286092 20260413130622.0 10.1016/j.neucom.2026.133460 doi 0925-2312 ISSN 1872-8286 ISSN DZNE-2026-00388 610 Chatterjee, Soumick 0000-0001-7594-1188 0 Weakly-supervised segmentation using inherently-explainable classification models and their application to brain tumour classification Amsterdam 2026 Elsevier article DRIVER Output Types/Journal article DataCite Journal Article journal journal PUB:(DE-HGF)16 1776078240_32195 PUB:(DE-HGF) ARTICLE BibTeX JOURNAL_ARTICLE ORCID Journal Article 0 EndNote Deep learning has demonstrated significant potential in medical imaging; however, the opacity of “black-box” models hinders clinical trust, while segmentation tasks typically necessitate laborious, hard-to-obtain pixel-wise annotations. To address these challenges simultaneously, this paper introduces a framework for three inherently explainable classifiers (GP-UNet, GP-ShuffleUNet, and GP-ReconResNet). By integrating a global pooling mechanism, these networks generate localisation heatmaps that directly influence classification decisions, offering inherent interpretability without relying on potentially unreliable post-hoc methods. These heatmaps are subsequently thresholded to achieve weakly-supervised segmentation, requiring only image-level classification labels for training. Validated on two datasets for multi-class brain tumour classification, the proposed models achieved a peak F1-score of 0.93. For the weakly-supervised segmentation task, a median Dice score of 0.728 (95% CI: 0.715–0.739) was recorded. Notably, on a subset of tumour-only images, the best model achieved an accuracy of 98.7%, outperforming state-of-the-art glioma grading binary classifiers. Furthermore, comparative Precision-Recall analysis validated the framework’s robustness against severe class imbalance, establishing a direct correlation between diagnostic confidence and segmentation fidelity. These results demonstrate that the proposed framework successfully combines high diagnostic accuracy with essential transparency, offering a promising direction for trustworthy clinical decision support. 353 - Clinical and Health Care Research (POF4-353) G:(DE-HGF)POF4-353 POF4-353 POF IV 0 Dataset connected to CrossRef, Journals: pub.dzne.de Yassin, Hadya 0000-0003-0987-505X 1 Dubost, Florian 2 Nürnberger, Andreas 0000-0003-4311-0624 3 Speck, Oliver P:(DE-2719)2810706 4 Last author 10.1016/j.neucom.2026.133460 Vol. 682, p. 133460 - PERI:(DE-600)1479006-3 133460 Neurocomputing 682 2026 0925-2312 http://pub.dzne.de/record/286092/files/DZNE-2026-00388_Restricted.pdf http://pub.dzne.de/record/286092/files/DZNE-2026-00388_Restricted.pdf?subformat=pdfa pdfa Deutsches Zentrum für Neurodegenerative Erkrankungen I:(DE-588)1065079516 DZNE 4 P:(DE-2719)2810706 DE-HGF Gesundheit Neurodegenerative Diseases G:(DE-HGF)POF4-350 G:(DE-HGF)POF4-353 G:(DE-HGF)POF4 G:(DE-HGF)POF4-300 G:(DE-HGF)POF Clinical and Health Care Research 0 JCR StatID:(DE-HGF)0100 StatID NEUROCOMPUTING : 2022 2025-11-11 DBCoverage StatID:(DE-HGF)0200 StatID SCOPUS 2025-11-11 DBCoverage StatID:(DE-HGF)0300 StatID Medline 2025-11-11 DBCoverage StatID:(DE-HGF)0600 StatID Ebsco Academic Search 2025-11-11 Peer Review StatID:(DE-HGF)0030 StatID ASC 2025-11-11 DBCoverage StatID:(DE-HGF)0199 StatID Clarivate Analytics Master Journal List 2025-11-11 DBCoverage StatID:(DE-HGF)1160 StatID Current Contents - Engineering, Computing and Technology 2025-11-11 DBCoverage StatID:(DE-HGF)0160 StatID Essential Science Indicators 2025-11-11 WoS StatID:(DE-HGF)0113 StatID Science Citation Index Expanded 2025-11-11 DBCoverage StatID:(DE-HGF)0150 StatID Web of Science Core Collection 2025-11-11 IF >= 5 StatID:(DE-HGF)9905 StatID NEUROCOMPUTING : 2022 2025-11-11 I:(DE-2719)1340009 AG Speck Linking Imaging Projects 0 journal EDITORS VDBINPRINT I:(DE-2719)1340009 UNRESTRICTED