristics into discrete groups (Silitonga 2017).

Gaussian mixture model (GMM) It is a probabilistic model that makes use of a mixture of Gaussian distributions to represent a variety of complicated data distributions in an adaptable manner (Bouman et al. 1997). GMM helps represent a variety of patient demographics and encapsulates data distribution uncertainty.

Expectation–maximization (EM) It is a framework for estimating maximum likelihood in probabilistic models, particularly useful in healthcare for latent health status modeling and data imputing from patient records, addressing missing data (Moon 1996).

Fuzzy This clustering offers a more nuanced representation by allowing data points to belong to multiple groups with varying membership degrees, particularly useful in cases of overlapping clinical symptoms (Höppner et al. 1999).

Hierarchical This clustering is a method used in healthcare to organize patient data in a tree-like hierarchy, revealing both broad and fine-grained patterns, and can be applied to reveal relationships between data points (Papin et al. 2021).

4.2.5.3 Segmentation

Segmentation is the process of breaking down a picture or set of data into distinct areas or segments according to attributes. This enables a more thorough examination of each component. To recognize and distinguish features like organs, tumors, or blood arteries, segmentation is applied to medical images (Aggarwal et al. 2011; Moriya et al. 2018).

Region-based convolutional neural network (RCNN) Region suggestions are used by the RCNN object detection framework to locate and categorize items in an image (Balasubramanian 2022). Using RCNN, anomalies or specific features in medical images may be found and localized.

Generative adversarial network (GAN) A generative model known as a GAN is made up of a discriminator and a generator that is taught concurrently via adversarial training (Aggarwal et al. 2021). GANs may be used to create artificial medical images for training or to enhance data in medical imaging.

Encoder and decoder Encoder–decoder architectures are used in medical image segmentation models to capture hierarchical features from input data and reconstruct the segmented output, enabling the extraction of intricate features (Gao et al. 2019).

UNet CNN architecture created for semantic segmentation is known as UNet. Basic U-Net architecture with feature map sample sizes is given in Fig. 16. It is made up of an expanding path, a bottleneck, and a shrinking path. UNet (Jain et al. 2022; Sharma et al. 2022), and hybrid UNet (Jain et al. 2021; Suri et al. 2022a, b) methods are widely used in medical imaging.

Fig. 16
figure 16

Basic U-Net architecture with feature map sample size (Courtesy of Dr. Luca Saba et al. 2021a, b, c, Italy)

Full size image
4.2.5.4 Mask R-CNN (MRCNN)

It is a modification of RCNN that enables instance segmentation by incorporating a mask prediction branch. It is useful in medical imaging for cell detection, and accurate separation of distinct instances, such as identifying multiple cancers in a single picture (Fujita and Han 2020).

Semantic This divides an image’s pixels into distinct groups, aiding in organ segmentation and target tissue identification in medical imaging (Scheikl et al. 2020).

Instance It is a method that classifies pixels and distinguishes between instances of the same class, useful in healthcare settings like identifying multiple nodules in lung CT scans (Frade et al. 2022).

Panoptic Panoptic segmentation unifies semantic and instance segmentation, providing a comprehensive understanding of both stuff (non-specific regions) and things (specific instances) (Kirillov et al. 2019). It allows a holistic analysis of medical images, incorporating both general structures and specific entities within the same framework.

4.2.5.5 Unsupervised capsule networks (CapsNets)

In the domain of unsupervised learning, CapsNets exhibit exceptional versatility and effectiveness. It facilitates a diverse range of applications such as clustering, unsupervised feature learning, generative modeling, and anomaly detection (Piciarelli et al. 2019; Fan et al. 2020a, b). CapsNet’s proficiency in capturing intricate data structures and providing interpretable features makes it a compelling approach for extracting valuable insights from unlabeled medical images (Sharma et al. 2023). This potent and interpretable tool holds promise for addressing COVID-19 challenges without extensive labeled data (Sandu and Karim 2020).

4.2.5.6 Unsupervised graph learning

Unsupervised graph learning, a facet of machine learning, uncovers patterns in graph-structured data without labeled information (Xia et al. 2021). Graph learning approaches, particularly GNNs, have been successfully applied in various domains, including recommendation systems (Fan et al. 2019; Wang et al. 2021a, b), social network analysis (Hamid et al. 2020), drug discovery (Bongini et al. 2021), and computer vision (Chen et al. 2022). In COVID-19 detection, GNNs have proven effective in unsupervised graph tasks (Dan-Sebastian et al. 2020; Chandra et al. 2023) and in predicting outcomes (Zhou et al. 2020). Recent research is exploring the integration of blockchain technology to enhance security and privacy when sharing weight files, input images, and predictions, aiming to address data security concerns in healthcare (Liu et al. 2020a, b; Wang et al. 2020).

4.2.6 Reinforcement learning (RL)

RL including Q-learning, is used in medical imaging, particularly during the COVID-19 pandemic, to identify patterns in unlabeled data. Figure 17 depicts a model of the reinforcement learning architecture. This error-by-trial learning approach, which includes Q-learning, is often employed with unlabeled data. In Shang and Li (2022), the authors demonstrated a hybrid combinatorial remanufacturing technique that produced RL models relying on Q-learning and a deep Q-network. In (Cockrell et al. 2022), the authors use deep reinforcement learning (DRL) similar to game-playing AI. This DRL uses deterministic policy gradient using a deep model while manipulating the six cytokines such as tumor necrosis factor, four famous interleukins (1, 4, 8, and 12), and interferon-gamma. In (Böck et al. 2022) the authors used RL for the treatment of sepsis, a medical life-threatening emergency during the COVID-19 pandemic.

Fig. 17
figure 17

Sample Reinforcement learning architecture. (Courtesy of Dr. Luca Saba et al. 2021a, b, c, Italy)

Full size image

Few authors also used the DRL framework to prioritize vaccines in COVID-19 applications (Bushaj et al. 2022), comparing age-based, comorbidity-based, or random-based paradigms.

4.2.7 Federated learning (FL)

ML with a decentralized structure, based on a star topology is known as FL (Yang et al. 2019). In FL, the models will be trained on individual user-edge devices, with the resulting updates being delivered to a central server (Lu et al. 2020; Shen et al. 2020; Feng et al. 2021). Predictions in subsequent cases are made using the revised model. In healthcare, the FL model shines in predicting illnesses using data sources with complicated pattern training, such as COVID-19 (Samuel et al. 2022) because recent advances in FL lead to fog computing (FC), intelligent devices can now collaborate with the server in the cloud (Celesti et al. 2020).

4.3 Artificial integration in healthcare: exclusion of blockchain

AI integration in healthcare without Blockchain can improve diagnostics, personalized medicine, and operational efficiency. However, challenges like data security, privacy, and interoperability persist. Ensuring patient information integrity is crucial, and the absence of Blockchain may hinder the establishment of a secure, decentralized data infrastructure, potentially limiting the seamless sharing and trustworthiness of healthcare data across different entities and systems. Table 3 discusses the areas of AI applications specific to healthcare, its challenges, and the real-world implementation considerations.

Table 3 AI in healthcare: application, challenges, and implementation considerations
Full size table

4.4 Artificial intelligence’s potential hazards in the healthcare system

Common errors that cause artificial intelligence to fail are as follows:

4.4.1 Using the incorrect data

Despite the enormous progress with the adoption of AI specific to healthcare, there are several significant challenges, especially data-related ones (Banerjee and Chanda 2020). Data must be precise and error-free before it can be used in AI applications. Moreover, inadequate and incorrect data for training and testing AI models may lead to bias in algorithms, especially unequal distribution of classes (FitzGerald and Hurst 2017, Norori et al. 2021a, b). For instance, an AI model trained on white-skinned patients may not be as accurate when applied to patients of different races. An incorrect diagnosis made by an AI-powered diagnostic system might have fatal repercussions.

AI is not always unbiased, and even slight prejudice can impact outcomes. Researchers use data augmentation to expand image access, but this doesn't always ensure strong validation (Monshi et al. 2021). To prevent bias, AI should be trained using diverse data, which can be achieved using Blockchain (Chen et al. 2018; Zheng et al. 2019). This combination could reduce bias and improve future prediction models.

4.4.2 Security

Data in healthcare include clinical reports, scientific studies, patient health records, and diagnostic details (Churi et al. 2021). They have been created daily in large volumes (especially due to mobile devices) during the past decade and must be stored securely for the future. Since information is precious, it must not only be stored securely but also transferred safely. Data breaches can be a common problem while transferring data (Dilmaghani et al. 2019). In these situations, Blockchain’s strong and dependable storage can safely convey data for AI training models and prediction (Zhang et al. 2021a, b).

