Abstract
Unsupervised anomaly detection (UAD) provides a powerful framework for identifying pathological regions in medical imaging without relying on annotations, thereby overcoming key limitations of supervised approaches. Although UAD methods ranging from classical reconstruction-based models to recent diffusion architectures have been extensively investigated in adult neuroimaging, their application to pediatric populations, which represent a more challenging setting for UAD exhibiting greater anatomical variability than adults, remains largely understudied. In this work, we present a comprehensive benchmark evaluation of diverse UAD frameworks applied to pediatric neuroimaging. We systematically evaluate models performance and limitations across a heterogeneous spectrum of abnormalities, including rare and subtle cases. Finally, to encourage research and address the scarcity of openaccess pathological pediatric data, we release a curated dataset of 2D T1-weighted brain MRI slices covering diverse abnormalities.
Links to Paper and Supplementary Materials
Main Paper (Open Access Version): https://papers.miccai.org/miccai-2026-sat/paper/BrainWorks_007.pdf
SharedIt Link: Not yet available
SpringerLink (DOI): Not yet available
Supplementary Material: Not Submitted
Link to Open Review
Open Review Page: Not Available
BibTex
@InProceedings{MacFra_Unsupervised_MICCAISAT2026,
author = { Maccarone, Francesca AND Melzi, Simone AND Bercea, Cosmin I. AND Arrigoni, Filippo AND Schnabel, Julia A. AND Felsner, Lina AND Peruzzo, Denis},
title = { { Unsupervised Anomaly Detection in Pediatric Brains: Method Comparison Across Multiple Clinical Conditions } },
booktitle = {Medical Image Computing and Computer Assisted Intervention -- MICCAI 2026 Workshops and Challenges},
year = {2026},
publisher = {Springer Nature Switzerland},
volume = {LNCS 17254},
month = {pending},
page = {pending}
}