Erik S. Anderson PhD, Presenter: Nothing to Disclose

Daniel Jason Aaron Margolis MD, Abstract Co-Author: Research Grant, Siemens AG

Mitchell Kamrava MD, Abstract Co-Author: Nothing to Disclose

PURPOSE/OBJECTIVE:To assess whether performing an apparent diffusion coefficient (ADC) histogram analysis on regions of interest (ROI) on prostate MRI will identify values that more strongly correlate with Gleason score (GS) than mean ADC value.MATERIALS AND METHODS:We performed a retrospective analysis of 19 prostate lesions from 17 individuals that were identified on multi-parametric MRI that included diffusion weighted imaging (DWI). The ROI was drawn on the most conspicuous slice of the DWI image. We then extracted the ADC values on a voxel by voxel basis, to assemble the entire distribution of ADC values within the ROI. We calculated ADC characteristics for each ROI, including mean, median, minimum, maximum, standard deviation, skewness, kurtosis and percentile values. We then assessed whether each of these parameters significantly correlated to GS based on targeted biopsy using the ANOVA statistical test.RESULTS:Median patient characteristics included: age 61 years (SD 6), PSA 8.1 (SD 12.8), prostate volume 43 cc (SD 14.2), and ROI volume 0.4 cm2 (SD 0.82). GS ranged from 6-9 (GS 6, n=8; GS 7, n=6; GS 9, n=5).Thirteen lesions were located within the peripheral zone. For peripheral zone lesions, multiple parameters were statistically correlated with increasing Gleason score including ADC median (p=0.02), 5th percentile (p=0.05), 10th percentile (p=0.03) and 25th percentile (p=0.02). Association of GS with ADC mean, standard deviation, minimum, maximum, skewness, kurtosis and additional percentile ranks of 75% and 90% did not achieve statistical significance. Associations remained significant when including all 19 lesions from the peripheral, central and transitional zones.CONCLUSIONS:Using ADC histogram analysis, the characteristics that achieved the highest correlation with GS were the median, 10th and 25th percentile values. This finding implies that the most predictive measurements within a ROI are the voxels that display the lowest diffusion, as opposed to the often reported mean. Future work will analyze the spatial relationship of low ADC voxels to determine whether anatomical sub-regions of low diffusion could represent targets for focused biopsy and radiation boost.

Anderson, E, Margolis, D, Kamrava, M, Does the ADC Mean Tell the Whole Story?: Using ADC Histogram Analysis to Determine the Ideal DWI Metric for Correlation with Gleason Score.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14042316.html