Class Visualizations and Activation Atlases for Enhancing Interpretability in Deep Learning-Based Computational Pathology
This work provides a framework for expert-centered interrogation of learned representations in deep learning models for computational pathology, addressing the interpretability gap for pathologists.
This paper systematically evaluated class visualizations (CVs) and activation atlases (AAs) for transformer-based models in computational pathology to enhance interpretability. They found that CVs preserved recognizability for distinct tissues (Fleiss k = 0.75 for real scans vs. k = 0.31 for CVs) but struggled with overlapping cancer subclasses, while AAs revealed layer-dependent organization with moderate agreement for tissue classification (k = 0.58) and low agreement for fine subclasses (k = 0.11).
The rapid adoption of transformer-based models in computational pathology has enabled prediction of molecular and clinical biomarkers from H&E whole-slide images, yet interpretability has not kept pace with model complexity. While attribution- and generative-based methods are common, feature visualization approaches such as class visualizations (CVs) and activation atlases (AAs) have not been systematically evaluated for these models. We developed a visualization framework and assessed CVs and AAs for a transformer-based foundation model across tissue and multi-organ cancer classification tasks with increasing label granularity. Four pathologists annotated real and generated images to quantify inter-observer agreement, complemented by attribution and similarity metrics. CVs preserved recognizability for morphologically distinct tissues but showed reduced separability for overlapping cancer subclasses. In tissue classification, agreement decreased from Fleiss k = 0.75 (scans) to k = 0.31 (CVs), with similar trends in cancer subclass tasks. AAs revealed layer-dependent organization: coarse tissue-level concepts formed coherent regions, whereas finer subclasses exhibited dispersion and overlap. Agreement was moderate for tissue classification (k = 0.58), high for coarse cancer groupings (k = 0.82), and low at subclass level (k = 0.11). Atlas separability closely tracked expert agreement on real images, indicating that representational ambiguity reflects intrinsic pathological complexity. Attribution-based metrics approximated expert variability in low-complexity settings, whereas perceptual and distributional metrics showed limited alignment. Overall, concept-level feature visualization reveals structured morphological manifolds in transformer-based pathology models and provides a framework for expert-centered interrogation of learned representations across label granularities.