MO-FG-CAMPUS-IeP3-03: Optimal Parameters to Determine the Apparent Diffusion Coefficient in Diffusion Weighted Imaging Via Simulation

Abstract

Purpose: To investigate in silico how calculation of apparent diffusion coefficient (ADC) is affected by image SNR, b-values, and the true tissue ADC. Methods: Diffusion-weighted (DW) images were generated assuming a mono-exponential signal model with two different b-values and known true ADC values. Rician noise of different levels was added to the DWI images to adjust the image SNR. Using the two DWI images, ADC was calculated using a mono-exponential model for each set of b-values, SNR, and true ADC. 40,000 simulations were performed for each parameter setting to determine the mean and the standard-deviation of the calculated ADC, as well as the percentage accuracy and precision with respect to the true ADC. In our study, we simulated two true ADCs (ADC 0.00102 for prostate tumor and 0.00180 mm2/s for normal prostate tissue). SNR was varied from 2 to 100 and b-values were varied from 0 to 2000s/mm2. Results: The accuracy (difference between known and calculated ADC) and precision (standard-deviation of calculated ADC) were increased with SNR. To increase SNR, 10 signal-averagings (NEX) were used. The optimal NEX combination for tumor and normal tissue was 2, 8, and the optimal b-value pair were 0, 1200s/mm2 and 0, 700s/mm2. The minimum accuracy and precision errors were obtained when low b-value is 0. For any choice of low b-value until 400s/mm2, the errors were minimized when b-value difference is approximately 1200 and 700s/mm2 for tumor and normal tissues respectively. For true ADC 0.00144 mm2/s, it is around 700s/mm2. Conclusion: Using the optimal choices of parameters for the prostate peripheral region both normal and tumor tissues were calculated with 99% accuracy and 94% precision. Suggestions for parameters choices could be made for any tissue ADC.