Machine learning-based prediction of upgrading on magnetic resonance imaging targeted biopsy in patients eligible for active surveillance

• Published in: Urologic oncology Vol. 40; no. 5; pp. 191.e15 - 191.e20
• Main Authors: ElKarami, Bashier, Deebajah, Mustafa, Polk, Seth, Peabody, James, Shahrrava, Behnam, Menon, Mani, Alkhateeb, Abedalrhman, Alanee, Shaheen
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.05.2022
• Subjects: Biopsy; Humans; Image-Guided Biopsy - methods; Machine Learning; Magnetic Resonance Imaging - methods; Male; Neoplasm Grading; Prostate cancer; Prostatic Neoplasms - diagnostic imaging; Prostatic Neoplasms - pathology; Retrospective Studies; Targeted biopsy; Watchful Waiting; Model; Targeted biopsy; Prostate cancer; Machine learning
• ISSN: 1078-1439; 1873-2496; 1873-2496
• DOI: 10.1016/j.urolonc.2022.01.012

•MRI and MRI-TB of the prostate may not be available to some patients.•Machine learning methods have high accuracy for predicting clinical outcomes.•Our model accurately predicts upgrading after MRI targeted biopsy. To examine the ability of machine learning methods to predict upgrading of Gleason score on confirmatory magnetic resonance imaging-guided targeted biopsy (MRI-TB) of the prostate in candidates for active surveillance. Our database included 592 patients who received prostate multiparametric magnetic resonance imaging in the evaluation for active surveillance. Upgrading to significant prostate cancer on MRI-TB was defined as upgrading to G 3+4 (definition 1 - DF1) and 4+3 (DF2). Machine learning classifiers were applied on both classification problems DF1 and DF2. Univariate analysis showed that older age and the number of positive cores on pre-MRI-TB were positively correlated with upgrading by DF1 (P-value ≤ 0.05). Upgrading by DF2 was positively correlated with age and the number of positive cores and negatively correlated with body mass index. For upgrading prediction, the AdaBoost model was highly predictive of upgrading by DF1 (AUC 0.952), while for prediction of upgrading by DF2, the Random Forest model had a lower but excellent prediction performance (AUC 0.947). We show that machine learning has the potential to be integrated in future diagnostic assessments for patients eligible for AS. Training our models on larger multi-institutional databases is needed to confirm our results and improve the accuracy of these models’ prediction.