Improving prototypical parts abstraction for case-based reasoning explanations designed for the kidney stone type recognition.

Fiche publication

Date publication

septembre 2025

Journal

Artificial intelligence in medicine

Auteurs

Membres identifiés du Cancéropôle Est :
Pr DAUL Christian

Tous les auteurs :
Flores-Araiza D, Lopez-Tiro F, Larose C, Hinojosa S, Mendez-Vazquez A, Gonzalez-Mendoza M, Ochoa-Ruiz G, Daul C

Lien Pubmed

https://www.ncbi.nlm.nih.gov/pubmed/40992030

Résumé

The in-vivo identification of the kidney stone types during an ureteroscopy would be a major medical advance in urology, as it could reduce the time of the tedious renal calculi extraction process, while diminishing infection risks. Furthermore, such an automated procedure would make possible to prescribe anti-recurrence treatments immediately. Nowadays, only few experienced urologists are able to recognize the kidney stone types in the images of the videos displayed on a screen during the endoscopy. This visual recognition by urologists is also highly operator dependent. Thus, several deep learning (DL) models have recently been proposed to automatically recognize the kidney stone types using ureteroscopic images. However, these DL models are of black box nature and do not establish the relationship of the visual features they used to take the decision with the color, texture and morphological features visually analyzed in biological laboratories to determine the type of extracted kidney stone fragments using the reference morphoconstitutional analysis (MCA) procedure. This contribution proposes a case-based reasoning DLmodel which uses prototypical parts (PPs) and generates local and global descriptors. The PPs encode for each class (i.e., kidney stone type) visual feature information (hue, saturation, intensity and textures) similar to that used by biologists during MCA. The PPs are optimally generated due a new loss function used during the model training. Moreover, the local and global descriptors of PPs allow to explain the decisions ("what" information, "where in the images") in an understandable way for biologists and urologists. The proposed DL model has been tested on a database including images of the six most widespread kidney stone types in industrialized countries. The overall average classification accuracy was 90.37±0.6%. When comparing this results with that of the eight other DL models of the kidney stone state-of-the-art, it can be seen that the valuable gain in explanability was not reached at the expense of accuracy which was even slightly increased with respect to that (88.2±2.1%) of the best method of the literature. These promising and interpretable results also encourage urologists to put their trust in AI-based solutions.