4.4.3 Data aggregation and automation

AI algorithms can never be relied on completely unless they are first constructed and then trained on a substantial amount of relevant and diverse data (Lee et al. 2017). However, accessing high-quality clinical datasets is challenging due to strict regulations and protections. Hospitals have limited access to EHRs due to the Food and Drug Administration (FDA) and the Health Insurance Portability and Accountability Act (HIPAA) (Shuaib et al. 2021). EHR exchange across databases requires additional work to ensure compliance. This has led to ongoing discussions on who can access protected health information. Acquiring and using a database with thousands of specific photos can be challenging, especially when working with uncommon ailments. Blockchain can help AI in data aggregation and update the network, accordingly, making AI more reliable in treating rare diseases (Dillenberger et al. 2019).

4.4.4 Privacy

Privacy and patient data problems are important considerations involving patient data gathering (Tith et al. 2020). Researchers have safeguards in place to protect patient data, but unscrupulous hackers still try to gain access to the data (Yampolskiy and Spellchecker 2016). AI’s ability to predict patient information even when the algorithm was not provided with such data also threatens patients’ privacy. Recently, the AIIMS reported that the entire network got compromised, and it was suspected that the ransomware assault affected the data of three to four hundred million patients (Mohurle and Patil 2017). Using Blockchain authorized people can have a copy of the information on their node and can get access to the data as and when needed and this can avoid this type of situation (Feng et al. 2019).

5 Synergies unveiled: the interdependence of artificial intelligence and blockchain technologies

This exploration deconstructs their symbiotic connection, highlighting magnified strengths for game-changing advances in security, transparency, and decentralized intelligence uncovering the collaborative future of technology.

5.1 Blockchain for artificial intelligence, decentralized computing in healthcare

The fusion of Blockchain with AI results in secure data sharing and marketplaces for AI systems. Combining Blockchain and AI has obvious benefits and provides decentralized computing for AI (Dinh and Thai 2018). Blockchain protects data, enables us to audit all intermediary steps, and even allows users to monetize their data (Chen et al. 2019; Jennath et al. 2020). Blockchain-enabled data marketplaces promote secure data incorporation from healthcare participants and vendors, enhancing AI training. Blockchain is essential for AI and other applications such as the financial market (Zhang et al. 2020), the Internet of Everything (IoE) (Singh et al. 2020; Alrubei et al. 2021), edge computing (Fan et al. 2020a, b; Shen et al. 2020; Manogaran et al. 2021), and fog computing (Qu et al. 2020). To protect the IoE, an access management system architecture has been proposed (Bera et al. 2020). So, it is evident that Blockchain can assist AI in becoming more independent, trustworthy, and intelligent. (Dinh and Thai 2018; Zhang et al. 2021a). Blockchain and AI are combining to create secure data sharing and marketplaces for AI systems. This decentralized computing approach protects data, audits intermediary steps, and allows users to monetize their data. Blockchain-enabled data marketplaces promote secure data incorporation from healthcare participants and vendors, enhancing AI training. Blockchain is essential for AI applications in financial markets, IoE, edge computing, and fog computing. To protect IoE, an AI-based Blockchain-envisioned access control architecture has been proposed. This fusion of Blockchain and AI can help AI become more independent, trustworthy, and intelligent, making it a valuable tool for various applications.

5.2 Artificial intelligence for blockchain in healthcare

By combining AI with Blockchain, the flaws of these technologies may be effectively corrected. Blockchain creates a safe and transparent distributed personal data marketplace by enabling secure data exchange between organizations (Mamoshina et al. 2018). On the other hand, AI technology enables the privacy-preserving personalization of patient records (Suo et al. 2018). The combination has been used to protect healthcare data privacy (Singh et al. 2022a, b), securely transfer imaging data (Orhan et al. 2021), a donor organ transplant system (Morande and Marzullo 2019), and even in healthcare workflow in a telemedical laboratory (Celesti et al. 2020), cardiovascular medicine (Krittanawong et al. 2020), novel coronavirus disease 2019 self-testing (Mashamba-Thompson and Crayton 2020). Secure and scalable Blockchains can be developed by interdependency on AI. Web 3.0 uses machine learning, AI, and Blockchain to improve human communication (Leeming et al. 2019). Figure 18 gives an overview of the reliance of AI on Blockchain and Blockchain on AI in various application objectives. This dependence may be beneficial to the healthcare system by reducing the amount of rework and reconciling making it an immensely powerful tripod for the future.

Fig. 18
figure 18

Dependency of AI and Blockchain: Blockchain for AI, and AI for Blockchain

Full size image

The combination of AI with Blockchain is essential because it builds a strong basis for the future of technology. This cooperation not only improves decision-making security and transparency but also instills confidence through verifiable and immutable records. Dynamic decision-making combined with data integrity develops an accountability culture, while distributed intelligence assures operational efficiency across decentralized networks. The anti-tampering and privacy-protection features address crucial issues, offering a safe foundation for sensitive data.

This integration streamlines processes, maintains dependability, and reshapes the landscape of industries, offering a future where trust, security, and efficiency meet effortlessly. Table 4 provides insights into AI and Blockchain properties and the astonishing benefits that result from their synergistic combination. The study explores the intricate interplay of various technologies, concentrating on their potential to improve integrity, openness, and efficiency in decision-making processes, thereby laying the foundation for a competent future.

Table 4 Blockchain and AI integration: features and synergistic benefits
Full size table

6 Blockchain integration with AI-enabled healthcare: the tripod

The fusion of AI and blockchain technology specific to healthcare is transforming the healthcare process, enhancing clinical operations, and service intensity, and managing predictive activities. This fusion reduces healthcare-related risks, making it a future healthcare tripod. This summary covers 51 articles on the integration of AI and blockchain in healthcare applications, highlighting their impact on diagnoses, treatment planning, data security, and administration.

6.1 Blockchain feature specifications in AI-enabled healthcare studies

Blockchain is indeed the technology that powers cryptocurrencies (King and Nadal 2012; Narayanan et al. 2016) and is being used in AI-enabled healthcare studies to securely transfer healthcare-related data and maintain patient privacy (Mantey et al. 2021). Table 5 summarizes the usage of blockchain features considered in AI-enabled healthcare studies. Smart contracts, which combine immutability, transparency, and decentralization, are a crucial feature in data authorization, particularly beneficial in healthcare for enhanced efficiency and transparency, making them a significant improvement in any industry (Zou et al. 2019).

Table 5 Features of blockchain and its applications in AI-enabled healthcare articles
Full size table

Various AI-enabled blockchain attributes are essential for accelerating data-specific AI application development. Figure 19 explores the fusion of AI-enabled Blockchain features in healthcare, highlighting its privacy and security benefits. Key attributes include immutability, availability, integrity, and authentication, while AI attributes like accuracy and prediction are crucial.

Fig. 19
figure 19

Blockchain features for AI-enabled healthcare system

Full size image

6.2 AI-enabled blockchain applications and case studies in healthcare: enhancing healthcare operations and patient outcomes

In the ever-changing environment of healthcare, the fusion of these technologies has sparked novel solutions. This section delves into AI-enabled blockchain applications and case studies, demonstrating their critical role in improving healthcare operations and maximizing patient outcomes.

6.2.1 Pandemic: detection and diagnosis to monitoring and prediction

COVID-19 initially posed significant challenges to the healthcare system due to infrastructure deficiencies, including a lack of integrity, immutability, and audit framework (Hamze 2021; Musamih et al. 2021). AI algorithms combined with blockchain technology enable real-time analysis of healthcare data, facilitating early detection, diagnosis, and monitoring of infectious diseases. To address these issues, the integration of Blockchain and AI has been crucial, enhancing the integrity and immutability of healthcare data, and ensuring a secure and transparent framework (Meghla et al. 2021; Kamenivskyy et al. 2022). This integration streamlines clinical operations by enabling timely interventions and containment measures, ultimately improving patient outcomes and public health, particularly in response to the COVID-19 pandemic. In the Rahman et al. (2020) study, the application focused on classifying Internet of Health Things (IoHT) data related to COVID-19 using Deep Neural Networks (DNN) within a Consortium blockchain. This integration aimed to ensure authentication, privacy protection, and data integrity, thereby enhancing clinical operations and patient outcomes. Similarly, Warnat-Herresthal et al. (2021) applied DNN-Swarm Learning and RNA sequencing in a private blockchain setting to predict and analyze COVID-19, addressing concerns of confidentiality.

Moreover, studies such as Muhammad and Hossain (2021) emphasized data privacy and security through Edge computing and Convolutional Neural Network (CNN) models. These efforts streamline clinical operations by ensuring the confidentiality and integrity of patient information, eventually fostering better patient health outcomes. Additionally, Tanwar et al. (2021) explored the use of long short-term memory (LSTM) techniques and other AI methodologies to enforce social distancing measures, enhancing security within healthcare environments.

Further, collaborations like Bera et al. (2021) utilized support vector machines (SVM) and RCNN models within Public blockchain frameworks to monitor COVID-19 patients in home isolation, ensuring security and trust. Meanwhile, efforts by Kumar et al. (2021a, b) and Mohsin et al. (2021) in Private blockchain environments leveraged machine learning and FL methods to identify COVID-19 patients using CT scans and to share and update COVID-19 data safely and reliably, respectively, thereby enhancing security and privacy. Similarly, other studies have revealed that by ensuring data security, privacy, and integrity, these technologies can enhance diagnostic accuracy, treatment efficacy, and patient care delivery. This approach can help healthcare organizations respond to COVID-19 challenges and establish a more patient-centric future.

