Article révisé par les pairs
Résumé : Objectives: A prospective multicenter validation of the ability of H magnetic resonance spectroscopic imaging (MRSI) to distinguish cancer from noncancer tissues throughout the prostate with histopathology of the resected organ as the standard of reference. Materials and methods: Institutional review board approval was obtained for all centers and all participating patients and volunteers provided written informed consent. Ninety-nine patients and 10 age-matched volunteers from 8 participating centers underwent magnetic resonance imaging and 3-dimensional MRSI with an endorectal coil at 1.5 T. Selected MRSI voxels were assigned to the peripheral zone (PZ), the central gland (CG), the periurethral area, and cancer tissue. Signal ratios of choline + creatine to citrate (CC/C) in spectra of these voxels were automatically calculated. Receiver operating characteristic curves were constructed to assess the accuracy by which this ratio can discriminate cancer from noncancer tissue. Results: A total of 70% of voxels in noncancer tissue and 90% of voxels in cancer tissue passed the quality check of the automatically fitted spectra. The median CC/C was significantly different between any noncancer and cancer tissue (P < 0.0001), but not between the different contributing centers. CC/C increased with cancer focus size (P = 0.0008) and certainty of voxel mapping to histopathologic cancer site (P < 0.0001). The area under the receiver operating characteristic curve for discriminating voxels of cancer tissue from noncancer tissue was 0.88 (confidence interval: 0.84-0.92) in the PZ and 0.76 (confidence interval: 0.71-0.81) in the CG. Conclusions: In patients with prostate cancer, recruited from different institutions, 3-dimensional MRSI is demonstrated to be a robust and quantitative technique, producing significantly different CC/C values for cancer compared with noncancer tissue for both the CG and the PZ. © 2010 by Lippincott Williams & Wilkins.

Documents en relation

DI-fusion