Breast Cancer Classification in Deep Ultraviolet Fluorescence Images Using a Patch-Level Vision Transformer Framework

2025-05-12

AI Summary

Intraoperative rapid assessment of surgical margins during breast-conserving surgery remains challenging. Method: This study proposes a patch-level Vision Transformer (Patch-ViT) for classifying deep-ultraviolet fluorescence whole-slide microscopy (DUV-FSM) images. To address discriminative difficulties arising from high resolution and complex tissue textures, the model jointly encodes local fine-grained features and global contextual information—marking the first application of Patch-ViT to DUV fluorescence histopathology. Additionally, we integrate a Grad-CAM++-based interpretability-weighting mechanism to enhance diagnostic transparency and clinical trustworthiness. Results: Evaluated on a real-world DUV-FSM dataset via five-fold cross-validation, the model achieves 98.33% accuracy, significantly outperforming mainstream CNN baselines. This work establishes a novel paradigm for intraoperative, label-free, and high-accuracy benign–malignant classification of breast tissue.

Technology Category

• Computer Vision: Interpretability, Explainability, and Transparency
• Machine Learning: Calibration & Uncertainty Quantification
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Responsible Web: Data and user privacy-enhancing technologies for the Web
• Security and Privacy: Data transparency and provenance

Abstract

Breast-conserving surgery (BCS) aims to completely remove malignant lesions while maximizing healthy tissue preservation. Intraoperative margin assessment is essential to achieve a balance between thorough cancer resection and tissue conservation. A deep ultraviolet fluorescence scanning microscope (DUV-FSM) enables rapid acquisition of whole surface images (WSIs) for excised tissue, providing contrast between malignant and normal tissues. However, breast cancer classification with DUV WSIs is challenged by high resolutions and complex histopathological features. This study introduces a DUV WSI classification framework using a patch-level vision transformer (ViT) model, capturing local and global features. Grad-CAM++ saliency weighting highlights relevant spatial regions, enhances result interpretability, and improves diagnostic accuracy for benign and malignant tissue classification. A comprehensive 5-fold cross-validation demonstrates the proposed approach significantly outperforms conventional deep learning methods, achieving a classification accuracy of 98.33%.

Problem

• Classifying breast cancer in DUV fluorescence images using ViT
• Improving intraoperative margin assessment for tissue conservation
• Enhancing diagnostic accuracy and interpretability with Grad-CAM++

Innovation

• Patch-level Vision Transformer for DUV WSI classification
• Grad-CAM++ saliency weighting enhances interpretability
• Achieves 98.33% accuracy via 5-fold cross-validation