A total of 16 research articles from 2020 to 2023 assess the fusion of AI and blockchain in healthcare, focusing on COVID-19-related situations, across various applications. Table 6 shows a comprehensive examination of AI and blockchain fusion in healthcare for combating COVID-19.

Table 6 Applications and case studies presenting AI-enabled Blockchain integration in healthcare for combating COVID-19
Full size table

6.2.2 Blockchain integration with AI: data redundancy and augmentation specific to COVID-19

Data Redundancy is a common problem in healthcare and we are aware that manually collected patient records can have redundancy and duplication (Gupta et al. 2019). The medical imaging field faces significant challenges due to the limited number and variety of samples in small datasets (Saba et al. 2019). Deep learning networks like CNN require large data for training, leading to data augmentation during the COVID-19 pandemic for improved detection using random and duplicated datasets (Monshi et al. 2021). Data duplication, a technique used in AI frameworks (Sanagala et al. 2021) aids in generalizing AI solutions and preventing overfitting problems (Agarwal et al. 2021a, b).

Blockchain can help prevent overfitting across multiple classes by enhancing data accuracy through data augmentation techniques like flipping or rotating images (Tian et al. 2021). Based on the findings of the European Parliamentary Research Service, Blockchain is regarded as one of the most important current technologies for COVID-19 (Mihalis 2020). During an epidemic, inadequate counting and data gathering were prevalent, exacerbated by late private clinic reports and non-automated processes (Kalla et al. 2020). Adopting Blockchain promotes trust across people, organizations, governments, and continents. Providers and users send anonymized patient data via the Blockchain paradigm (Zhuang et al. 2020). Countries can use private Blockchain solutions or government health databases to send daily test results and other information, but both must maintain hash values (Nakamoto 2008). Blockchain technology can be effectively used to monitor and combat infectious diseases like COVID-19, and also for other applications like patient data exchange, and contact tracing (Kalla et al. 2020).

As the Blockchain network allows only tamper-proof data sharing (Cheng et al. 2020), this may result in data exchange safeguarded by Blockchain and further advance AI (Bhattacharya et al. 2019; Kumar et al. 2021a, b). Therefore, enhanced AI may result in increased data security and efficiency (Wang et al. 2019a, b). The study aims to improve healthcare models by combining Blockchain with AI, resulting in (i) superior training (ii) increased accessibility of data required for AI enhancement, and (iii) the distribution of exclusive AI generalization algorithms among healthcare providers, resulting in a decentralization of datasets (Kumar et al. 2022a, b, c, d) thereby enhancing its functionality and effectiveness by this superior tripod design.

AI and blockchain technologies are revolutionizing healthcare by improving security, interoperability, and intelligence in managing health data. AI safeguards medical information confidentiality, while blockchain analyzes large datasets for critical insights. Research shows these technologies significantly enhance privacy and security in EHR and PHR records.

6.2.3 Case studies demonstrate the elimination of falsification and forgery through AI and blockchain

Blockchain integrated with AI algorithms eliminates the falsification and forgery of medical records by ensuring the authenticity and confidentiality of healthcare data. The focus of research conducted by Kim and Huh (2020) is on improving anonymity and enhancing the sensitivity, availability, and security of AI in medical records. Their objective was to ensure the authenticity of medical records, including EMRs, PACS data, and PHRs by preventing forgery and fabrication. They achieved this by utilizing a mini-batch dataset, and randomly selecting some training data to aid in the learning process. This integration streamlines clinical operations by improving data accuracy, enhancing patient safety, and fostering trust in healthcare systems, ultimately leading to better patient outcomes.

6.2.4 Case studies demonstrating mitigation of data breaches, manipulation, and privacy through AI and blockchain

Traditional databases, relying on a single point of failure, are susceptible to hacking and data breaches. In contrast, the blockchain’s distributed network disperses data across multiple nodes, thwarting malicious tampering. Empirical case studies demonstrate how using these technologies can preserve privacy, guard against security lapses, and safeguard confidential information across a range of industries, including healthcare.

Kumar et al. (2022a, b, c, d) presented a study focusing on enhancing attack detection and preventing data breaches within industrial healthcare systems by integrating permissioned blockchain and smart contracts. The utilization of cryptographic techniques and consensus mechanisms inherent in blockchain technology establishes a tamper-resistant and transparent ledger of transactions. This enhances the detection of malicious activity and prevents unauthorized alterations to sensitive healthcare data. Additionally, the implementation of smart contracts serves as automated protocols governing data access, sharing, and authentication, thereby reducing the risks associated with unauthorized breaches or fraudulent activities. While the study provides a comprehensive approach to enhancing security and integrity within industrial healthcare systems, scalability and comprehensive comparisons with existing methods remain areas for further exploration and refinement.

Furthermore, Alzubi et al. (2022) introduced a privacy-enhancing system that amalgamates DL, blockchain technology, and FL to address concerns surrounding EHR privacy and security. By leveraging CNN and FL techniques, the study aimed to achieve high accuracy rates, scalability, data privacy, and detection of malicious activity within healthcare systems. A key contribution of the study lies in its proposed model designed to identify abnormal users and restrict their database accessibility, thus preventing the risk of unauthorized access and data breaches to sensitive EHRs. The model utilizes DL algorithms to analyze user behavior patterns and identify deviations indicative of potential malicious intent or abnormal activity. With blockchain fusion, the system ensures the immutability and transparency of access logs, providing a comprehensive audit trail for monitoring user interactions with EHRs. The integration of FL further enhances privacy protection by enabling decentralized model training across multiple data sources thereby avoiding the need for aggregated data in a central location. This approach ensures that sensitive patient data remains localized and encrypted.

Lakhan et al. (2023) addressed the challenges faced by dynamic Internet of Medical Things (IoMT) systems in healthcare applications, particularly regarding data fraud in distributed environments. The study introduced an innovative framework that employs dynamic heuristics named FL-BETS (FL-based Blockchain-enabled Task Scheduling) to balance load, manage energy consumption, and meet deadlines across cloud and fog nodes. FL-BETS aims to ensure the integrity and privacy of data while minimizing energy consumption and delay, thus optimizing healthcare workload management. The study evaluates FL-BETS against existing blockchain and FL mechanisms, demonstrating superior performance in energy efficiency, data validation, fraud analysis, and meeting healthcare application constraints. Similarly, Reegu et al. (2023) demonstrated that blockchain-based EHR systems reduce data breaches compared to traditional centralized databases and also minimize the risk of cyberattacks. As we continue to explore innovative approaches, the synergy between blockchain and AI will undoubtedly shape a more secure and resilient digital landscape for the future.

6.2.5 Patient privacy and data security through AI and blockchain

The fusion of AI with blockchain technology appears to be a viable option for improving data security in healthcare systems (Wang et al. 2019a, b). Researchers have carefully investigated novel strategies to improve data security, notably in the realms of (EHRs/PHRs) (Jennath et al. 2020). AI algorithms serve as key enablers in enhancing data security by proactively identifying and mitigating potential threats. While AI algorithms are adept at evaluating large volumes of data and recognizing patterns suggestive of security problems, they are also vulnerable to manipulation and exploitation by malicious actors (Yampolskiy and Spellchecker 2016). Furthermore, AI algorithms may struggle to ensure the integrity and immutability of data, making them insufficient as standalone solutions for data security (Jensen et al. 2020). The flaw in relying solely on AI for data security lies in its vulnerability to adversarial attacks and limitations in ensuring data integrity and tamper-proofing (Hu et al. 2021a, b).

Blockchain technology addresses these limitations and enhances data security in several ways. First and foremost, blockchain provides a decentralized and immutable ledger system that protects data integrity and authenticity (Hepp et al. 2018). As we know blockchain’s transactions are cryptographically linked, resulting in a tamper-resistant record of data exchanges (Funk et al. 2018). The immutability of blockchain data makes it extremely difficult for malevolent people to change or manipulate records without detection (Bathula et al. 2022a, b). Furthermore, blockchain’s decentralized architecture removes one point of failure and lowers the danger of unwanted access, enhancing the overall resilience of data storage and management systems (Noveck 2011). Because of both technologies' complementary qualities, combining AI with blockchain is critical for improving data security in healthcare and other domains (Lu et al. 2021; Ali et al. 2023). While AI algorithms excel at evaluating data and detecting security risks, blockchain provides a safe and open platform for storing and distributing sensitive data. By combining AI with blockchain, enterprises can use AI’s analytical powers to uncover possible security breaches and abnormalities in real-time, while blockchain provides data integrity, transparency, and immutability (Vyas et al. 2019). This integration allows for safe and privacy-preserving data exchange, auditable transactions, and proactive threat detection, thus enhancing overall data security and trust in the digital ecosystem (Mamoshina et al. 2018).

