Characterization and stratification of prostate lesions based on comprehensive multiparametric MRI using detailed whole‐mount histopathology as a reference standard. (29th September 2017)

• Record Type: Journal Article
• Authors: Starobinets, Olga; Simko, Jeffry P.; Kuchinsky, Kyle; Kornak, John; Carroll, Peter R.; Greene, Kirsten L.; Kurhanewicz, John; Noworolski, Susan M.
• Is Part Of: NMR in biomedicine. Volume 30:Number 12(2017:Dec.)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2017-09-29
• Subjects: diffusion‐weighted imaging -- dynamic contrast‐enhanced imaging -- histopathology -- magnetic resonance spectroscopy imaging -- multiparametric MRI -- prostate cancer; Nuclear magnetic resonance -- Periodicals; Magnetic Resonance Spectroscopy -- Periodicals; 574
• DOI: 10.1002/nbm.3796
• Languages: English
• ISSNs: 0952-3480

Abstract

Abstract : The purpose of this study was to characterize prostate cancer (PCa) based on multiparametric MR (mpMR) measures derived from MRI, diffusion, spectroscopy, and dynamic contrast‐enhanced (DCE) MRI, and to validate mpMRI in detecting PCa and predicting PCa aggressiveness by correlating mpMRI findings with whole‐mount histopathology. Seventy‐eight men with untreated PCa received 3 T mpMR scans prior to radical prostatectomy. Cancerous regions were outlined, graded, and cancer amount estimated on whole‐mount histology. Regions of interest were manually drawn on T 2 ‐weighted images based on histopathology. Logistic regression was used to identify optimal combinations of parameters for the peripheral zone and transition zone to separate: (i) benign from malignant tissues; (ii) Gleason score (GS) ≤3 + 3 disease from ≥GS3 + 4; and (iii) ≤ GS3 + 4 from ≥GS4 + 3 cancers. The performance of the models was assessed using repeated fourfold cross‐validation. Additionally, the performance of the logistic regression models created under the assumption that one or more modality has not been acquired was evaluated. Logistic regression models yielded areas under the curve (AUCs) of 1.0 and 0.99 when separating benign from malignant tissues in the peripheral zone and the transition zone, respectively. Within the peripheral zone, combining choline, maximal enhancement slope, apparent diffusion coefficient (ADC), and citrate measures for separating ≤GS3 + 3 from ≥GS3 + 4 PCa yielded AUC = 0.84. Combining creatine, choline, and washout slope yielded AUC = 0.81 for discriminating ≤GS3 + 4 from ≥GS4 + 3 disease. Within the transition zone, combining washout slope, ADC, and creatine yielded AUC = 0.93 for discriminating ≤GS3 + 3 and ≥GS3 + 4 cancers. When separating ≤GS3 + 4 from ≥GS4 + 3 PCa, combining choline and washout slope yielded AUC = 0.92. MpMRI provides excellent separation between benign tissues and PCa, and across PCa tissues of different aggressiveness. The final models prominently feature spectroscopy and DCE‐derived metrics, underlining their value within a comprehensive mpMRI examination. Abstract : This study aimed to characterize prostate cancer (PCa) based on multiparametric MR (mpMR) measures and to validate mpMRI in detecting PCa and predicting PCa aggressiveness through correlations with whole‐mount histopathology. MpMRI provided excellent separation of benign tissues and PCa, and across PCa tissues of different aggressiveness in the peripheral and in the transition zones, even for small lesions. The final models prominently featured spectroscopy and dynamic contrast‐enhanced imaging metrics, underlining the value of these measures within a comprehensive mpMRI examination.