Weakly supervised attention-based learning supports breast cancer detection on deep ultraviolet-excitation fluorescence images
5 affiliations
- 1Department of Pathology and Cell Regulation, Graduate School of Medical Science, Kyoto Prefectural University of Medicine
- 2Hyogo Prefectural Nishinomiya Hospital
- 3Department of Surgical Pathology, Graduate School of Medical Science, Kyoto Prefectural University of Medicine
- 4Department of Endocrine and Breast Surgery, Graduate School of Medical Science, Kyoto Prefectural University of Medicine
- 5Division of Healthcare Digital Transformation, Data-Driven Innovation Initiative, Kyushu University
Abstract
Microscopy with ultraviolet surface excitation (MUSE) enables rapid fluorescence imaging of fresh tissue surfaces without conventional sectioning, but its appearance differs from that of haematoxylin and eosin-stained slides. Pixel- or patch-level annotation for deep learning is laborious, particularly for emerging imaging modalities. Here, we evaluated weakly supervised learning (WSL) for breast cancer detection on MUSE images using attention-based multiple instance learning trained with microscopic image-level labels alone. Fresh breast tissues from 35 mastectomy patients were stained with Hoechst 33,342 and terbium and imaged by MUSE. We analysed 700 images, comprising 10 cancerous and 10 non-cancerous images per case. Data were split at the case level into an independent test set of 140 images and a training/validation set of 560 images with fixed four-fold cross-validation. To assess the effect of patch size, we trained models using 224 × 224- or 64 × 64-pixel patches. The models achieved high test performance, with receiver operating characteristic area under the curve values of 0.983 and 0.979 for the 224 × 224- and 64 × 64-pixel models, respectively. Attention maps highlighted tumour-rich regions and cancer-associated stroma. These findings support the feasibility of WSL for MUSE-based breast cancer detection while reducing the need for exhaustive annotation.
Related
- Transcranial photobiomodulation rebalances cortical excitation/inhibition in young adults with attention deficit/hyperactivity disorder
- Weakly-Supervised Gated Attention Multiple Instance Learning for Multi-Center Cervical Cytology Slide Triage
- MULTIMODAL EEG WORLD MODEL: SELF-SUPERVISED LATENT TRANSITION LEARNING FOR WEARABLE EEG SEIZURE DETECTION
- Single-view echocardiography-based deep learning for pulmonary hypertension detection: a multi-level supervised and explainable approach
- High-Performance Liquid Chromatography-Ultraviolet Detection and Gas Chromatography-Mass Spectrometry Analyses of Bioactive Compounds of Urtica dioica Leaf Extracts Using Various Solvents
BibTeX
@article{nakao2026weakly,
title = {{Weakly supervised attention-based learning supports breast cancer detection on deep ultraviolet-excitation fluorescence images}},
author = {Ryuta Nakao and Tomoya Matsui and Masatoshi Abe and Mizuki Honda and Osamu Inamori and Ippei Takada and Yoshinori Harada and Midori Morita and Yasuto Naoi and Hirohiko Niioka and Tetsuro Takamatsu},
journal = {Scientific Reports},
year = {2026},
month = jul,
doi = {10.1038/s41598-026-63683-w},
url = {https://doi.org/10.1038/s41598-026-63683-w},
note = {CC BY},
}
RIS
TY - JOUR
TI - Weakly supervised attention-based learning supports breast cancer detection on deep ultraviolet-excitation fluorescence images
AU - Ryuta Nakao
AU - Tomoya Matsui
AU - Masatoshi Abe
AU - Mizuki Honda
AU - Osamu Inamori
AU - Ippei Takada
AU - Yoshinori Harada
AU - Midori Morita
AU - Yasuto Naoi
AU - Hirohiko Niioka
AU - Tetsuro Takamatsu
PY - 2026
DA - 2026/07/01
JO - Scientific Reports
DO - 10.1038/s41598-026-63683-w
UR - https://doi.org/10.1038/s41598-026-63683-w
AB - Microscopy with ultraviolet surface excitation (MUSE) enables rapid fluorescence imaging of fresh tissue surfaces without conventional sectioning, but its appearance differs from that of haematoxylin and eosin-stained slides. Pixel- or patch-level annotation for deep learning is laborious, particularly for emerging imaging modalities. Here, we evaluated weakly supervised learning (WSL) for breast cancer detection on MUSE images using attention-based multiple instance learning trained with microscopic image-level labels alone. Fresh breast tissues from 35 mastectomy patients were stained with Hoechst 33,342 and terbium and imaged by MUSE. We analysed 700 images, comprising 10 cancerous and 10 non-cancerous images per case. Data were split at the case level into an independent test set of 140 images and a training/validation set of 560 images with fixed four-fold cross-validation. To assess the effect of patch size, we trained models using 224 × 224- or 64 × 64-pixel patches. The models achieved high test performance, with receiver operating characteristic area under the curve values of 0.983 and 0.979 for the 224 × 224- and 64 × 64-pixel models, respectively. Attention maps highlighted tumour-rich regions and cancer-associated stroma. These findings support the feasibility of WSL for MUSE-based breast cancer detection while reducing the need for exhaustive annotation.
SN - 2045-2322
ER -