Reliability and Spatial Consistency of MR Diffusion Tensor Imaging Measures Along the Cerebral Perivascular Space

Diffusion tensor imaging analysis along the perivascular space (DTI-ALPS) has emerged as a promising noninvasive method for evaluating water motion that may reflect glymphatic system function. However, the reliability of DTI-ALPS measurements across different region-of-interest (ROI) selection methods remains underinvestigated. This study aimed to assess the interrater reliability among three neuroradiologists in native space and compare DTI-ALPS indices derived from ROIs placed in subjects' native space versus standardized Montreal Neurological Institute (MNI) space.

DTI-ALPS indices from 16 healthy subjects were calculated from both left and right hemispheres using two ROI placement approaches: (1) native space ROIs manually placed by three neuroradiologists, and (2) standardized ROIs in MNI space based on the fractional anisotropy template. Interrater reliability was assessed using intraclass correlation coefficients (ICCs). The proportion of ROI overlaps among the three neuroradiologists was also evaluated. Differences between native and MNI space measurements were evaluated using related-samples Friedman's analysis with post hoc pairwise comparisons.

Interrater reliability for native space ROI placement was moderate for left-sided DTI-ALPS indices (ICC = 0.599) and good for right-sided DTI-ALPS indices (ICC = 0.807). Spatial overlap analysis revealed poor Dice similarity coefficients across all ROI types (range: 0.047-0.312), with right association ROIs showing higher spatial consistency. Significant differences were found between native and MNI space measurements for left-sided DTI-ALPS indices (p = 0.002) but not for right-sided DTI-ALPS indices (p = 0.913).

These findings highlight the importance of standardized ROI selection approaches for clinical applications of DTI-ALPS.