Researchers have introduced Tripath, a new set of deep learning models designed to analyze 3D pathology datasets and make predictions about clinical outcomes. The team used two advanced 3D imaging techniques to capture images of prostate cancer specimens, and then trained the models to predict the risk of cancer recurrence using volumetric human tissue biopsies. Tripath outperformed pathologists and other deep learning models that only rely on 2D morphology and thin tissue slices by capturing comprehensive 3D morphologies from the entire tissue volume. The intricate and complex nature of human tissue is best represented in three dimensions.In a recent study, Mass General Brigham researchers and their partners have introduced Tripath: advanced deep learning models designed to utilize 3D pathology datasets for predicting clinical outcomes. This signifies a shift towards examining tissue in a three-dimensional form, as opposed to traditional two-dimensional slices. The 3D datasets contain significantly more data than the 2D ones, making manual examination impractical.In collaboration with the University of Washington, the research team used two 3D high-resolution imaging techniques to image prostate cancer specimens. They then trained models to predict the risk of prostate cancer recurrence on volumetric human tissue biopsies. By capturing 3D morphologies from the entire tissue volume, the Tripath method outperformed pathologists and deep learning models that rely on 2D morphology and thin tissue slices. The results of the study are published in Cell.
While the new approach still needs to be validated in larger datasets before it can be used in clinical settings, the researchers are hopeful about its potential for further development.The potential of the Tripath approach to inform clinical decision making is being met with optimism. Lead author Andrew H. Song, PhD, from the Division of Computational Pathology at Mass General Brigham, emphasized the importance of analyzing the entire volume of a tissue sample for accurate patient risk prediction. This comprehensive analysis is made possible by the 3D pathology paradigm developed by the team. By leveraging AI and 3D spatial biology techniques, Tripath offers a framework for clinical decision support and has the potential to uncover new biomarkers for prognosis and therapy.The co-corresponding author Faisal Mahmood, PhD, from the Division of Computational Pathology at Mass General Brigham, stated that Tripath is their first attempt to use deep learning to extract sub-visual 3D features for risk stratification, which has promising potential for guiding critical treatment decisions. Co-corresponding author Jonathan Liu, PhD, from the University of Washington, added that in their prior work in computational 3D pathology, they focused on specific structures such as the prostate gland network. He also mentioned that Tripath shows promising potential for guiding critical treatment decisions. Song and Mahmood have a provisional patent related to this work.This study focuses on the technical and methodological aspects. Liu is a co-founder and board member of Alpenglow Biosciences, Inc., which has licensed the OTLS microscopy portfolio developed at the University of Washington.
The authors received funding from Brigham and Women’s Hospital (BWH) President’s Fund, Mass General Hospital (MGH) Pathology, the National Institute of General Medical Sciences (R35GM138216), Department of Defense (DoD) Prostate Cancer Research Program (W81WH-18-10358 and W81XWH-20-1-0851), the National Cancer Institute (R01CA268207), the National Institute of Biomedical Imaging and Bioengineering.The study was funded by the National Institutes of Health (R01EB031002), the Canary Foundation, the NCI Ruth L. Kirschstein National Service Award (T32CA251062), the Leon Troper Professorship in Computational Pathology at Johns Hopkins University, UKRI, mdxhealth, NHSX, and Clarendon Fund.
