Impact of post-processing methods on apparent diffusion coefficient values

The apparent diffusion coefficient (ADC) is increasingly used as a quantitative biomarker in oncological imaging. ADC calculation is based on raw diffusion-weighted imaging (DWI) data, and multiple post-processing methods (PPMs) have been proposed for this purpose. We investigated whether PPM has an impact on final ADC values. Sixty-five lesions scanned with a standardized whole-body DWI-protocol at 3 T served as input data (EPI-DWI, b-values: 50, 400 and 800 s/mm(2)). Using exactly the same ROI coordinates, four different PPM (ADC_1-ADC_4) were executed to calculate corresponding ADC values, given as [10(-3) mm(2)/s] of each lesion. Statistical analysis was performed to intra-individually compare ADC values stratified by PPM (Wilcoxon signed-rank tests: alpha = 1 %; descriptive statistics; relative difference/a dagger; coefficient of variation/CV). Stratified by PPM, mean ADCs ranged from 1.136-1.206 *10(-3) mm(2)/s (a dagger = 7.0 %). Variances between PPM were pronounced in the upper range of ADC values (maximum: 2.540-2.763 10(-3) mm(2)/s, a dagger = 8 %). Pairwise comparisons identified significant differences between all PPM (P <= 0.003; mean CV = 7.2 %) and reached 0.137 *10(-3) mm(2)/s within the 25th-75th percentile. Altering the PPM had a significant impact on the ADC value. This should be considered if ADC values from different post-processing methods are compared in patient studies. Post-processing methods significantly influenced ADC values. The mean coefficient of ADC variation due to PPM was 7.2 %. To achieve reproducible ADC values, standardization of post-processing is recommended.