Related Experiment Video
Updated: Aug 11, 2026
Detection and Isolation of Cancer in Prostate Biopsies Using Stimulated Raman Histology and Artificial Intelligence
Published on: June 10, 2025
Classification and regression tree analysis for the prediction of aggressive prostate cancer on biopsy
Stephen E F Spurgeon1, Yi-Ching Hsieh, Adrianna Rivadinera
1Divisions of Hematology and Medical Oncology, Oregon Health and Science University, Portland, Oregon, USA.
Purpose:
Prostate cancer screening allows early cancer detection but not all patients benefit from subsequent therapy. Thus, identifying patients who are likely to harbor aggressive cancer could significantly decrease the number of prostate biopsies performed.
Materials And Methods:
Data were collected on 1,563 consecutive referred men with serum PSA 10 ng/ml or less who underwent an initial prostate biopsy. Predictors of aggressive cancer (Gleason sum 7 or greater) were identified using CART analysis. Model building was done in a randomly selected training set (70% of the data) and validation was completed using the remaining data.
Results:
Cancer was detected in 406 men (26.1%). Gleason 7 or greater cancer was found in 130 men (8.3%). CART created a decision tree that identified certain groups at risk for aggressive cancer, namely 1) PSAD greater than 0.165 ng/ml/cc, and 2) PSAD greater than 0.058 to 0.165 ng/ml/cc or less, age greater than 57.5 years and prostate volume greater than 22.7 cc. The incidence of aggressive prostate cancer was 1.1% when PSAD was 0.058 ng/ml/cc or less in the validation set. The sensitivity and specificity of CART for identifying men with aggressive cancer were 100% and 31.8% for model building data, and 91.5% and 33.5% for the validation data set, respectively.
Conclusions:
CART identified groups at risk for aggressive prostate cancer. Application of this CART could decrease unnecessary biopsies by 33.5% when only a diagnosis of high grade prostate cancer would lead to subsequent therapy.
