Summary The incidence of esophageal adenocarcinoma (EAC) has been rising dramatically in Western populations over the last several decades. EAC arises from Barrett’s esophagus (BE), a specialized intestinal metaplasia of the esophagus associated with chronic acid reflux. Patients with BE are recommended to undergo periodic endoscopic screening, in which biopsies are obtained along the BE-lined esophagus to detect progression to EAC. Here, the goal is to detect the presence of neoplasia, and if present, to categorize the lesion as low-grade dysplasia (LGD), high-grade dysplasia (HGD) or cancer (EAC), all of which have unique treatment implications. Unfortunately, there is high inter-pathologist disagreement in distinguishing between LGD and HGD. In particular, LGD is notoriously difficult to diagnose by histopathology, and has variable progression rates to HGD and/or EAC. Therefore, there is a need to improve our methods for obtaining a definitive diagnosis and treatment recommendation based on endoscopic biopsies. A significant contributing factor to this problem is that pathological grading (risk assessment) of esophageal biopsies currently relies upon the subjective interpretation of 2D sections that only represent ~1% of the total volume of the biopsies. Our team at the Univ. of Washington (UW), the Fred Hutch Cancer Center (FHCC), and the Brigham and Women’s Hospital (BWH) is pioneering the development of non-destructive 3D pathology and associated computational methods (2D and 3D) for clinical decision support. Non-destructive 3D pathology has the potential to greatly improve diagnostic determinations by enabling: (1) orders-of-magnitude greater sampling of tissue specimens, (2) volumetric imaging of cell distributions, tissue structures, and other novel 3D spatial biomarkers that are prognostic/predictive, and (3) non- destructive imaging, which preserves valuable tissues (e.g. whole biopsies) for downstream molecular assays. Over the past few years, we have conducted a series of studies to evaluate our nondestructive 3D pathology methods for the diagnosis and grading of BE-related lesions (LGD, HGD, and EAC). We have demonstrated the ability of 3D pathology to elucidate esophageal lesions that are ambiguous with conventional 2D histology. We have also developed deep-learning-assisted pipelines to analyze massive 3D pathology datasets for prognostication of cancer outcomes. Here, we will continue to refine these collective technologies to improve the treatment and outcomes for patients with BE. Our project goals include: Aim 1 (standardization and quality control) – to develop a 3D pathology pipeline to enable reproducible (>95% yield) generation of clinical-grade 3D datasets; Aim 2 (AI triage to assist pathologists) – to develop weakly supervised deep-learning triage methods, based on annotated 2D image levels within 3D pathology datasets, for time-efficient pathologist interpretation of 3D pathology datasets; and Aim 3 (AI decision suppor...