Abstract
Calcium Pyrophosphate Deposition Disease (CPPD) is a common crystal-induced arthropathy whose radiographic assessment is challenging due to subtle calcific deposits, inter-grade ambiguity, and the ordinal nature of disease severity. Conventional deep learning methods typically formulate severity estimation as a standard classification or regression task, overlooking both the progressive transition between adjacent grades and the structure of the learned feature space. We propose a prototype-based representation learning framework for automated CPPD severity assessment from knee radiographs, where a learnable prototype-based classifier replaces the conventional linear classification head. To account for the ordinal progression of disease, the model is trained using a Soft-Target Cross-Entropy loss with ordinal soft labels. Experiments on a retrospective dataset of 2,172 knee radiographs annotated into four CPPD severity grades demonstrate that the proposed framework achieves the best overall performance, reaching a QWK of 82.24 ± 0.71, WF1 of 63.01 ± 1.73, MF1 of 58.91 ± 2.01, accuracy of 62.95 ± 1.54, and MAE of 0.415 ± 0.017 over 10 random seeds. Feature-space analysis further reveals improved clustering quality and more structured representations than the best baseline. These findings demonstrate the effectiveness of combining prototype-based representation learning with ordinal supervision for robust and clinically meaningful CPPD severity assessment.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/RHEUM_AI_003.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: https://papers.miccai.org/miccai-2026-sat/supp/RHEUM_AI_003_supp.pdf
Link to Open Review
Open Review Page: https://openreview.net/forum?id=aoXtikiaYb
BibTex
@InProceedings{FedLor_Prototypical_MICCAISAT2026,
author = { Federici, Lorenzo AND Fiorentino, Maria Chiara AND Rosati, Riccardo AND Zingaretti, Primo},
title = { { Prototypical Severity Representation Learning for Automated Calcium Pyrophosphate Deposition Disease Assessment } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17271},
month = {pending},
page = {pending}
}