Deep-learning triage of three-dimensional pathology datasets for comprehensive and efficient pathologist assessments

Standard slide-based two-dimensional (2D) histopathology severely undersamples spatially heterogeneous tissue, with each thin 2D section representing <1% of the entire biopsy volume. Recent advances in non-destructive three-dimensional (3D) pathology, such as open-top light-sheet microscopy, enable comprehensive high-resolution imaging of large clinical specimens. Since manual review of these massive and complex 3D datasets is infeasible in clinical practice, we present TRICARE, a deep-learning triage framework that identifies high-risk 2D cross sections within 3D pathology datasets to enable time-efficient pathologist evaluation, which offers a lower-risk route for accelerated adoption by retaining pathologists for final diagnoses. TRICARE assigns risk scores to all 2D levels within a tissue volume by leveraging context from a subset of neighbouring depth levels, outperforming models in which predictions are based on isolated 2D levels. In two use cases-risk stratification based on prostate cancer biopsies and screening for dysplasia/cancer in endoscopic biopsies of Barrett's esophagus-AI-triaged 3D pathology, enabled by TRICARE, demonstrates the potential to improve the detection of high-risk diseases compared with slide-based 2D histopathology while optimizing pathologist workloads.