Michael Donovan

PhD, MD

University of Miami

Vice Chair Translational Research

PLCO (Learn more about this study)

PLCOI-803

Jul 2, 2021

Use of Artificial Intelligence to develop an automated, outcome-baaed, prostate cancer grading system.

We have developed an AI-digital multi-feature array platform to extract cell and tissue quantitative architectural attributes of various disease processes including prostate cancer. Of relevance, there are an estimated 191,930 new cases of prostate cancer diagnosed in the US in 2020 representing 21% of new cancers in men. The challenge is that early detection of prostate cancer can lead to over treatment without a benefit of improving life expectancy while leading to a negative impact on quality of life. Treatment decisions including surgery, radiation and active surveillance are impacted by the initial grade of the prostate cancer which remains subjective and limited in accurately predicting an outcome.
Over the past 15+ years myself and colleagues from Mount Sinai) have published extensively on the importance of standardizing and quantifying prostate cancer grading systems to improve risk discrimination and most importantly guide treatment. We focused on predicting likelihood of disease recurrence based on retrospective outcomes using the prostate needle biopsy. Although our earlier published studies have focused on the use of quantitative immunofluorescence combined with morphometry, we have been able to advance the field, by automating and quantifying prostate cancer grade solely from the prostate needle biopsy H&E whole slide digital image and then integrating these features with clinical data to predict outcomes. Such approaches, once validated, will allow informed treatment decisions including the use of active surveillance vs more interventional therapy such as surgery or radiation, These types of investigations require large amounts of heterogeneous image and clinical data sets to impact the clinical decision process and ultimately improve upon outcomes.