The artificial intelligence challenge in rare disease diagnosis: a case study on collagen VI muscular dystrophy

Abstract

© 2025 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).

The use of artificial intelligence (AI) techniques is significantly changing the analysis of medical images, accelerating and standardizing the diagnosis process. To train an AI model, however, a large dataset is typically required, especially when using the most powerful techniques. Therefore, not all specialties are taking advantage of AI techniques in the same way. For instance, they are seldomly used in areas such as the diagnosis of rare diseases since, due to their low prevalence, not enough data are typically available to train an AI model. In this paper, we address the use of AI techniques to diagnose a particular rare disease: Collagen VI-related Congenital Muscular Dystrophy from confocal microscopy images. We apply both classical machine learning and modern deep learning techniques and we show that, when using the appropriate data management and training procedures, one can successfully derive a highly-accurate classifier even with a limited amount of training data. Due to the generality of the explored techniques, this conclusion is likely to hold also for most of the rare diseases whose diagnosis relies on the examination of histological images.

This work was supported by the European Union: HORIZON–MSCA–2022–DN, Improving BiomEdical diagnosis through LIGHT-based technologies and machine learning ‘‘BE-LIGHT’’ (GA n◦ 101119924 – BE-LIGHT); and the Instituto de Salud Carlos III, Spain (PI22/01382), FEDER (A way of making Europe) and Fundación Noelia.

Peer Reviewed

Postprint (published version)