The fusion of AI and blockchain increases data security by merging AI’s analytical skills with blockchain’s safe and transparent data management architecture, making it an essential option for protecting sensitive information in today’s digital era (Chen et al. 2018; Abou El Houda et al. 2023).

6.2.6 Case studies demonstrating patient privacy and data security specific to EHR /personal health records PHR)

In the realm of healthcare, the research on the fusion of AI and blockchain technologies underscores the substantial enhancement of privacy and security in EHR and PHR records.

Several researchers offered cutting-edge frameworks and models to emphasize the privacy and security of patient data such as EHR/ PHR in this thorough analysis of advances in safe healthcare (Jennath et al. 2020). They provided a patient-driven data-sharing paradigm that prioritizes privacy without disclosing personally identifying information and is particularly designed for Secure Personal Health and Care Records (PHCR). Alruwaili (2020) proposed a hybrid strategy combining AI-based intelligent agents and BC to protect EHR databases. Bhattacharya et al. (2019) proposed Blockchain-based DL as a service for exchanging EHR information about patients with diabetes, and obesity, and improves the security of EHRs by developing cryptographic authentication and signature techniques. The authors Kumar et al. (2020) proposed a method for choosing miners using supervised learning and AI technology to maintain fairness in a healthcare-based system for mining data. Al-Safi et al. (2022) in their research, offered a decentralized method to preserve patient privacy in medical data using blockchain technology and utilizing an AI algorithm for classification and accuracy. Furthermore, Alzubi et al. (2022) provide a privacy-preserving EHR paradigm that identifies typical users and limits database access. Sai et al. (2023) investigates secure smart healthcare diagnostics by offering an NFT marketplace for patient-controlled healthcare data, utilizing IPFS in FL, and solving privacy problems in cloud-based solutions.

6.2.7 Case studies demonstrating chronic disease management

AI-powered predictive analytics combined with blockchain technology enabled personalized treatment plans and proactive interventions for chronic disease management. This integration streamlines clinical operations by providing timely interventions.

6.2.7.1 Prediction of lung cancer

In their paper Kumar et al. (2021a, b), the authors present a framework integrating DL and Blockchain for decentralized data learning in the context of Lung cancer prediction. For safe real-time data sharing, they made use of a customized smart contract. Zerka et al. (2020) called this chained distributed machine learning C-DistriM, which is a revolutionary distributed learning approach that blends sequential distributed learning with a Blockchain-based platform. It created a model with high-quality data that can be verified for integrity.

6.2.7.2 Detection of myopic macular degeneration and high myopia.

In their study, Tan et al. (2021) developed retinal photograph-based DL algorithms and assessed for the diagnosis of myopic macular degeneration and extreme myopia using a retrospective multicohort investigation. Also, they highlighted how deep learning algorithms and Blockchain are leveraged to increase transparency, auditability, and security.

6.2.7.3 Predicting diabetes

In their work El Rifai et al. (2020), the authors proposed a diabetic predicting solution that addresses private data leakage in FL algorithms by integrating public Blockchain with FL and DL techniques. They demonstrate their methodology using a predictive decision support tool for a dataset (diabetes), emphasizing the importance of protecting sensitive health information.

6.2.8 Case studies demonstrating dietary guidance and customized medical care

AI-driven dietary guidance and customized medical care programs empower patients to make informed lifestyle choices and adhere to personalized treatment plans. Blockchain technology ensures secure data sharing between patients and healthcare providers, streamlining clinical operations and improving patient outcomes through personalized care plans.

The integration of blockchain and AI technologies, as exemplified by the work of Mantey et al. (2021), has profound implications for streamlining clinical operations and enhancing patient outcomes, particularly in the realm of dietary guidance and customized medical care. Here, blockchain is used by the system to guarantee the privacy and reliability of health data, while AI algorithms analyze individual health information, tailor diet advice, and deliver targeted medication notifications. This combination empowers individuals with specialized care based on their health profiles.

6.2.9 Case studies demonstrating telesurgery or robotic surgery

The fusion of blockchain, AI, and telesurgery signifies a groundbreaking era in surgical innovations. Gupta et al. (2020a, b) introduced a self-managed, secure, transparent, and trustable system for telesurgery or robotic surgery, revolutionizing the healthcare industry by enhancing accuracy, safety, and technology.

6.2.10 Case studies demonstrating clinical/genomic/biomedical research

Kuo et al. (2020) in their study have developed a model ensemble with hierarchical consensus, using Blockchain-based distribution and level-wise learning. Their findings show that this architecture improves prediction accuracy while preserving privacy. Mamoshina et al. (2018) use AI and Blockchain to decentralize biomedical research, encouraging continuous health monitoring and accelerating research.

6.2.11 Case studies demonstrating IoMT (Internet of Medical Things) and AI application

The papers discuss the fusion of AI and Blockchain on the IoMT. They focused on developing a healthcare-based system for neural network training, decentralizing healthcare processes through Blockchain-based FL, and personalized FL, and improving healthcare model efficiency, demonstrating the blockchain’s and AI’s potential for IoMT applications.

Samuel et al. present an IOMT privacy architecture (Samuel et al. 2022), highlighting the efficiency of AI-enabled big data analytics in safely examining patient data connected to COVID-19. A two-stage FL system for IoMT devices is presented in Lian et al. (2022, 2023), which makes use of blockchain to improve training accuracy and privacy over a dispersed network suggests the FL-BETS framework for healthcare, which protects patient privacy and makes fraud detection possible. A decentralized IoMT learning paradigm for real-time patient monitoring is presented in Khan and AbaOud (2023), along with privacy-focused measures including blockchain authentication and homomorphic encryption. A method for integrating AI into IoMT and improving data privacy for Secure PHCR is put forward (Połap et al. 2020). Kalapaaking et al. (2023), they focused on enhancing the security of healthcare ML data by using privacy-preserving strategies to secure a worldwide healthcare ailment. Jin et al. (2021) suggested a cross-cluster FL framework through a cross-chain approach (CFL) to cope with privacy leakage and data sparsity issues while ensuring IoMT system efficiency.

6.2.12 Vaccine distribution

The fusion of distributed ledger technology with AI might potentially give the optimal platform for immunization in supply chain management, demand predictions, and vaccine distribution can be further optimized and streamlined (Rahman et al. 2021). Blockchain technology helps distribute vaccines by securely and transparently recording transactions (Abbas et al. 2020). It also tracks the movement of vaccines from manufacturers to healthcare providers and recipients (Kamenivskyy et al. 2022). Blockchain technology can ensure the accurate handling and preservation of vaccine supply information by connecting information silos controlled by suppliers, manufacturers, distributors, and medical professionals (Das 2021). The system utilized AI to monitor quality standards, production parameters, and Blockchain data for distribution optimization, trend identification, and vaccination demand forecasting (Antal et al. 2021).

Blockchain and AI can also be used to create a comprehensive framework for vaccine distribution, as illustrated in Fig. 20. The World Health Organization (WHO), manufacturers, suppliers, distributors, national stores, doctors, hospitals, pharmacies, and individual patients participate in vaccination deployment using Blockchain technology (Kazancoglu et al. 2022). Blockchain nodes verify information on vaccinations, connecting blocks that precede and follow it (Sunny et al. 2020). Post-WHO approval, manufacturers supply vaccines to distributors. Blockchain nodes ensure transparency, linking verified vaccination data through cryptographic hash functions (WHO 2020). These systems automate distribution, ensuring timely delivery.

Fig. 20
figure 20

An overall framework for vaccine distribution using Blockchain and AI

Full size image

ML is used to evaluate vast quantities of data from the Blockchain to forecast the demand for a certain vaccine in a specific location (Das et al. 2021; Meghla et al. 2021). Early in the pandemic, COVID-19 vaccine delivery was challenging because the system infrastructure lacked integrity, immutability, and audit framework (Hamze 2021; Musamih et al. 2021). Blockchain and AI have significantly enhanced the accuracy of COVID-19 result forecasts and vaccine traceability by enabling big data analysis on Blockchain (Meghla et al. 2021; Kamenivskyy et al. 2022). The vaccine distribution was made more efficient and transparent by Blockchain, ensuring (i) traceability and a thorough audit of the supply chain, (ii) storage, and (iii) delivery conditions (Antal et al. 2021).

This fusion of these technologies across various applications and its outcomes is summarized in Table 7.

Table 7 Comprehensive table summarizing case studies, applications, and outcomes of AI and blockchain fusion in healthcare
Full size table

6.3 Enhancing healthcare efficiency with AI-blockchain

The fusion of AI and blockchain technologies presents a revolutionary opportunity to enhance healthcare efficiency and effectiveness on multiple fronts. By leveraging AI algorithms to analyze extensive datasets stored securely on blockchain platforms, healthcare systems can realize significant improvements in assessment metrics and diagnostic accuracy, ultimately leading to streamlined processes and better patient outcomes.

