Prostate cancer: Comparison of dynamic contrast-enhanced MRI techniques for localization of peripheral zone tumor

OBJECTIVE. The objective of this study was to compare the performance of different methodologies for interpretation of dynamic contrast-enhanced MRI (DCE-MRI) in localization of peripheral zone prostate cancer. MATERIALS AND METHODS. Forty-three men (mean age, 59 ± 8 years) with biopsyproven prostate cancer who underwent prostate MRI including DCE-MRI before prostatectomy were included. Two observers independently reviewed DCE-MRI data using three methodologies: qualitative, in which kinetic curves of signal intensity versus time were generated for foci showing rapid enhancement on subtracted contrast-enhanced images; semiquantitative, in which a biexponential heuristic model was used to generate color-coded maps depicting maximum slope and washout of contrast enhancement; and quantitative, in which a Tofts model was used to generate color-coded influx rate transfer constant (Ktrans) and efflux rate transfer constant (Kep) maps. Findings were stratified by whether suspicious foci showed evidence of washout with each method and compared with histopathologic results in each sextant. RESULTS. There was similar accuracy for the semiquantitative and quantitative models for both observers irrespective of requiring evidence of washout. For the more experienced observer, requiring washout resulted in lower sensitivity and higher specificity for the qualitative and semiquantitative models. Also for the more experienced observer, use of either a semiquantitative or quantitative model provided greater sensitivity compared with a qualitative model when requiring washout. There was no association between tumor detection and Gleason score for any DCE-MRI methodology for either reader. CONCLUSION. For the experienced reader, sensitivity for peripheral zone tumor was increased by use of either a semiquantitative or quantitative model compared with a qualitative model and decreased by requiring washout. We failed to identify a difference in performance between semiquantitative and quantitative models.