A framework to analyze partial volume effect on gray matter mean diffusivity measurements

Analyzing gray matter diffusion properties can be challenging due to possible measurement biases originating from averaging of gray matter (GM) and cerebrospinal fluid (CSF) signals. Therefore, a better characterization of CSF contamination effects in different cortical regions is required in order to disentangle actual changes in microstructure of GM itself from changes due to other effects such as macroscopic morphological changes. We propose a localized analysis framework for the CSF contamination effect on GM mean diffusivity measurement and applied this framework to measurements on 15 subjects. Our proposed modeling framework was compared to fluid-attenuated inversion recovery (FLAIR) DTI technique from the same subjects. The results of our studies suggest that GM mean diffusivity value was significantly biased by the CSF contamination effect, and that the amount of contamination strongly depended on the local morphology of the peripheral brain. Expected biases had their maxima in the motor and the somatosensory association cortex, and their minima in mid and inferior temporal areas of the brain where the cortical thicknesses are particularly pronounced. We conclude from our studies that regional differences in tissue compounding ratio must be taken into account when assessing localized GF diffusivity differences.