The researchers analyzed 811 prostate multiparametric MRI exams from four tertiary care centers to develop the machine-learning algorithm to predict lesions categorized as PI-RADS 4 or 5.
Among the factors considered were the patient's age, prostate-specific antigen (PSA) level, and prostate volume. Subjects had either no prior prostate biopsy or a negative prior prostate biopsy.
The model achieved 73% accuracy for predicting PI-RADS 4 or 5 lesions on the basis of 10-fold cross-validation. At a cutoff threshold of 43% for predicting these lesions, the algorithm has a sensitivity of 75% and specificity of 82%, the researchers found.
Dr. Zachary Nuffer from the University of Rochester Medical Center is scheduled to present the findings at ARRS 2018 in Washington, DC. The meeting is being held April 22-27.
Copyright © 2018 AuntMinnie.com