๐ค AI Summary
Existing methods struggle to jointly model global structural patterns and local histopathological details in whole-slide images (WSIs), resulting in suboptimal accuracy for grading and survival prognosis, as well as limited clinical interpretability. To address this, we propose an end-to-end self-calibrating network comprising three synergistic modules: a global branch, a focal predictor, and a detail branchโintegrated via feature consistency constraints to enhance region-specific attention and discriminative robustness. We further introduce a novel prognostic biomarker discovery paradigm grounded in feature cluster uniqueness and spatial tissue distribution. Our method unifies thumbnail-based global classification, attention-guided focal prediction, and multi-scale detail resampling with spatial alignment. Evaluated on multiple public benchmarks, it achieves significant improvements in both pathological grading accuracy and survival prediction performance, operates 3.2ร faster than state-of-the-art methods, and generates clinically interpretable lesion-localization heatmaps.
๐ Abstract
Pathology images are considered the ``gold standard"for cancer diagnosis and treatment, with gigapixel images providing extensive tissue and cellular information. Existing methods fail to simultaneously extract global structural and local detail features for comprehensive pathology image analysis efficiently. To address these limitations, we propose a self-calibration enhanced framework for whole slide pathology image analysis, comprising three components: a global branch, a focus predictor, and a detailed branch. The global branch initially classifies using the pathological thumbnail, while the focus predictor identifies relevant regions for classification based on the last layer features of the global branch. The detailed extraction branch then assesses whether the magnified regions correspond to the lesion area. Finally, a feature consistency constraint between the global and detail branches ensures that the global branch focuses on the appropriate region and extracts sufficient discriminative features for final identification. These focused discriminative features prove invaluable for uncovering novel prognostic tumor markers from the perspective of feature cluster uniqueness and tissue spatial distribution. Extensive experiment results demonstrate that the proposed framework can rapidly deliver accurate and explainable results for pathological grading and prognosis tasks.