Firstly, the combination of AI and blockchain harnesses the power of big data analytics in ways previously unattainable by healthcare providers. AI algorithms excel at processing and interpreting vast amounts of healthcare data, including electronic health records, medical imaging scans, genomic sequences, and real-time monitoring data from wearable devices (Baucas et al. 2023). By accessing this wealth of information stored on blockchain platforms, AI systems can identify patterns, trends, and correlations that might otherwise go unnoticed. This comprehensive analysis empowers clinicians to make more informed decisions, tailor treatments to individual patient needs, and optimize healthcare delivery pathways. For instance, a study by Tripathi et al. (2024) showed how AI might evaluate big datasets to enhance diagnosis precision and provide customized treatment suggestions. Similarly, blockchain’s decentralized nature allows for secure, tamper-proof storage of medical data, ensuring that AI systems have access to reliable and consistent information for analysis (Wu et al. 2022).

Moreover, the immutable and transparent nature of blockchain technology ensures the integrity and security of healthcare data throughout its lifecycle (Hölbl et al. 2018). By storing data in tamper-proof blocks distributed across a decentralized network, blockchain platforms provide a robust framework for safeguarding sensitive patient information from unauthorized access, tampering, or data breaches (Sonkamble et al. 2023). This heightened level of data security instills trust among patients, healthcare providers, and stakeholders, facilitating greater collaboration and information sharing across the healthcare ecosystem.

Furthermore, AI algorithms can leverage the structured and standardized data stored on blockchain platforms to develop more accurate assessment metrics and diagnostic tools. Jennath et al. (2020) conducted an in-depth analysis of longitudinal patient records, genetic profiles, treatment outcomes, and population health trends. They found that AI-driven systems can detect risk factors, early warning signs, and predictive biomarkers associated with various diseases and conditions. This predictive analytics capability enables healthcare professionals to proactively intervene, monitor disease progression, and personalize treatment plans to optimize patient outcomes. A study by Durga and Poovammal (2022) highlighted how AI, when combined with blockchain technology, could enhance predictive analytics for chronic diseases, allowing for early interventions and improved patient management strategies.

Thus, the fusion of AI and blockchain technology implies a paradigm change in healthcare delivery. Healthcare systems may improve patient care throughout the continuum by providing precise evaluation metrics and more accurate diagnoses through the analysis of massive amounts of data stored on blockchain platforms. The future of healthcare will be provided with unparalleled opportunities for innovation, efficiency, and impact as we continue to realize the synergistic potential of AI and blockchain.

6.4 Summary of the intersection of blockchain and various AI techniques in healthcare

6.4.1 Machine learning-enabled blockchain technologies in healthcare

Table 8 summarizes research works aimed at integrating Machine Learning-enabled Blockchain Technologies into healthcare applications. The research spans from 2020 to 2022 and addresses a wide range of topics, including telesurgery, patient data exchange, COVID-19 analytics, and vaccination delivery. AI approaches, illness categorization algorithms, and data integrity by smart contracts on the blockchain are among the main contributions to these studies. However, there are constraints, insufficient empirical validation, and ethical issues about data privacy and algorithmic bias in the published papers. Some research lacks clarity on AI methodology, dataset specifications, and consensus processes, stressing the importance of more complete, transparent, and ethically based ways to successfully use ML and Blockchain Technologies in healthcare.

Table 8 A Summary of the published research on the application of Machine Learning-enabled Blockchain Technologies in Healthcare
Full size table

6.4.2 Deep learning-enabled blockchain technologies in healthcare

The use of DL-enabled Blockchain Technologies in healthcare applications, particularly private blockchains like CNN and DNN, is prevalent. However, limited disclosure of consensus algorithms and smart contract implementations highlights the need for increased transparency. Challenges include dataset size limitations, scalability concerns, and ethical considerations. These observations underscore the evolving landscape of integrating DL and Blockchain in healthcare as shown in Table 9. It emphasizes the necessity for comprehensive reporting and addressing practical implementation challenges for future advancements.

Table 9 A Summary of the published research on the application of Deep Learning-enabled Blockchain Technologies in Healthcare
Full size table

6.4.3 Federated learning -enabled blockchain technologies in healthcare

The fusion of FL and blockchain represents a significant advancement in addressing critical challenges such as privacy risks and data reliability inherent in collaborative ML frameworks. Qammar et al. (2023) explore the intersection of Blockchain technology and FL in smart healthcare to tackle challenges in decentralized data environments. Their systematic review synthesizes current literature to show how Blockchain’s decentralized ledger can mitigate privacy risks, unreliable model uploads, and high communication costs inherent in FL systems. Through decentralized coordination, communication, and adopting Blockchain’s distributed consensus mechanisms, the authors proposed solutions that enhance data privacy and reliability in FL implementations for healthcare. The study discusses recent advancements and ongoing challenges that require further research to fully exploit Blockchain’s potential in FL integration. Similarly, Myrzashova et al. (2023) contribute to this discourse by systematically reviewing the literature on blockchain-enabled FL, emphasizing its potential to decentralize healthcare applications, manage dispersed clinical data, and bolster security and privacy in digital healthcare systems. They highlight blockchain’s role in safeguarding sensitive medical information and optimizing collaborative machine learning across distributed networks. Their analysis supports blockchain adoption to address and meet evolving demands for secure and privacy-preserving healthcare data management and analysis.

Furthermore, Zhu et al. (2023) conducted a systematic survey investigating the integration of blockchain technology with FL to address challenges inherent in decentralized machine learning environments. Their study classifies current integration models into three main architectures namely coupled, decoupled, and overlapped, systematically evaluating their impacts on scalability, data privacy, and security within FL systems. The survey underscores blockchain’s potential to enhance FL by leveraging its decentralized and immutable ledger, effectively mitigating vulnerabilities associated with centralized coordination and ensuring transparent consensus mechanisms. Technical challenges such as blockchain network scalability, optimization of consensus algorithms, and seamless interoperability with FL frameworks are meticulously analyzed. The study contributes by proposing innovative solutions and outlining future research directions aimed at advancing blockchain-enabled FL systems across diverse applications.

Figure 21 illustrates the integrated architecture of FL with blockchain technology. FL participants begin by receiving an initial model and updates to the model based on their respective datasets. These updates are transmitted to miners via APIs like REST, JSON-RPC and gRPC. Miners, serving as nodes in the blockchain network, validate and aggregate these updates using consensus algorithms such as (PoW or PoS). Smart contracts enforce transparent agreements among participants, managing tasks like registration, coordination of training, and reward distribution. Verified blocks containing aggregated updates are appended to the blockchain, allowing FL clients to securely access and download the finalized global model. This architecture enhances collaboration and transparency in FL processes, leveraging blockchain’s decentralized nature for improved model training across distributed participants.

Fig. 21
figure 21

Architecture integrating federated learning with blockchain technologies. gRPC gRPC Remote Procedure Call, RESTful API Representational State Transfer (REST), JSON-RPC JavaScript Object Notation Remote Procedure Call

Full size image

6.4.4 Heterogeneous federated learning and isomerism learning with blockchain

In the realm of collaborative ML, Heterogeneous Federated Learning (HFL) addresses the diversity of devices and data (Yu et al. 2020). Emphasizing the integration of HFL with blockchain, particularly through methodologies like Isomerism Learning, holds considerable potential for enhancing research quality.

6.4.4.1 Isomerism learning, blockchain, and artificial intelligence

Isomerism learning harmonizes FL’s adaptive characteristics with blockchain’s security features, paving the way for robust and privacy-preserving collaborative ML frameworks (Aich et al. 2021; Abou El Houda et al. 2023). By incorporating Isomerism Learning into the research framework, the study can capitalize on intelligent collaboration among entities to dynamically tackle time-varying challenges in decentralized collaborative learning. Hao et al. (2023) introduced an innovative approach to analyzing public sentiment in social media, considering various factors such as geography, politics, and ideology. Their study emphasizes advancements in model architecture, incorporating techniques like embedding tables and gating mechanisms to enhance performance. Furthermore, the authors proposed isomerism learning, a new distributed deep learning model that utilizes blockchain technology for secure collaboration and parallel training across distributed nodes. This method dynamically adjusts model weights based on event objectivity, leading to notable improvements in experimental performance evaluations. The integration of blockchain ensures secure data sharing and enhances scalability in collaborative learning environments, highlighting its potential to reshape decentralized ML methodologies.

Biscotti, introduced by Shayan et al. (2020) is a decentralized peer-to-peer (P2P) scheme based on blockchain technology. It emphasizes maintaining privacy and security among FL peers by securely exchanging secrets using verifiable random functions (VRFs). The framework addresses challenges such as Sybil attacks and data poisoning by authenticating and verifying each peer’s contributions within the decentralized network. Biscotti utilizes a blockchain-inspired approach with off-chain storage and adopts a private blockchain model organized around Merkel trees. It also features a consensus mechanism optimized for integration with Hyperledger Fabric, enhancing resilience and confidentiality in collaborative ML environments.

