Structured micro-ultrasonography training improves prostate cancer detection and management decisions

To evaluate whether a structured 2-day micro-ultrasonography (microUS) training curriculum improves diagnostic performance and management decisions for clinically significant prostate cancer (csPCa) detection in novice readers.

A total of 20 participants with limited microUS experience completed the Mannheim Micro-Ultrasound Curriculum, comprising didactic modules, supervised case interpretation, and 85 reference cases. Pre- and post-training performance was assessed on a fixed, biopsy confirmed test set of 100 cases (50 csPCa confirmed on radical prostatectomy, 50 biopsy negative). Each participant interpreted 10 randomised cases per timepoint, with no case assessed twice per reader and each case being assessed once before and after training in each training session. The primary endpoint was change in cohort-level accuracy for csPCa detection; secondary endpoints included sensitivity, specificity, predictive values, area under the receiver operating characteristic curve (AUC), Prostate Risk Identification using microUS (PRI-MUS) performance, and management recommendations. Mixed-effects models were applied as a robustness analysis.

Training significantly improved pooled cohort level accuracy from 50.5% (95% confidence interval [CI] 43.4-57.6%) to 64.5% (95% CI 57.4-71.1%; P = 0.006), sensitivity from 70.0% (95% CI 60.0-78.8%) to 92.0% (95% CI 84.8-96.5%; P = 0.001), and negative predictive value from 50.8% (95% CI 37.7-63.9%) to 82.2% (95% CI 67.9-92.0%; P = 0.001), whereas specificity did not increase significantly (31.0% vs 37.0%; P = 0.46). The AUC for the overall subjective risk score rose from 0.53 (95% CI 0.45-0.61) to 0.70 (95% CI 0.63-0.77; P = 0.002). Mixed-effects modelling confirmed a significant training effect (odds ratio 2.13, 95% CI 1.20-3.79; P = 0.01), reproducible across two independent sessions. PRI-MUS scoring improvements were not statistically significant. Overall decision correctness improved from 67% to 83% (P = 0.001), with poor-quality decisions decreasing from 33% to 17%.

Structured image-based microUS training significantly improved cohort-level diagnostic performance and management decisions in novice readers in a controlled reader-study. These findings support structured training and quality assurance before independent clinical implementation of microUS.