Wang J, et al. - This study attempted to assess whether machine learning-based analysis of MR radiomics can help improve the performance PI-RADS v2 in clinically relevant prostate cancer (PCa). The obtained data investigates that machine learning analysis of MR radiomics can help enhance the performance of PI-RADS in clinically relevant prostate cancer.

Methods

• This IRB-approved study enrolled fifty-four patients with PCa undergoing multi-parametric (mp) MRI before prostatectomy.
• They conducted an imaging analysis on 54 tumours, 47 normal peripheral (PZ) and 48 normal transitional (TZ) zone based on histological-radiological correlation.
• Mp-MRI was scored via PI-RADS, and quantified by assessing radiomic features.
• They established predictive model applying a novel support vector machine trained with: (i) radiomics, (ii) PI-RADS scores, (iii) radiomics and PI-RADS scores.
• In this study, paired comparison was made via ROC analysis.

Results

• It was noted that the model trained with radiomics had a significantly higher area under the ROC curve (Az) (0.955 [95% CI 0.923–0.976]) than PI-RADS (Az: 0.878 [0.834–0.914], p 0.001
• The obtained data indicates that Az between them was insignificant for PCa versus PZ (0.972 [0.945–0.988] vs. 0.940 [0.905–0.965], p=0.097).
• In context of the findings, when radiomics was added, performance of PI-RADS was significantly improved for PCa versus PZ (Az: 0.983 [0.960–0.995]) and PCa versus TZ (Az: 0.968 [0.940–0.985]).