Non-destructive 3D pathology and analysis: a new perspective on cancer

We are developing non-destructive slide-free 3D pathology methods for clinical decision support. In comparison to conventional slide-based pathology, 3D pathology provides: (1) vastly greater sampling of tissue specimens including whole biopsies, (2) volumetric imaging of cell distributions and 3D tissue structures that are prognostic and predictive, and (3) a non-destructive and reversible workflow that preserves valuable biopsies for downstream molecular assays. Due to the immense size of feature-rich 3D pathology datasets, new challenges exist in terms of data management, human visualization, and computer-aided interpretation. We have been working on a full stack of technologies to facilitate the clinical adoption of 3D pathology, from sample preparation (reversible optical clearing and fluorescence labeling), high-throughput imaging with open-top light-sheet (OTLS) microscopes developed in our lab, to data processing and AI-based image triage and analysis. For AI analyses, we are developing both traditional machine classifiers based on intuitive “hand-crafted” 3D features and deep-learning classifiers based on sub-visual features. Our non-destructive large-volume digital pathology methods are synergistic with the growing fields of radiomics and genomics, which collectively have the potential to improve treatment decisions for diverse patient populations.

After viewing this lecture, participants should be able to:
1. State the advantages of 3D pathology over standard 2D histology
2. Identify clinical use cases in which 3D pathology could add value
3. Describe various approaches for AI-assisted analysis of 3D pathology datasets


Jonathan T.C. Liu, PhD
Professor; Mechanical Engineering, Bioengineering, and Laboratory Medicine & Pathology
Director, Molecular Biophotonics Laboratory
University of Washington

06/26/24