Mitigating Spurious Correlations in Patch-wise Tumor Classification on High-Resolution Multimodal Images

2025-11-17

AI Summary

In high-resolution multimodal histopathological image patch-level tumor classification, spurious correlations—arising from inter-patch variations in tissue region size (e.g., tumor patches containing large tissue areas versus non-tumor patches dominated by background)—hinder model generalization to challenging cases such as small tumors or large background regions. To address this, we propose GERNE, a debiasing learning strategy that explicitly models and mitigates spurious associations between patch composition and class labels within a binary patch classification framework, optimizing worst-group accuracy. GERNE employs multimodal nonlinear feature extraction and robust sensitivity analysis via thresholded salient feature evaluation. On clinically critical hard cases, it significantly improves performance: worst-group accuracy increases by approximately 7% over empirical risk minimization. This yields enhanced generalization and robustness—particularly for detecting small tumors—while preserving interpretability and clinical relevance.

Technology Category

• Machine Learning: Multi-class/Multi-label Learning & Extreme Classification
• Computer Vision: Multi-modal Vision
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Web Mining and Content Analysis: Robustness and generalizability of Web mining methods
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Patch-wise multi-label classification provides an efficient alternative to full pixel-wise segmentation on high-resolution images, particularly when the objective is to determine the presence or absence of target objects within a patch rather than their precise spatial extent. This formulation substantially reduces annotation cost, simplifies training, and allows flexible patch sizing aligned with the desired level of decision granularity. In this work, we focus on a special case, patch-wise binary classification, applied to the detection of a single class of interest (tumor) on high-resolution multimodal nonlinear microscopy images. We show that, although this simplified formulation enables efficient model development, it can introduce spurious correlations between patch composition and labels: tumor patches tend to contain larger tissue regions, whereas non-tumor patches often consist mostly of background with small tissue areas. We further quantify the bias in model predictions caused by this spurious correlation, and propose to use a debiasing strategy to mitigate its effect. Specifically, we apply GERNE, a debiasing method that can be adapted to maximize worst-group accuracy (WGA). Our results show an improvement in WGA by approximately 7% compared to ERM for two different thresholds used to binarize the spurious feature. This enhancement boosts model performance on critical minority cases, such as tumor patches with small tissues and non-tumor patches with large tissues, and underscores the importance of spurious correlation-aware learning in patch-wise classification problems.

Problem

• Addresses spurious correlations in patch-wise tumor classification on multimodal images
• Mitigates bias between patch composition and labels in binary classification
• Improves worst-group accuracy for tumor detection with small tissue areas

Innovation

• Patch-wise binary classification for tumor detection
• GERNE debiasing method to mitigate spurious correlations
• Maximizing worst-group accuracy for minority cases