Quantitative (1) H MR spectroscopic imaging of the prostate gland using LCModel and a dedicated basis-set: correlation with histologic findings

Abstract

Proton magnetic resonance spectroscopic imaging ((1) H-MRSI) has been advocated as a valuable tool for prostate cancer diagnosis. However, a barrier to widespread clinical use of this technique is the lack of robust quantification methods that yield reproducible results in an institution-independent manner. The main goal of this study was to develop a standardized and fully automated approach (LCModel-based) for quantitative prostate (1) H-MRSI. To this end, a dedicated basis set was constructed by the combination of simulated (citrate, Cit; choline, Cho, and creatine, CR) and experimentally acquired (spermine, Spm) spectra. The overlapping Spm, Cho, and Cr could be resolved and quantified individually, thus allowing for the independent assessment of glandular (Cit and Spm) and proliferative (Cho) components. Several metabolite ratios were calculated and compared to the histologic findings of prostatectomy specimens from 10 prostate cancer patients with Gleason scores (3 + 3) and (3 + 4). The Cho mole fraction and the Cho/(Cit + Spm) ratio were found to best discriminate between prostate cancer and healthy tissue. The comparison between the quantitative MRSI results and the histologic findings suggests that no correlation exists between the detected metabolic alterations and the Gleason score of low-grade tumors.