In a recent study, researchers from Mass General Brigham and the University of Washington have developed deep learning models called Tripath that can use 3D pathology datasets to predict clinical outcomes. By imaging curated prostate cancer specimens using high-resolution 3D imaging techniques, the models were trained to predict prostate cancer recurrence risk on volumetric human tissue biopsies. These models outperformed pathologists and other deep learning models that rely on 2D morphology and thin tissue slices. The results of the study, published in Cell, highlight the potential of Tripath to improve clinical decision-making and offer novel insights into prognosis and therapeutic response.
Lead author Andrew H. Song emphasized the importance of analyzing the entire volume of tissue samples to accurately predict patient risk, a feat only possible with the 3D pathology paradigm. Co-corresponding author Faisal Mahmood noted that Tripath, leveraging AI and 3D spatial biology techniques, could provide a framework for clinical decision support and the discovery of new biomarkers for prognosis and treatment response. Co-corresponding author Jonathan Liu highlighted the promise of using deep learning to extract sub-visual 3D features for risk stratification, potentially guiding critical treatment decisions.
Despite the promising results, the new approach will need further validation in larger datasets before it can be considered for clinical use. The researchers are optimistic about the potential of Tripath to improve patient outcomes through more accurate risk prediction and personalized treatment strategies. Their innovative use of AI and 3D imaging techniques represents a significant advancement in the field of pathology, offering a more comprehensive understanding of tissue complexity and the potential to uncover new insights into disease progression and treatment.
The researchers involved in the study have disclosed that they hold a provisional patent related to the technical and methodological aspects of Tripath. Jonathan Liu has additional ties to Alpenglow Biosciences, Inc., a company that has licensed the microscopy portfolio from his lab at the University of Washington. The study was funded by various organizations, including the Brigham and Women’s Hospital, Mass General Hospital, the National Institute of General Medical Sciences, the Department of Defense, the National Cancer Institute, and other funding sources, highlighting the collaborative effort and diverse support behind this groundbreaking research.
Overall, the development of Tripath represents a significant advancement in the field of pathology, with the potential to revolutionize clinical decision-making and personalized medicine. By leveraging deep learning models and 3D pathology datasets, researchers have demonstrated the enhanced predictive capabilities of Tripath in identifying prostate cancer recurrence risk. Continued research and validation of this approach in larger datasets will be critical to further refine and optimize this innovative tool for clinical use, offering new possibilities for improved patient outcomes and treatment strategies in the future.
