Suśruta: Artificial Intelligence and Bayesian Knowledge Network in Health Care – Smartphone Apps for Diagnosis and Differentiation of Anemias with Higher Accuracy at Resource Constrained Point-of-Care Settings

Abstract

Anemia in India carries a major disease burden. This includes both nutritional anemias, of which iron deficiency anemia (IDA) is the commonest and inherited hemolytic anemias like β thalassemias (β-TT). In Eastern and North-Eastern India about 25–40% of the public blood bank blood is consumed by patients with β Thalassemia. Moreover, 51% of Indian women in the reproductive age suffer from IDA. Most vulnerable group of anemic patients is in the rural underserved regions. This underserved population can only be served through the use of artificial intelligence (AI), automation, supported by telemedicine. To combat the problems of both IDA and β-thalassemia by early diagnosis at the point-of-care– we have developed Suśruta - an Artificial Intelligence (AI) driven robust smartphone-based health care application. This App uses five major components of AI; namely: (1) Natural language processing (NLP) to analyze the unstructured clinical data and translate it into computer understandable 3^rd Generation SNOMED and ICD10 ontologies; (2) Speech Synthesis; (3) Artificial Neural Network (ANN) with Machine Learning and Deep Learning (ML/DL) on 60,283 labelled common blood counts (CBC) and High Performance Liquid Chromatography (HPLC) data collected over 8 years by a teaching hospital in Kolkata for β-TT screening; (4) Computer Vision and Image Processing techniques to interpret hemoglobin content in blood through non-invasive analysis of conjunctiva and nailbed images; and (5) NoSQL and Big-data Graph database-driven Bayesian Knowledge Network for Evidence Based Medicine and Bayesian Outcome Tracing for Predictive Medicine. Unlike previous systems, the ML/DL technique of β-Thalassemia carrier screening with CBC improved the accuracy of screening by two folds compared similar approaches analyzing CBC with Mentzer Index. Moreover, the uniqueness of Suśruta is that the App is robust and works both in an offline and online mode at resource constrained point-of-care. This is the first time AI is used for comprehensive anemia care by early diagnosis, which empowers the ordinary health workers in rural underserved communities. Furthermore, it will introduce the concept of patient empowerment and person centered care by changing the definition of point-of-care in rural India.