BlocFL, introduced by Lei et al. (2023), merges FL with blockchain by replacing the central server with a consortium blockchain. This enables local training of multiple neural networks across various nodes. Researchers have proposed a blockchain-based client selection method to optimize resource allocation, particularly with nodes of varied capabilities. The method showed similar accuracy to the baseline and significantly improved resource utilization, enhancing system scalability and efficiency. In contrast, GCFL (Graph with Coordinator FL), as presented by Ying et al. (2023), merges FL with a DAG blockchain to minimize data redundancy and facilitate seamless data exchange across devices. GCFL introduces a Two-Phase tips selection consensus algorithm, which significantly reduces resource usage and enhances stability compared to traditional FL systems. This innovative approach effectively improves FL performance while ensuring data privacy and scalability. It surpasses the limitations of single chain blockchains and conventional consensus methods, providing a robust solution for the evolving needs of decentralized ML. Institutional and healthcare collaboration (IFL) occurs securely by sharing model updates on the blockchain, enabling accurate predictions from diverse data. Isomerism Learning adapts the global model to various data distributions, ensuring accuracy. overcoming challenges like scalability, energy consumption, and algorithmic complexities is crucial for fully integrating IFL with blockchain. Nguyen et al. (2021a, b) and Lu et al. (2019) have highlighted these challenges, emphasizing the need for ongoing research endeavours to refine methodologies, optimize system architectures, and develop novel algorithms. Advancing the integration of HFL with blockchain, particularly through approaches like Isomerism Learning, holds promise for fostering innovation and enabling transformative applications across various domains.

Table 10 summarizes key findings on integrating FL-enabled Blockchain Technologies in healthcare, focusing on COVID-19 identification, fraud detection, and real-time patient monitoring. It highlights Consortium and private blockchains, local gradients, and FL algorithms used. Challenges include scalability and privacy safeguards, emphasizing ongoing refinement needed. Therefore, FL within the tripod framework of AI, healthcare, and blockchain technologies exemplifies a transformative approach in collaborative ML. By integrating FL with blockchain, FL ensures data privacy, enhances collaboration, and drives innovation. As technological advancements continue and regulatory frameworks evolve, FL is poised to revolutionize ML applications, offering economic benefits and advancing data-driven solutions globally.

Table 10 A Summary of the published research on the application of Federated Learning-enabled Blockchain Technologies in Healthcare
Full size table

6.4.5 Machine learning and deep learning -enabled blockchain technologies in healthcare

Table 11 presents an overview of research efforts combining ML and DL with blockchain technologies in healthcare applications. Notably, we found that the research includes a broad range of applications from 2020 to 2023, including securing patient health records, clinical information systems, COVID-19 detection, nutritional recommendations, and illness categorization. While many ML and DL approaches are used, such as DNN, CNN, and transfer learning, the common drawbacks across research include a lack of understanding of blockchain implementation specifications, consensus algorithms, and smart contract details. Concerns concerning suggested frameworks' generalizability, scalability, and real-world application appear as recurring themes. The fusion of AI and blockchain can improve healthcare systems, but further study and precise reporting are required to solve present limits and ensure success.

Table 11 A Summary of the published research on the application of Machine Learning and Deep Learning-enabled Blockchain Technologies in Healthcare
Full size table

6.4.6 Deep learning and federated learning-enabled blockchain technologies in healthcare

Table 12 provides a thorough analysis of the usage of DL and FL with blockchain technology in the healthcare field. The research, spanning 2020–2023, will focus on COVID-19 detection, illness prediction, and healthcare model efficiency. Common DL approaches include CNN, ResNet, and CapsNets, while FL methods aim to protect privacy in IoMT and tailored systems. Significant contributions include enhanced data privacy, secure medical imaging diagnostics, and enhanced ML data security. While stressing data protection and easy integration into the IoMTs, research on blockchain implementations frequently lacks definition, leaving critical information on types, smart contracts, and consensus methods unknown. Common restrictions include real-world implementation obstacles, legal frameworks, and scalability issues. Clear analysis of blockchain features is critical for establishing successful and scalable healthcare solutions in this changing context.

Table 12 A Summary of the published research on the application of Deep Learning and Federated Learning-enabled Blockchain Technologies in Healthcare
Full size table

6.4.7 Reinforcement learning-enabled blockchain technologies in healthcare

The research explores the fusion of Reinforcement Learning (RL) with Blockchain technology in the healthcare industry as shown in Table 13, focusing on applications like EHRs, COVID-19 services, and safe data offloading. It uses multi-agent systems and deep reinforcement learning algorithms to improve data protection, secure healthcare services, and optimize medical data offloading. However, uncertainties, lack of AI approach specifics, and insufficient knowledge of Blockchain technology are common issues. More research is needed to improve RL and Blockchain integration in healthcare.

Table 13 A Summary of the published research on the application of Reinforcement Learning-enabled Blockchain Technologies in Healthcare
Full size table

6.4.8 Explainable AI-enabled blockchain technologies in healthcare

Table 14 provides an overview of Explainable Artificial Intelligence (XAI) integrated with Blockchain Technologies in healthcare applications. The study from 2023 focuses on smart and secure healthcare facilities using XAI methods like Grad CAM/LIME. This trend aligns with the AI community’s focus on transparency and interpretability. XAI’s usage in healthcare is expected to grow, indicating a shift towards more accountable and understandable AI systems. However, the study highlights ongoing challenges and the potential for continued advancements in this dynamic field.

Table 14 A Summary of the published research on the application of Explainable AI-enabled Blockchain Technologies in Healthcare
Full size table

6.5 Fusion of AI-enabled blockchain performance metrics in healthcare

Table 15 provides a thorough overview of research publications examining the convergence of AI and blockchain technology in healthcare, displaying a broad spectrum of applications and addressing important performance criteria year-wise. The research evaluates AI metrics like accuracy and AUC, along with blockchain measures like security and privacy. It draws attention to concerns about security and confidentiality, as well as the use of permissioned and public blockchains and the investigation of cutting-edge technology like NFTs for secure medical diagnostics. Many studies emphasize the need for real-time scalability, FL, interpretability, and energy efficiency. Blockchain plays an important role in ensuring trust, accountability, and trust. AI-Blockchain integration is revolutionizing global collaboration in medical learning processes, reshaping healthcare operations like a future Tripod.

Table 15 A Summary of the published research on the Fusion of AI and Blockchain with performance metrics in healthcare
Full size table

7 Critical discussion

7.1 Comparative and comprehensive analysis

This narrative review offers a thorough overview of blockchain-based AI healthcare applications considering articles published between 2005 and 2023. It covers various healthcare domains, including infectious disease, COVID-19, e-Health, EHR/PHR management systems, and healthcare supply chains. The review explores advanced solutions to Blockchain-based healthcare challenges using AI-powered techniques. It examines the interdependence of various AI methodologies, including Deep Learning (DL), Machine Learning (ML), Reinforcement Learning (RL), Federated Learning (FL), and Explainable AI (XAI). The review explores the potential of the integration of Blockchain in AI healthcare, highlighting its multifaceted impact and potential. It serves as a point of reference for various applications, emphasizing the growing integration of Blockchain in AI healthcare to ensure data privacy and security, establishing it as a resilient tripod for the future.

7.2 Major findings

Derived from the integration and implementation of AI and Blockchain technology in the medical sector, we have drawn certain findings in our review. We have arrived at the following main findings based on these observations and challenges, which are covered in Sects. 2, 3, 4, 5, and 6:

  1. (i)

    Applications: The features of Blockchain, smart contracts, security, and privacy are widely addressed in a variety of applications such as securing EHRs, combating COVID-19, and e-health. This is shown in Table 5. Various types of blockchains exist, with public and private blockchains emerging as the predominant choices for healthcare applications.

  2. (ii)

    Blockchain architecture and its working: although many papers have been written on the integration of Blockchain into healthcare applications, none have described how Blockchain architecture works and how Blockchain technologies, such as Ethereum and Hyperledger, connect with healthcare applications. Figure 11 depicts the detailed architecture of layers with its examples.

  3. (iii)

    Blockchain solution for healthcare applications: Current Blockchain technology issues that might jeopardize the system’s fundamental nature and gave recent efficient Blockchain solutions, demonstrating this technology a dependable answer to all Blockchain-based healthcare applications. It is presented in Table 2.

  4. (iv)

    Data challenges in healthcare AI: leveraging blockchain solutions:

In healthcare AI, data collection is a significant challenge. AI model training demands extensive, unbiased, high-quality datasets for effective development. Recent healthcare applications increasingly leverage Blockchain in Sect. 6.2 for securing data transmission and records, emphasizing its role in overcoming data challenges.

  1. (v)

    Decentralized computing using blockchain and AI: Blockchain and AI interdependence can help both the technologies with decentralized computing and scalable Blockchains. The detailed dependency is given clearly in Table 4.

