To evaluate reproducibility of voxel-based apparent diffusion coefficient (ADC).ADC of the phantom containing 0.2 mM gadolinium and 80% polyvinyl alcohol (Nikkofines (90-401 type), Tokyo, Japan) was measured in a whole-body 3T clinical MRI system (Ingenia, Philips Medical Systems, Best, The Netherland). The measurements were performed three times for diffusion weighted imaging with both single shot echo planar imaging (DWI-EPI) (Fig. 1) and multi-shot turbo spin echo (DWI-TSE) (Fig. 2) . 1^st and 2^nd examinations were performed at an identical positioning and 3^rd examination was done at a slightly different positioning. The sequence parameters were as follows: field of view=220 mm, matrix size=112x112, number of averages=4, slice thickness=4 mm, SENS factor=2, repetition time=3000 ms, echo time=81.1 ms for DWI-EPI and 87.6 ms for DWI-TSE, b-factor=0 and 1000 s/mm², three orthogonal motion proving gradients (MPG). Each voxel ADC was calculated by fitting signal intensity change within the region of interest containing 197 voxels into mono-exponential curve. ADCs measured by DWI-EPI and DWI-TSE with right-to-left, anterior-to-posterior, foot-to-head motion probing gradients, and three synthesized imaging were described as ADC-EPI-RL, ADC-EPI-AP, ADC-EPI-FH, ADC-EPI-ISO, ADC-TSE-RL, ADC-TSE-AP, ADC-TSE-FH, and ADC-TSE-ISO, respectively. And ADCs from 1^st, 2^nd, and 3^rd examinations were differentiated by adding 1, 2, and 3 at the end, for example ADC-EPI-RL1 means ADC calculated by 1^st DWI-EPI with right-to-left MPG. The voxel-based ADCs measured by several sequences were compared using paired t-test and t-test for evaluating MPG direction effect (ADC-EPI-RL vs. ADC-EPI-AP vs. ADC-EPI-FH, and ADC-TSE-RL vs. ADC-TSE-AP vs. ADC-TSE-FH) and reproducibility using both identical (1^st ADC-EPI vs. 2^nd ADC-EPI, and 1^st ADC-TSE vs. 2^nd ADC-TSE) and not identical positioning (1^st ADC-EPI vs. 3^rd ADC-EPI, and 1^st ADC-TSE vs. 3^rd ADC-TSE). p<0.05 was considered statistically significant.Regarding effect of different MPG directions (Fig. 3), significant difference was observed between ADC-EPI-RL1 and ADC-EPI-AP1, between ADC-EPI-AP1 and ADC-EPI-FH1, between ADC-TSE-RL1 and ADC-TSE-AP1, and between ADC-TSE-RL1 and ADC-TSE-FH1. Regarding reproducibility for identical positioning (Fig. 4), significant difference was observed between ADC-EPI-RL1 and ADC-EPI-RL2, between ADC-EPI-ISO1 and ADC-EPI-ISO2, and between ADC-TSE-FH1 and ADC-TSE-FH2. Regarding reproducibility for different positioning (Fig. 5), significant difference was observed in all DWI-EPI and DWI-TSE comparisons between 1^st and 3^rd examinations except between ADC-TSE-AP1 and ADC-TSE-AP3.

The present study demonstrated that voxel-based ADC could not be reproduced using even phantom. Many studies have been reported regarding histogram analysis based on voxel-based ADC distribution^1-4. We consider enough attention should be payed when analyzing voxel-based ADC distribution.

Voxel-based ADC values differed among different MPG directions in both DWI-EPI and DWI-TSE. As phantom was considered having no anisotropy, these differences could be originated in the difference of MPG strength and/or image distortion. Voxel-based ADC values differed between 1^st and 2^nd identical position examinations. As DWI-EPI showed significant difference in 2 comparisons, between ADC-EPI-RL1 and ADC-EPI-RL2 and between ADC-EPI-ISO1 and ADC-EPI-ISO2 but DWI-TSE showed significance in only between ADC-TSE-FH1 and ADC-TSE-FH2, this difference may be due to strong distortion artifact in DWI-EPI and/or low signal-to noise ratio of DWI-TSE. Voxel-based ADC values differed between 1^st and 3^rd slightly different position examinations even in synthesized images. We consider this result should be kept in mind when analyzing sequential voxel-based ADC histogram changes of tumor, e.g. pre- and post-chemoradiotherapy.

The reproducibility of voxel-based ADC is not high enough even in a phantom study. It would be recommended to pay enough attention when performing voxel-based ADC study like histogram analysis for tumor ADC.