Abstract:Artificial intelligence (AI) methods including deep neural networks can provide rapid molecular classification of tumors from routine histology with accuracy that can match or exceed human pathologists. Discerning how neural networks make their predictions remains a significant challenge, but explainability tools can help provide insights into what models have learned when corresponding histologic features are poorly understood. Conditional generative adversarial networks (cGANs) are AI models that generate synthetic images and illustrate subtle differences between image classes. Here, we describe the use of a cGAN for explaining models trained to classify molecularly-subtyped tumors, exposing associated histologic features. We leverage cGANs to create class- and layer-blending visualizations to improve understanding of subtype morphology. Finally, we demonstrate the potential use of synthetic histology for augmenting pathology trainee education and show that clear, intuitive cGAN visualizations can reinforce and improve human understanding of histologic manifestations of tumor biology
Abstract:A model's ability to express its own predictive uncertainty is an essential attribute for maintaining clinical user confidence as computational biomarkers are deployed into real-world medical settings. In the domain of cancer digital histopathology, we describe a novel, clinically-oriented approach to uncertainty quantification (UQ) for whole-slide images, estimating uncertainty using dropout and calculating thresholds on training data to establish cutoffs for low- and high-confidence predictions. We train models to identify lung adenocarcinoma vs. squamous cell carcinoma and show that high-confidence predictions outperform predictions without UQ, in both cross-validation and testing on two large external datasets spanning multiple institutions. Our testing strategy closely approximates real-world application, with predictions generated on unsupervised, unannotated slides using predetermined thresholds. Furthermore, we show that UQ thresholding remains reliable in the setting of domain shift, with accurate high-confidence predictions of adenocarcinoma vs. squamous cell carcinoma for out-of-distribution, non-lung cancer cohorts.