A Weakly Supervised Deep Learning Framework for Whole Slide Classification to Facilitate Digital Pathology in Animal Study

The pathology of animal studies is crucial for toxicityevaluationsand regulatory assessments, but the manual examination of slides bypathologists remains time-consuming and requires extensive training.One inherent challenge in this process is the interobserver variability,which can compromise the consistency and accuracy of a study. Artificialintelligence (AI) has demonstrated its ability to automate similarexaminations in clinical applications with enhanced efficiency, consistency,and accuracy. However, training AI models typically relies on costlypixel-level annotation of injured regions and is often not availablefor animal pathology. To address this, we developed the PathologAIsystem, a "weakly" supervised approach for WSI classificationin rat images without explicit lesion annotation at the pixel level.Using rat liver imaging data from the Open TG-GATEs system, PathologAIwas applied to predict necrosis of n = 816 WSIs (377controls). TG-GATEs studied 170 compounds at three dose levels (low,middle, and high) for four time points (3, 7, 14, and 28 days). PathologAIfirst preprocessed WSIs at the tile level to generate a high-levelrepresentation with a Generative Adversarial Network architecture.The prediction of liver necrosis relied on an ensemble model of 5CNN classifiers trained on 335 WSIs. The ensemble model achieved notableclassification accuracy on the holdout test set: 87% among 87 controlslides free of findings, 83% among 120 controls with spontaneous necrosis,67% among 147 treated animals with spontaneous minimal or slight necrosis,and 59% among 127 treated animals with minimal or slight necrosiscaused by the treatment. Importantly, PathologAI was able to discriminateWSIs with spontaneous necrosis from those with treatment related necrosisand discriminated mild lesion level findings (slight vs minimal) andbetween treatment dose levels. PathologAI could provide an inexpensiveand rapid screening tool to assist the digital pathology analysisin preclinical applications and general toxicological studies.