Analysis of 3D pathology samples using weakly supervised AI

Andrew H Song; Mane Williams; Drew F K Williamson; Sarah S L Chow; Guillaume Jaume; Gan Gao; Andrew Zhang; Bowen Chen; Alexander S Baras; Robert Serafin; Richard Colling; Michelle R Downes; Xavier Farré; Peter Humphrey; Clare Verrill; Lawrence D True; Anil V Parwani; Jonathan T C Liu; Faisal Mahmood

doi:10.1016/j.cell.2024.03.035

Analysis of 3D pathology samples using weakly supervised AI

Cell. 2024 May 9;187(10):2502-2520.e17. doi: 10.1016/j.cell.2024.03.035.

Authors

Andrew H Song¹, Mane Williams², Drew F K Williamson¹, Sarah S L Chow³, Guillaume Jaume¹, Gan Gao³, Andrew Zhang⁴, Bowen Chen¹, Alexander S Baras⁵, Robert Serafin³, Richard Colling⁶, Michelle R Downes⁷, Xavier Farré⁸, Peter Humphrey⁹, Clare Verrill¹⁰, Lawrence D True¹¹, Anil V Parwani¹², Jonathan T C Liu¹³, Faisal Mahmood¹⁴

Affiliations

¹ Department of Pathology, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA; Department of Pathology, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA; Cancer Program, Broad Institute of Harvard and MIT, Cambridge, MA, USA; Data Science Program, Dana-Farber Cancer Institute, Boston, MA, USA.
² Department of Pathology, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA; Cancer Program, Broad Institute of Harvard and MIT, Cambridge, MA, USA; Data Science Program, Dana-Farber Cancer Institute, Boston, MA, USA; Department of Biomedical Informatics, Harvard Medical School, Boston, MA, USA.
³ Department of Mechanical Engineering, Bioengineering, and Laboratory Medicine & Pathology, University of Washington, Seattle, WA, USA.
⁴ Department of Pathology, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA; Cancer Program, Broad Institute of Harvard and MIT, Cambridge, MA, USA; Data Science Program, Dana-Farber Cancer Institute, Boston, MA, USA; Harvard-MIT Division of Health Sciences and Technology, Massachusetts Institute of Technology, Cambridge, MA, USA.
⁵ Department of Pathology, Johns Hopkins University School of Medicine, Baltimore, MD, USA; Department of Biomedical Engineering, Johns Hopkins University School of Medicine, Baltimore, MD, USA.
⁶ Nuffield Department of Surgical Sciences, University of Oxford, UK; Department of Cellular Pathology, Oxford University Hospitals NHS Foundations Trust, John Radcliffe Hospital, Oxford, UK.
⁷ Sunnybrook Health Sciences Centre, University of Toronto, Toronto, ON, Canada.
⁸ Public Health Agency of Catalonia, Lleida, Spain.
⁹ Department of Pathology, Yale School of Medicine, New Haven, CT, USA.
¹⁰ Nuffield Department of Surgical Sciences, University of Oxford, UK; Department of Cellular Pathology, Oxford University Hospitals NHS Foundations Trust, John Radcliffe Hospital, Oxford, UK; NIHR Oxford Biomedical Research Centre, Oxford University Hospitals NHS Foundation Trust, Oxford, UK.
¹¹ Department of Laboratory Medicine & Pathology, University of Washington School of Medicine, Seattle, WA, USA.
¹² Department of Pathology, The Ohio State University, Columbus, OH, USA.
¹³ Department of Mechanical Engineering, Bioengineering, and Laboratory Medicine & Pathology, University of Washington, Seattle, WA, USA. Electronic address: jonliu@uw.edu.
¹⁴ Department of Pathology, Brigham and Women's Hospital, Harvard Medical School, Boston, MA, USA; Department of Pathology, Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA; Cancer Program, Broad Institute of Harvard and MIT, Cambridge, MA, USA; Data Science Program, Dana-Farber Cancer Institute, Boston, MA, USA. Electronic address: faisalmahmood@bwh.harvard.edu.

PMID: 38729110
DOI: 10.1016/j.cell.2024.03.035

Abstract

Human tissue, which is inherently three-dimensional (3D), is traditionally examined through standard-of-care histopathology as limited two-dimensional (2D) cross-sections that can insufficiently represent the tissue due to sampling bias. To holistically characterize histomorphology, 3D imaging modalities have been developed, but clinical translation is hampered by complex manual evaluation and lack of computational platforms to distill clinical insights from large, high-resolution datasets. We present TriPath, a deep-learning platform for processing tissue volumes and efficiently predicting clinical outcomes based on 3D morphological features. Recurrence risk-stratification models were trained on prostate cancer specimens imaged with open-top light-sheet microscopy or microcomputed tomography. By comprehensively capturing 3D morphologies, 3D volume-based prognostication achieves superior performance to traditional 2D slice-based approaches, including clinical/histopathological baselines from six certified genitourinary pathologists. Incorporating greater tissue volume improves prognostic performance and mitigates risk prediction variability from sampling bias, further emphasizing the value of capturing larger extents of heterogeneous morphology.

Keywords: 3D deep learning; 3D microscopy; 3D pathology; computational pathology; deep learning; intratumoral heterogeneity; microCT; patient prognosis; slide-free microscopy.

Publication types

Research Support, N.I.H., Extramural
Research Support, Non-U.S. Gov't
Research Support, U.S. Gov't, Non-P.H.S.

MeSH terms

Deep Learning
Humans
Imaging, Three-Dimensional* / methods
Male
Prognosis
Prostatic Neoplasms* / diagnostic imaging
Prostatic Neoplasms* / pathology
Supervised Machine Learning
X-Ray Microtomography / methods