  2. (vi)

    AI in healthcare challenges and blockchain solutions: Table 3 under Sect. 4 highlights challenges in real-time AI implementation in healthcare, with data privacy and security at the forefront. Blockchain emerges as a robust solution, enhancing the diagnostic process by ensuring heightened privacy and security.

  3. (vii)

    Power of privacy and security in tripod design: the power of Tripod is dependent on the many AI-enabled Blockchain features, especially privacy, and security, which are essential for speeding up the deployment of data-specific AI applications. From Table 15 it is evident that the most widely used metrics in implemented integration papers of AI and blockchain are security and privacy in addition to AI metrics such as accuracy and AUC.

  4. (viii)

    AI and blockchain in healthcare applications: Fig. 6 shows that AI and Blockchain are primarily used in healthcare applications for securing EHRs and COVID-19 data transmission, accounting for 39% and 30% of published articles from 2017 to 2023.

  5. (ix)

    Hypothesis

  6. (x)

    Validation: The most prominent AI assessment metrics for the research under consideration are accuracy and performance, showing that it relies on reliable and unbiased data, requiring Blockchain’s involvement. This proves that our idea of combining Blockchain with AI may enhance healthcare organizations, making our tripod dependable and efficient.

7.3 Claims

  1. (i)

    Rise in fusion of AI and blockchain publications in healthcare: Our first claim from Fig. 3a illustrates that there is a substantial rise in the publication of articles from 2017 to 2023, ranging from 1 to 24 articles, underscores the increasing recognition of AI adopting blockchain in healthcare. This trend affirms the pivotal role of this fusion as a potent tripod shaping the future landscape of healthcare research and applications.

  2. (ii)

    Use of AI techniques and rise of FL and DL in healthcare: Our first claim from Fig. 5a is proof that the adoption of FL and the combination of FL and DL started rising in the healthcare industry too, and it can be used in future research because holds the capability to address limitations inherent in methodologies dependent on a singular centralized data pool. This, therefore, allows us to be able to provide large-scale precision medicine, leading to models that give unbiased judgments. Figure

  3. (iii)

    Use of blockchain type and rise of private blockchain in healthcare: Fig. 4a, shows the proof that private blockchain is widely used in the healthcare industry, with 20 of the 51 AI and blockchain fusion implemented articles utilizing this type of application. However, 14 articles selected public blockchain, demonstrating the general inclination toward private blockchain when integrating AI in healthcare.

  4. (iv)

    Implementation of AI and blockchain in healthcare: We found that though Blockchain is mostly adopted in integration with AI in selected articles, most of the articles just mention their usage without detailing their implementation. Figure 3b presents proof that the number of articles not implemented is 28 out of the total 79 articles published. However, Fig. 3c indicates a positive trend, showing a rise in practical implementation from 2020 onwards.

7.4 A special note on datasets

The narrative evaluation requires a specific mention of the sorts of datasets used, particularly device-based datasets. Numerous prominent sensor-based datasets have been developed by researchers. Table 16 outlines the different datasets used in healthcare with Blockchain and AI integration with attributes such as (i) data source, (ii) dataset quantity employed for the study, (iii) dataset magnitude, (iv) device used, and (v) type of data.

Table 16 Different datasets are used with the integration of Blockchain and AI in healthcare
Full size table

This research explores the use of various datasets in various applications, including blockchain applications in EHR record sharing, real-time data, and X-ray COVID-19 data categorization. The studies examined in this research employed diverse datasets, with some relying on a single dataset and others utilizing multiple datasets.

In 2021, X-ray COVID-19 and COCO played crucial roles in COVID-19 screening and social distancing. In 2022, ToN-IoT/IoT-Botnet and MNIST/CIFAR-10 (Jennath et al. 2020; Kim and Huh 2020) datasets focused on secure data exchange and privacy-preserving healthcare. In 2023, CDC data, Medical MNIST/Tissue MNIST/OCTMNIST (Bhattacharya et al. 2019; Połap et al. 2020), and Monkeypox addressed COVID-19 detection, healthcare efficiency, and patient privacy issues. MNIST, CIFAR-10, CC-19/COVID-19-CT, MIMIC-III (Johnson et al. 2016), and 2-D Colon Pathology/Breast Tumor datasets (Fleurence et al. 2014), were frequently used, demonstrating their adaptability in applications ranging from privacy protection to infectious disease diagnostics (Kumar et al. 2021a, b). Tuberculosis Chest X-ray Image Data Sets Heart Disease UCI (Jaeger et al. 2014). All this data requires security and privacy while sharing or gathering, and Blockchain was used to accomplish this. Some datasets are publicly available, requiring minimal prior approval due to confidentiality. Devices for data collection include CCTV cameras, smartphones, smart wearable devices, CT scanners, hospitals, patients, and edge nodes. Blockchain was employed to ensure the security and privacy of sharing and gathering these data.

8 Future work plan

Expanding advanced features of AI and their potential interfacing with Blockchain in the future.

  1. (i)

    Pruning for artificial intelligence storage reduction: Pruning involves removing unnecessary nodes or filters from deep learning (DL) models to minimize their size and enhance their prediction time (Xu et al. 2019). If the prediction time decreases, the DL models will be suitable for usage in edge devices, particularly in health care (Oguntola et al. 2018). The storage requirements of the DL model are the major focus of the pruning process, and its final goal is to make the model more storage-friendly (Choudhary et al. 2020). Weight pruning (Han et al. 2015), neuron pruning (Srinivas and Babu 2015; Skandha et al. 2022a, b), filter pruning for CNN (Li et al. 2016), and layer pruning (Chen and Zhao 2018) are a few pruning techniques. There has yet to be seen AI application with Blockchain using pruning strategies.

  2. (ii)

    Explainable artificial intelligence (XAI): Explainable AI helps to comprehend and interpret machine learning model predictions, and using it can debug, enhance, and explain model behavior completely (Chalkiadakis 2018; Linardatos et al. 2020). In healthcare, clinicians must comprehend the functionality and workings of black-box models, including their decision-making processes, to validate the reasoning behind the model’s conclusions (Tjoa and Guan 2020; Zhang et al. 2022). The recent adoption of transparent AI-powered systems indicates a promising future trajectory towards fully explainable AI, creating opportunities for further developments in XAI.

8.1 Strengths, weakness, and extensions

This review is a trailblazing attempt that presents and explores the integration of blockchain and AI in various healthcare applications, including COVID-19 and EHR/PHR, providing a comprehensive overview of the field. Emphasized existing blockchain technology issues that might ruin the system’s core nature and presented recent efficient blockchain solutions hence proving this technology’s reliability for Blockchain-based AI healthcare applications. We have discussed the blockchain architecture which can give an understanding of blockchain frameworks integrated with healthcare applications. Thorough and comparative research was conducted throughout the benchmarking phase. A significant amount of work was done in this healthcare study using the fusion of Blockchain and AI, but some limitations and drawbacks need to be addressed.

  1. (i)

    Each of these proposed frameworks and architectures has its benefits, but none has yet used blockchain technology to help AI be used in a decentralized and fully distributed way in healthcare systems.

  2. (ii)

    Many articles have developed and addressed AI implants, but few have covered the full implementation of the Blockchain architecture. Blockchain technology is being applied in the healthcare sector, where special issues such as security must be managed.

  3. (iii)

    The main issue associated with this modern technology in medical institutions is a lack of knowledge and implementation cost.

There are several aspects which can be considered for the future. Note that the outcomes of AI are mostly determined by the data, however, there are several key concerns, such as user privacy and security. The answer to these issues is Blockchain distributed architecture with transparent data. The public Blockchain and FL combination has been one of the most popular ways in the previous three years. Finally, the use of FL and DL in private Blockchain framework has shown remarkable results in healthcare. But there is truly little work done on DRL and consortium Blockchain, which can be the next revolution in the healthcare domain.

9 Ethical implications and societal impacts of AI and blockchain in healthcare

The fusion of blockchain and AI in healthcare is promising for improved efficiency and patient care. However, ethical, and societal impacts must be carefully examined to ensure (i) responsible deployment, (ii) safeguard patient rights, and (iii) promote equitable access to quality care.

9.1 Ethical implications

9.1.1 Algorithmic bias and fairness

The fusion of blockchain with AI algorithms in healthcare raises ethical concerns about algorithmic fairness and bias. Blockchain’s transparency in data transactions is pivotal, yet ensuring AI algorithms provide equitable treatment to all patient groups remains essential. Bias is important to study when it comes to AI in healthcare (Suri et al. 2021a, b; Das et al. 2022; Paul et al. 2022). Ethical guidelines are necessary to mitigate biases, promote fairness, and prevent discriminatory outcomes in healthcare decision-making. In the healthcare landscape, data’s pivotal role underscores concern regarding the bias in acquiring accurate data for training AI models (Estiri et al. 2022). AI bias encompasses data or algorithmic bias and societal AI bias. Algorithmic bias results from biases in training data or algorithm design, potentially leading to disparities in healthcare services (Panch et al. 2019; Kumar et al. 2024). Societal AI bias reflects societal intolerance or institutional prejudice (Norori et al. 2021). Addressing bias in healthcare AI involves proper feature extraction, crucial for mitigating bias, and effective performance assessment methods (FitzGerald and Hurst 2017).

