Home // eTELEMED 2020, The Twelfth International Conference on eHealth, Telemedicine, and Social Medicine // View article

An Approach to Explainable AI for Digital Pathology

Authors:
Juan Manuel Montes-Sanchez
Luis Muñoz-Saavedra
Francisco Luna-Perejon
Javier Civit-Masot
Satur Vicente-Diaz
Anton Civit

Keywords: DigitalPathology;ExplainableArtificialIn-teligence; Deep learning

Abstract:
Many medical diagnostics are based, at least, in parton medical imaging. The development of machine learning and,in particular Deep Learning (DL) based image processing in thelast decade has led to the growth of diagnostic support aidsbased on these technologies. A problem regarding the adoptionof this systems the lack of understandability of their diagnosticsuggestions due to their Blackbox nature. Several approacheshave been proposed to increase their explainability includingevaluation of the internal layer contributions to outputs, networkmodifications to make these contributions more meaningful andmodel agnostic explanations. Medical systems are consideredthe paradigmatic case where understandability is of outmostimportance. Digital Pathology (DP) is an especially difficult, butespecially interesting case for image based diagnostic supportaids. This is due, among other factors, to the fact that DP imagesare very large and multidimensional with the information noteasily available at first sight. It is important to develop toolsthat let the pathologists apply their available knowledge easilywhile improving the diagnostic quality and their productivity.The design and evaluation of an interpretable digital pathologydiagnosis aid would open the possibility for developing anddeploying larger scale systems that would provide pathologistswith reliable and trustworthy tools to help them in their dailydiagnosis tasks.

Pages: 9 to 10

Copyright: Copyright (c) IARIA, 2020

Publication date: March 22, 2020

Published in: conference

ISSN: 2308-4359

ISBN: 978-1-61208-763-4

Location: Valencia, Spain

Dates: from November 21, 2020 to November 25, 2020