"No negatives needed": weakly-supervised regression for interpretable tumor detection in whole-slide histopathology images

2025-02-28

AI Summary

Tumor detection in whole-slide images (WSIs) for digital pathology critically depends on scarce tumor-free negative samples—a major clinical bottleneck. Method: We propose a novel weakly supervised regression paradigm that directly predicts the clinically reported tumor percentage, eliminating reliance on negative cases. Our approach employs a multi-instance learning–based regression architecture, incorporates a region magnification strategy to enhance sensitivity to small tumors, and integrates visual attention with logit maps to enable interpretable, localization-aware quantitative predictions. A noise-robust training scheme further improves generalizability. Contribution/Results: Validated across multi-organ and multi-specimen datasets, our method significantly improves detection of microscopic lesions while producing predictions that are both clinically interpretable and spatially traceable. To our knowledge, it is the first end-to-end weakly supervised regression solution for tumor burden assessment in WSIs.

Technology Category

• Computer Vision: Medical and Biological Imaging
• Machine Learning: Multi-instance/Multi-view Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Accurate tumor detection in digital pathology whole-slide images (WSIs) is crucial for cancer diagnosis and treatment planning. Multiple Instance Learning (MIL) has emerged as a widely used approach for weakly-supervised tumor detection with large-scale data without the need for manual annotations. However, traditional MIL methods often depend on classification tasks that require tumor-free cases as negative examples, which are challenging to obtain in real-world clinical workflows, especially for surgical resection specimens. We address this limitation by reformulating tumor detection as a regression task, estimating tumor percentages from WSIs, a clinically available target across multiple cancer types. In this paper, we provide an analysis of the proposed weakly-supervised regression framework by applying it to multiple organs, specimen types and clinical scenarios. We characterize the robustness of our framework to tumor percentage as a noisy regression target, and introduce a novel concept of amplification technique to improve tumor detection sensitivity when learning from small tumor regions. Finally, we provide interpretable insights into the model's predictions by analyzing visual attention and logit maps. Our code is available at https://github.com/DIAGNijmegen/tumor-percentage-mil-regression.

Problem

• Weakly-supervised tumor detection without negative examples
• Regression-based tumor percentage estimation in whole-slide images
• Improving sensitivity for small tumor regions via amplification

Innovation

• Weakly-supervised regression for tumor detection
• Tumor percentage estimation without negative examples
• Amplification technique for small tumor sensitivity