> Our Work > The Work We Fund

2025 AIRA Matrix – PCF Special Challenge Award

Testing Deep Learning Algorithms for Prognostication and Prediction in Prostate Cancer: A Clinical Validation Study

Principal Investigators: Tamara Lotan, MD (Johns Hopkins University), Angelo De Marzo, MD, PhD (Johns Hopkins University), Alex Baras, MD, PhD (Johns Hopkins University)

Description:

  • Clinical decision-making in localized prostate cancer relies on pathologist evaluations of prostate tumor biopsies, using a visually estimated 60-year-old “Gleason grading” system. Patients with aggressive-looking tumors are likely to be treated with surgery, radiation and/or hormonal therapy, while patients with tumors that appear pathologically indolent are often monitored with active surveillance. However, there is high variability in Gleason grading by pathologists with different experience levels, and better methods are needed.
  • Artificial Intelligence (AI) / Deep Learning digital pathology algorithms for prostate cancer grading hold incredible promise to eliminate subjectivity, increase usage and utility of quantitative grading metrics (quantification of tumor areas and amount of high-risk disease); and enable equal access to expert pathology grading throughout the world.
  • While several pathology-based AI algorithms have been developed for prostate cancer identification, grading, and prognosis, few have gained FDA approval. The AIRA Matrix AIRAProstate algorithm has demonstrated high performance for predicting patient outcomes in head-to-head tests with other algorithms and similar or better performance than expert human pathologists.
  • In this project, Dr. Tamara Lotan and team will leverage several large prostate cancer and non-cancer prostate digital pathology datasets to provide key validation data required to take AIRAProstate through FDA clearance.
  • The team will test AIRAProstate on a large prostate cancer biopsy cohort with existing expert pathologist grading determinations from Johns Hopkins. They will also test AIRAProstate on a large collection of benign and atypical prostate biopsy cores, and a collection of 500 whole slide images of prostate tissues with non-adenocarcinoma diagnoses, including both common and uncommon and benign and diseased states, to ensure the algorithm can distinguish these from prostate adenocarcinoma.
  • If successful, this project will result in a new FDA-approved digital pathology test that provided improved prostate cancer diagnosis and prognostication.  

What this means to patients: Clinical decision-making in localized prostate cancer is guided by subjective pathologic grading of tumor histopathologic images, based on a semi-quantitative, visually-based grading system developed 60 years ago. AI-based digital pathology algorithms have the potential to revolutionize pathologic assessment of prostate cancer by markedly improving reproducibility and accuracy of tumor grading, and enabling patients in low-resource settings anywhere in the world to have access to this technology. Dr. Lotan and team will perform validation studies necessary for the high-performing AI-based digital pathology algorithm AIRA Prostate, to move toward FDA clearance for prostate cancer identification and grading, which will ultimately greatly improve outcomes for patients around the world.