Supervised risk predictor of central gland lesions in prostate cancer using 1 H MR spectroscopic imaging with gradient offset-independent adiabaticity pulses
Neda Gholizadeh,Peter B Greer,John Simpson,Caixia Fu,Oun Al-Iedani,Peter Lau,Arend Heerschap,Saadallah Ramadan
DOI: https://doi.org/10.1002/jmri.26803
Abstract:Background: Due to the histological heterogeneity of the central gland, accurate detection of central gland prostate cancer remains a challenge. Purpose: To evaluate the efficacy of in vivo 3D 1 H MR spectroscopic imaging (3D 1 H MRSI) with a semi-localized adiabatic selective refocusing (sLASER) sequence and gradient-modulated offset-independent adiabatic (GOIA) pulses for detection of central gland prostate cancer. Additionally four risk models were developed to differentiate 1) normal vs. cancer, 2) low- vs. high-risk cancer, 3) low- vs. intermediate-risk cancer, and 4) intermediate- vs. high-risk cancer voxels. Study type: Prospective. Subjects: Thirty-six patients with biopsy-proven central gland prostate cancer. Field strength/sequence: 3T MRI / 3D 1 H MRSI using GOIA-sLASER. Assessment: Cancer and normal regions of interest (ROIs) were selected by an experienced radiologist and 1 H MRSI voxels were placed within the ROIs to calculate seven metabolite signal ratios. Voxels were split into two subsets, 80% for model training and 20% for testing. Statistical tests: Four support vector machine (SVM) models were built using the training dataset. The accuracy, sensitivity, and specificity for each model were calculated for the testing dataset. Results: High-quality MR spectra were obtained for the whole central gland of the prostate. The normal vs. cancer diagnostic model achieved the highest predictive performance with an accuracy, sensitivity, and specificity of 96.2%, 95.8%, and 93.1%, respectively. The accuracy, sensitivity, and specificity of the low- vs. high-risk cancer and low- vs. intermediate-risk cancer models were 82.5%, 89.2%, 70.2%, and 73.0%, 84.7%, 60.8%, respectively. The intermediate- vs. high-risk cancer model yielded an accuracy, sensitivity, and specificity lower than 55%. Data conclusion: The GOIA-sLASER sequence with an external phased-array coil allows for fast assessment of central gland prostate cancer. The classification offers a promising diagnostic tool for discriminating normal vs. cancer, low- vs. high-risk cancer, and low- vs. intermediate-risk cancer. Level of evidence: 2 Technical Efficacy: Stage 2 J. Magn. Reson. Imaging 2019;50:1926-1936.