Digital image analysis and artificial intelligence in pathology diagnostics—the Swiss view

Digital pathology (DP) is increasingly entering routine clinical pathology diagnostics. As digitization of the routine caseload advances, implementation of digital image analysis algorithms and artificial intelligence tools becomes not only attainable, but also desirable in daily sign out. The Swiss Digital Pathology Consortium (SDiPath) has initiated a Delphi process to generate best-practice recommendations for various phases of the process of digitization in pathology for the local Swiss environment, encompassing the following four topics: i) scanners, quality assurance, and validation of scans; ii) integration of scanners and systems into the pathology laboratory information system; iii) the digital workflow; and iv) digital image analysis (DIA)/artificial intelligence (AI). The current article focuses on the DIA-/AI-related recommendations generated and agreed upon by the working group and further verified by the Delphi process among the members of SDiPath. Importantly, they include the view and the currently perceived needs of practicing pathologists from multiple academic and cantonal hospitals as well as private practices.

In early 2021, the Swiss Digital Pathology Consortium (SDiPath) initiated a Delphi process to generate best-practice recommendations for various phases of the process of digitization in pathology for the local Swiss environment, encompassing the following four topics: (i) scanners, quality assurance, and validation of scans; (ii) integration of scanners and systems into the pathology laboratory information system; (iii) the digital workflow; and (iv) digital image analysis (DIA)/artificial intelligence (AI).As digital pathology (DP) is increasingly integrated into routine clini-cal pathological diagnostics by academic and privatepathologyinstitutes, setting up national recommendations to standardize procedures and support implementation was perceived as a current need [1][2][3][4].This includes the implementation of DIA algorithms and AI tools, as they have become increasingly attainable and desirable in daily sign out.
Herein, we briefly outline the recommendations related to the validation and use of DIA and AI as it applies to the algorithm-aided extraction of information from histological images to support clinical decision making.The recommendations generated were drafted by a dedicated DIA/AI working group and were subsequently verified by members of SDi-Path (N = 25) using a Delphi process.All members of SDiPath (N > 170) were encouraged to participate.Importantly, the recommendations include the view and the currently perceived needs of practicing pathologists from multiple academic and cantonal hospitals as well as private practices, who comprised 76% of the responding participants.The rest of the participants were scientists, IT personnel, and laboratory staff.We refer the reader to the original guideline publication for an in-depth discussion of the recommendations [5].

Compliance with general quality guidelines
The SDiPath guidelines affirm that DIA/AI intended for diagnostic use needs to be designed in compliance with current laws and regulatory requirements for medical devices, i.e., only officially certified systems (e.g., IVD-CE certified, FDA-approved) or lab-developed systems properly validated with respect to quality control and quality assurance can be applied.Also, and equally important, it is the boardcertified pathologist who determines and is legally responsible for the finally rendered diagnosis.If DIA/AI output is used without being approved by a board-certified pathologist, this is considered offlabel use.In line with the existing legal requirement to store all glass slides the diagnoses were generated on, we recommend storing the results of the DIA/AI analysis to allow diagnosis retracing.All steps should be documented in a standard operating procedure (SOP) document.
An internal validation of the system on relevant specimens mirroring the realworld clinical use is recommended, even if the system is officially certified (e.g., IVD-CE certified, FDA-approved) to ensure consistency of performance.Documentation of metadata of the scans, the software used, and the validation protocol as well as the performance of the software (reproducibility, intra-observer variability, and known problems regarding performance) should be provided.When validating the system, a concordance testing needs to be carried out and revalidation shall be performed whenever changes are made in the workflow.

Workflow considerations
Quality control steps verifying that the scans are suitable for analysis and that all relevant areas have been analyzed are required.User training, personnel, and technical requirements, as well as SOPs for hardware and software malfunctions and management updates, need to be provided.In-depth information about the type and application of the DIA/AI system used in the diagnostic process (including quality control measures and validation steps) needs to be included into the pathology report.There are also multiple technical properties that were agreed on to be paramount.In short, the DIA/AI systems need to be well integrated into the pathology workflow environment and automated, in order to enable streamlining of the diagnostic process in the era of increasing workload.
Importantly, for the pathologist, it is most desirable to have visual control of the algorithm's results, and this should highlight the regions which determined its output.Understanding how the system is generating output and decisions becomes increasingly more critical as the level of autonomy of these systems will rise in the future.It is suggested that the systems should enable a feedback loop to improve performance.
As a last comment, finding parallels with the multiple comments regarding where "Back to the Future II" was wrong in predicting current everyday life and technology [6,7], we appreciate that the technological possibilities and standards as well as knowledge in the field of digital pathology are developing rapidly.The current recommendations, published elsewhere in full length [5], have been generated to facilitate current development and implementation and shall be reviewed and updated as technology advances.For this article no studies with human participants or animals were performed by any of the authors.All studies mentioned were in accordance with the ethical standards indicated in each case.

Prof
The supplement containing this article is not sponsored by industry.
Open Access.This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made.The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material.If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder.To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Declarations Conflict of interest
. V.H. Koelzer reports being an invited speaker for Sharing Progress in Cancer Care (SPCC) and Indica Labs; advisory board of Takeda; sponsored research agreements with Roche and IAG, all unrelated to the current study.S. Berezowska, G. Cathomas, R. Grobholz, M. Henkel, W. Jochum, M. Kreutzfeldt, K.D. Mertz, M. Rössle, D. Soldini, I. Zlobec and A. Janowczyk declare that they have no competing interests.