Simple models improve the discrimination of prostate cancers from the peripheral gland by T1-weighted dynamic MRI

Evaluation of the accuracy of descriptive and physiological parameters calculated from signal intensity-time curves using T1-weighted dynamic contrast enhanced MRI (DCE MRI) to differentiate prostate cancers from the peripheral gland. Twenty-seven patients with prostate cancers were examined with DCE MRI prior radical prostatectomy. Regions of interest were defined in tumors and non-affected areas in the peripheral zone. Dynamic data were parameterized in amplitude and exchange rate constant (k(ep)) using a two-compartment model. Additionally, relative slope during 26, 39, 52 and 65 s, areas under the curve (AUC) and time to start of signal intensity increase (t(lag)) were determined. Vessel density (VD) of excised prostates was quantified in tumor areas using a CD34 stain. The parameter slope(52) showed 20 % higher values (P < 0.001) in tumors than in the peripheral gland and compared with the other parameters the largest area under the ROC curve (0.81). The minimum total error rate was attained at a cut-point of 0.021, yielding a sample value of sensitivity and specificity of 70 % and 88 %, respectively, and a bias-corrected sum of sensitivity and specificity of 1.54. In addition, amplitude (P < 0.001), k(ep) (P = 0.03) and AUC (P < 0.001) were significantly higher in tumors. t(lag) did not discriminate carcinomas from glandular tissue. VD was higher in tumors than in the non-affected peripheral prostate (P = 0.05). However, none of the dynamic parameters in carcinomas showed a significant correlation with VD or Gleason score. Although pharmaco-kinetic modeling in DCE MRI showed potential to discriminate prostate cancers from peripheral prostate tissue, descriptive parameters of the early signal enhancement after contrast media injection reached higher sensitivity and specificity.