Purpose: To evaluate the length of capsular contact of dominant lesions on multiparametric prostate magnetic resonance imaging (MRI) for predicting extraprostatic extension (EPE) and to determine a threshold value to apply in clinical practice.
Materials and methods: Ninety patients undergoing 3T prostate MRI before prostatectomy were included. Two independent readers (R1, R2) recorded for each lobe the presence or absence of capsular irregularity on T2 -weighted imaging (T2 WI) and of overt measurable EPE. Readers also recorded the length of capsular contact of each lobe's dominant lesion for T2 WI and the apparent diffusion coefficient (ADC) map. Based on prostatectomy specimens, EPE was recorded for each lobe and classified as focal (single focus ≤0.5 mm in depth) vs. established. Receiver operating characteristic analysis, logistic regression, and kappa coefficients were used to assess interpretive approaches on a side-specific basis.
Results: The optimal thresholds were 6 mm and 7 mm of contact using T2 WI and ADC for any EPE, and 10 mm and 7 mm using T2 WI and ADC for nonfocal EPE (AUCs 81.0-82.5%). Capsular contact had higher sensitivity, yet lower specificity, than subjective interpretations for any EPE and for nonfocal EPE (all P ≤ 0.018, aside from any EPE for R2 using ADC). Length of contact exhibited more substantial gains in sensitivity (9-20% for any EPE; 34-41% for nonfocal EPE) than losses in specificity (6-13% for any EPE; 17-27% for nonfocal EPE) compared with subjective interpretations. Interreader agreement: 0.70 for assessments based on length of contact; 0.49-0.59 for subjective assessments.
Conclusion: Length of capsular contact of dominant lesions can improve interreader agreement and sensitivity for EPE compared with subjective features, with relatively mild specificity loss.
Keywords: MRI; diffusion-weighted imaging; prostate cancer; staging.
© 2015 Wiley Periodicals, Inc.