Anatomical determinants of DTI-ALPS: effects of ROI definition, ventricular morphology, and periventricular...

The diffusion tensor image analysis along the perivascular space (DTI-ALPS) index is increasingly used as a non-invasive MRI biomarker of glymphatic and perivascular function, yet the anatomical validity and measurement stability of the metric remain incompletely characterised. Using diffusion MRI data from 850 healthy young adults and 150 healthy ageing participants from the Human Connectome Project, I systematically evaluated the influence of region-of-interest (ROI) placement and ventricular anatomy on ALPS measurements. Reference ALPS implementations demonstrated substantial hemispheric variability, with a median left-right difference of 12.5% and marked asymmetry in the underlying numerator and denominator tensor components. A two-stage optimisation framework incorporating fibre-pool alignment, hemispheric symmetry, component stability, and directional purity identified anatomically improved ROI configurations that significantly increased fibre specificity and reduced measurement variability in independent validation cohorts. Despite these improvements, residual hemispheric asymmetry persisted, suggesting an intrinsic anatomical contribution to ALPS variability. In the healthy ageing cohort, ventricular volume emerged as the strongest predictor of ALPS, explaining substantially more variance than chronological age. Voxel-wise deformation-based morphometry demonstrated that lower ALPS values were associated with ventricular and periventricular expansion, while optimisation increased coupling between ALPS and ventricular anatomy. Collectively, these findings indicate that ALPS measurements are strongly influenced by ROI definition, ventricular morphology, and surrounding periventricular tissue architecture. Rather than functioning as a direct measure of glymphatic transport in isolation, ALPS appears to represent a composite anatomical diffusion biomarker shaped by both methodological implementation and underlying neuroanatomy. These results provide a framework for improving methodological standardisation and interpretation of ALPS measurements in future neuroimaging studies.