Evaluation of MRI for diagnosis of extraprostatic extension in prostate cancer

PurposeTo assess the ability of magnetic resonance imaging (MRI) to diagnose extraprostatic extension (EPE) in prostate cancer. Materials and MethodsWith Institutional Review Board (IRB) approval, 149 men with 170 0.5mL tumors underwent preoperative 3T MRI followed by radical prostatectomy (RP) between 2012-2015. Two blinded radiologists (R1/R2) assessed tumors using Prostate Imaging Reporting and Data System (PI-RADS) v2, subjectively evaluated for the presence of EPE, measured tumor size, and length of capsular contact (LCC). A third blinded radiologist, using MRI-RP-maps, measured whole-lesion: apparent diffusion coefficient (ADC) mean/centile and histogram features. Comparisons were performed using chi-square, logistic regression, and receiver operator characteristic (ROC) analysis. ResultsThe subjective EPE assessment showed high specificity (SPEC=75.4/91.3% [R1/R2]), low sensitivity (SENS=43.3/43.6% [R1/R2]), and area-under (AU) ROC curve=0.67 (confidence interval [CI] 0.61-0.73) R1 and 0.61 (CI 0.53-0.70) R2; (k=0.33). PI-RADS v2 scores were strongly associated with EPE (P<0.001 / P=0.008; R1/R2) with AU-ROC curve=0.72 (0.64-0.79) R1 and 0.61 (0.53-0.70) R2; (k=0.44). Tumors with EPE were larger (18.87.8 [median 17, range 6-51] vs. 18.84.9 [12, 6-28] mm) and had greater LCC (21.114.9 [16, 1-85] vs. 13.66.1 [11.5, 4-30] mm); P<0.001 and 0.002, respectively. AU-ROC for size was 0.73 (0.64-0.80) and LCC was 0.69 (0.60-0.76), respectively. Optimal SENS/SPEC for diagnosis of EPE were: size 15mm=67.7/66.7% and LCC 11mm=84.9/44.8%. 10(th)-centile ADC and ADC entropy were both associated with EPE (P=0.02 and<0.001), with AU-ROC=0.56 (0.47-0.65) and 0.76 (0.69-0.83), respectively. Optimal SENS/SPEC for diagnosis of EPE with entropy 6.99 was 63.3/75.0%. 25(th)-centile ADC trended towards being significantly lower with EPE (P=0.06) with no difference in other ADC metrics (P=0.25-0.88). Size, LCC, and ADC entropy improved sensitivity but reduced specificity compared with subjective analysis with no difference in overall accuracy (P=0.38). ConclusionMeasurements of tumor size, capsular contact, and ADC entropy improve sensitivity but reduce specificity for diagnosis of EPE compared to subjective assessment. Level of Evidence: 3 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2018;47:176-185.