Employing appropriate techniques and methods helps minimize bias in healthcare AI applications, ensuring fair and equitable healthcare delivery (Suri et al. 2021a, b, 2022a). Moreover, by using augmentation methods like the Adaptive Synthetic Sampling Method (ADASYN) and (Synthetic Minority Over-sampling Technique) SMOTE, the balanced data architecture may reduce bias brought on by unbalanced data (Alex et al. 2024). By creating artificial data points for the minority class, these techniques even out the distribution of classes within the dataset. By deliberately employing these augmentation techniques, AI systems may learn from a more balanced representation, lessening the influence of bias.

9.1.2 Transparency and accountability

Blockchain technology serves as a powerful tool in fostering transparency and accountability in the utilization of AI algorithms (Zhou et al. 2024). The unalterable ledger of blockchain assures that every transaction and activity carried out by AI algorithms is forever recorded and cannot be changed (Wang et al. 2019a, b). Blockchain enhances accountability, openness, and fairness in decision-making processes by providing a complete record of AI activity, enabling stakeholders to evaluate and analyze data to detect and prevent bias in AI systems (Lo et al. 2022). The system’s decentralized nature minimizes centralized manipulation and promotes fairness by distributing decision-making data across multiple nodes for transparent recording.

9.1.3 Automation with smart contracts

Smart contracts automate the enforcement of established rules and criteria without the need for human interaction, hence preventing bias (Omar et al. 2021). They use fairness principles and ethical rules to detect and rectify biases, eliminating subjective human judgment and reducing the likelihood of biased decisions. Blockchain’s Audibility features enable comprehensive auditing, tracing biased decisions back to their source. A study by Bose et al. (2024), explored addressing bias using blockchain smart contracts and the results show that blockchain contracts help maintain accurate data, which is crucial for reducing bias. Overall, blockchain enhances transparency, accountability, and audibility in AI decision-making.

9.1.4 Data ownership and patient consent

It is critical in both AI and blockchain implementations to guarantee that patients own their health data and are fully informed about its collection, storage, and use. Blockchain technology presents a unique opportunity to improve data ownership and patient consent management (Mann et al. 2021). Patients may take control of their health data by managing access rights, keeping an eye on data consumption, and adopting blockchain-based decentralized platforms (Jabarulla and Lee 2021a, b). Moreover, clear standards for the collection, storage, and usage of data should be established, and patients should be informed about their data usage. Sophisticated permission management systems that enable patients to quickly grant, cancel, or monitor their consent for the use of their data while maintaining their autonomy can be provided using blockchain-connected AI systems (Tith et al. 2020).

9.1.5 Equity in access to healthcare services

The integration of AI and blockchain technology in healthcare has the potential to greatly improve equity in access to advanced healthcare services by decentralizing them, decreasing inequities, and increasing transparency and trust (Till et al. 2017). AI-powered telemedicine can provide high-quality treatment to rural and underprivileged populations, while blockchain assures safe, transparent data sharing and empowers patients with access to their health data (Abugabah et al. 2020). Personalized treatment plans and AI-optimized resource allocation guarantee that varied populations receive individualized care (Frank and Olaoye 2024). Addressing digital gaps, algorithmic bias, and data privacy is crucial for ethically implementing AI technologies in healthcare. Inclusive development, robust infrastructure, and continuous monitoring of AI algorithms are essential for achieving fair healthcare outcomes. Additionally, investing in digital infrastructure and providing education can help bridge the digital divide, ensuring that all patients can benefit from the advancements in AI and blockchain healthcare technologies (Ragnedda and Destefanis 2019).

9.1.6 Data privacy and security

The integration of AI and blockchain technologies requires robust measures to protect patient confidentiality and prevent unauthorized access to sensitive healthcare information, ensuring patient data privacy and security (Zhang et al. 2021a, b). The implementation of encryption techniques, strict access control policies, and secure data storage on blockchain networks are crucial for ensuring privacy and security (Kumar et al. 2022a, b, c, d). Regular audits, data minimization practices, and employee training are essential for mitigating risks. Compliance with regulations like HIPAA and GDPR is crucial for upholding patient rights and privacy (Bakare et al. 2024). These measures protect patient data confidentiality and maintain trust in AI and blockchain technologies.

9.2 Societal impacts

9.2.1 Enhanced healthcare accessibility and equity

The fusion of AI and blockchain technology can bridge regional barriers and eliminate healthcare inequities by enabling telemedicine, remote consultations, decentralized networks, AI-driven resource allocation, and personalized healthcare solutions (Guo et al. 2019). This democratizes access to high-quality services, ensuring everyone, regardless of location or socioeconomic status, has access to necessary treatments and interventions (Yang et al. 2020).

9.2.2 Empowerment of patients barriers and healthcare consumers

The empowerment of patients and healthcare consumers is one important social impact. Blockchain technology provides healthcare consumers with transparent access to their health information and decision-making processes, empowering patients, and consumers (Hang, Kim et al. 2021). Patients are now more equipped to design individualized treatment plans and make informed decisions. Increased satisfaction with care and better health outcomes are likely to be experienced by patients due to the shift toward patient-centered care (Abbas et al. 2020).

9.2.3 Transformation of healthcare workforce and skills

This fusion of AI and blockchain technology transforms the healthcare workforce and skill sets (Sousa and Rocha 2019). Healthcare personnel must adapt to new technologies and workflows made by AI automation and blockchain data management solutions. This transition includes initiatives to upskill and reskill healthcare workers so that they can effectively employ new technologies to offer high-quality care. Furthermore, new roles such as AI specialists, blockchain developers, and data analysts arise, transforming the healthcare workforce and creating new opportunities.

9.2.3.1 Ethical and regulatory considerations

Adhering to strict regulations on data protection, security, and compliance creates challenges for deploying AI and blockchain technology in healthcare (Al Meslamani 2023). It has significant societal implications, including protecting health data privacy and security, preventing bias in AI algorithms, and upholding accountability in AI-driven decision-making (Chen et al. 2018; Stephanie et al. 2023). To address these challenges, regulatory frameworks must change, including creating guidelines for platform interoperability and data exchange and ensuring adherence to data protection regulations like GDPR (Shuaib et al. 2021).

9.2.3.2 Economic implications

There are significant financial implications when using blockchain and AI in healthcare. These technologies have the power to improve resource allocation, save costs, and streamline healthcare delivery processes (Munoz et al. 2019). They also promote entrepreneurship and innovation in the healthcare industry, which leads to the creation of new goods, services, and business ideas (Pergher et al. 2016). Additionally, the development of jobs and economic growth and prosperity are facilitated by investments in blockchain and AI technologies.

9.2.3.3 Cultural and social change

The widespread use of AI and blockchain in healthcare is transforming the industry, fostering innovation and cooperation among practitioners, researchers, and technology (Zhang et al. 2021). It also promotes patient involvement in decision-making and democratizes healthcare information, empowering individuals to take control of their health and well-being (Jabarulla and Lee 2021a, b). However, realizing the complete capabilities of these technologies necessitates addressing ethical, regulatory, and cultural considerations to ensure their deployment in a socially accountable and equitable manner.

10 Conclusion

In conclusion, the fusion of blockchain and AI technologies presents a transformative opportunity to revolutionize the healthcare sector. Our analysis underscores the growing adoption of blockchain for securing EHRs, enhancing data privacy, and facilitating secure data transmission in healthcare. Concurrently, the rise of federated learning and deep learning techniques highlights AI’s potential to drive precision medicine and personalized healthcare solutions. The preference for private blockchain implementation underscores the healthcare sector’s commitment to data security and privacy, reflecting a broader shift towards real-world integration and deployment of AI and blockchain technologies.

Blockchain’s ability to secure data transfer complements AI’s need for secure storage, facilitating increased data security, improved service efficiency, and enhanced accessibility in healthcare. By granting healthcare stakeholders access to patient medical records on the blockchain, this fusion can streamline procedures, potentially saving billions for the industry while ensuring enhanced security, privacy, and accessibility. The convergence of blockchain and AI offers solutions across healthcare domains, including infectious diseases, COVID-19, e-Health, and EHR/PHR management, with the potential to significantly enhance disease identification, response, and overall healthcare efficacy. The future tripod of blockchain, AI, and healthcare promises to improve patient outcomes, and address key challenges in the healthcare sector.

Looking ahead, exploring advanced AI features such as pruning and explainable AI (XAI) integrated with blockchain holds promise for further enhancing healthcare outcomes. As these technologies evolve, their collaborative potential will continue to shape the future of healthcare delivery worldwide, guaranteeing data security and privacy while driving unprecedented innovation and efficiency.