Study population: We reviewed the medial records and imaging studies of the patients who underwent prostatectomy for prostate cancer from January 2015 to October 2016 and performed preoperative multiparamteric prostate MRI in our institution. Total 50 patients with 3T MR imaging including DCE-MRI and wash-in color-map and pathologically confirmed prostate cancer were retrospectively analyzed.

Inline wash-in parameter images: The standard option of the MR machine includes automatic generation of parametric DCE color coded wash-in maps after imaging acquisition. These maps reflect percentage of enhancement based on signal intensity changes between the first image and the image obtained one minute after contrast infusion and. The maximum value of percentage change in signal intensity was selected to display on the parametric map. The degree of percentage enhancement was coded using different color; high value as red and low value as blue.

Image analysis: Two radiologists reviewed randomly ordered either color-map or DCE-whole-images in two different sessions separated by at least 2 weeks, without any information about pathologic result of the patients. Reader 1 (R1) was experienced radiologist with specialty in abdominal radiology and reader 2 (R2) was second grade trainee in radiology. Reader 1 selected one focal lesion with T2-low signal intensity in each patient and indicated the lesion using an arrow. Each reader was asked 1) to grade the confidence of enhancement of the marked focal lesion from grade 1 (definite no enhancement) to grade 6 (definite enhancement) and 2) to record the time consumed for evaluation. The results of DCE-whole-images by R1 were regarded as gold-standard. R1 started to review images one month after lesion selection. Fig 1 showed the images of a representative case.

Statistical Analysis: Confidence grades of the gold-standard were compared to results of wash-in color-map by two readers and DCE-whole-images R2 using weighted kappa statistics. The consumed time for evaluating two image sets by two readers were compared using paired t-test; the differences were compared between readers and between methods.