Clinicians and researchers interested in detection and staging of prostate cancer.To introduce an MRI technique for detection and grading of prostate tumors, and characterization of prostatic tissue.MR multi-exponential T₂ mapping is a well-known imaging technique that has been applied for tracking histo-pathological changes in organs such as brain¹. Evidence of multi-exponential T₂ decay from prostatic tissue has been presented before², but the suitability of this technique for diagnosis of prostate cancer has not been studied in details. Using multi-exponential T₂ mapping, the fractional volume of the luminal space, or so called luminal water fraction (LWF), in the prostatic tissue can be determined. Because of the difference of tissue composition and lumen percentage in normal and cancerous prostatic tissues, we hypothesized that MR multi-exponential T₂ mapping can be used for the detection and staging of prostate cancer. We have acquired and analyzed MR images of 11 patients, and investigated the accuracy of this technique in detection and evaluation of prostatic tumors by performing multi-parametric statistical analysis.Data acquisition was carried out at the University of British Columbia (UBC) MRI Research Centre, using a 3.0T whole body MR scanner [Achieva 3.0T, Philips Medical Systems, Best, The Netherlands]. Eleven patients with biopsy proven cancer underwent MRI scan with an endorectal coil, prior to undergoing prostatectomy. A 3D multi-echo spin echo sequence (TR/TE=3061/25ms, NE=64, FOV=240x240x40mm³, voxel-size=1x1x4mm³, matrix-size=240x240) was used for scanning of the entire prostate gland. Images were analyzed with Matlab [The MathWorks Inc,Natick, MA, USA]. The analysis involved regularized Non-Negative Least Squares (NNLS)^3,4 fitting of multi-exponential decay curves, which generated T₂ distributions for every pixel (see Figure 1). The following parameters were defined and used to describe the T₂ distributions: number of distinguishable T₂ components (N) determined by counting the number of peaks in the distribution; geometric mean of the short (T_2-short) and long (T_2-long) components, as well as the geometric mean of the entire distribution (gmT₂); ratio of area under the long component over the total area under the entire distribution (Luminal Water Fraction – LWF); and areas under the short (A₁) and long (A₂) components. Average values of these parameters were calculated within 255 ROIs manually outlined on digitized images of the whole-mount histology sections registered to MRI images⁵. Selection of ROIs included: cancerous peripheral (PZ) and transition (TZ) zones, and normal PZ, TZ, Anterior Fibromuscular Stroma (AFMS), and Periurethral Fibromuscular stroma (PFMS). Statistical analyses were performed with MedCalc [MedCalc Software, Mariakerke, Belgium]. Significant differences between MR parameters of tumor and normal tissues were determined with Kruskal-Wallis test. Correlations between MR parameters and Gleason score were evaluated in PZ and TZ with Spearman’s rank correlation test. Sensitivity and specificity were calculated by Receiver Operating Characteristic (ROC) analysis of individual and combined MR parameters. Representative MR parametric maps are shown in Figure 2. Kruskal-Wallis test indicated that the average values of T_2-short, gmT₂, A₁, A₂, and LWF were significantly different between tumor and normal tissue in PZ and TZ, suggesting that these five parameters can be used as tumor indicator in prostate. The average values of N were significantly different between the glandular tissue (i.e. PZ and TZ) and non-glandular tissue (i.e. AFMS and PFMS); hence, N can be used for tissue classification and machine learning purposes. The highest correlation with Gleason score was obtained for LWF (-0.734, p <0.001 in PZ, and -0.712, p <0.001 in TZ). The values of sensitivity, specificity, and Area Under the ROC Curve (AUC) demonstrated high accuracy of tumour detection with the proposed technique (Table 1). The results of this pilot study demonstrate the suitability of MR multi-exponential T₂ mapping for prostate cancer diagnosis. Several parameters were defined to characterize T₂ decay curve of prostatic tissue, and it was shown that the defined parameters were successfully applied for prostate tissue classification, detection of its cancerous tissue with high sensitivity and specificity, and grading of prostate cancer.