PURPOSE: To investigate the accuracy and interobserver variability of a standardized evaluation system for endorectal three-dimensional (3D) magnetic resonance (MR) spectroscopic imaging of the prostate. MATERIALS AND METHODS: The human research committee approved the study, and all patients provided written informed consent. Endorectal MR imaging and MR spectroscopic imaging were performed in 37 patients before they underwent radical prostatectomy. For the 22 patients with good or excellent MR spectroscopic imaging data, step-section histopathologic tumor maps were used to identify spectroscopic voxels of unequivocally benign (n = 306) or malignant (n = 81) peripheral zone tissue. Two independent spectroscopists, unaware of all other findings, scored the spectra of the selected voxels by using a scale of 1 (benign) to 5 (malignant) that was based on standardized metabolic criteria. Descriptive statistical, receiver operating characteristics (ROC), and kappa statistical analyses of the data obtained by both readers were performed by using two definitions of cancer: one based on a voxel score of 3-5 and the other based on a score of 4 or 5. RESULTS: The scoring system had good accuracy (74.2%-85.0%) in the differentiation between benign and malignant tissue voxels, with areas under the ROC curve of 0.89 for reader 1 and 0.87 for reader 2. Specificities of 84.6% and 89.3% were achieved when a voxel score of 4 or 5 was used to identify cancer, and sensitivities of 90% and 93% were achieved when a score of 3-5 was used to identify cancer. Readers demonstrated excellent interobserver agreement (kappa values, 0.79 and 0.80). CONCLUSION: The good accuracy and excellent interobserver agreement achieved by using the standardized five-point scale to interpret peripheral zone metabolism demonstrate the potential effectiveness of using metabolic information to identify prostate cancer, and the clinical usefulness of this system warrants testing in prospective clinical trials of MR imaging combined with MR spectroscopic imaging. (c) RSNA, 2004.
ASJC Scopus subject areas
- Radiology Nuclear Medicine and imaging