Diffusion tensor imaging along the perivascular space: the bias from crossing fibres.

Abstract

Non-invasive evaluation of glymphatic function has emerged as a crucial goal in neuroimaging, and diffusion tensor imaging along the perivascular space (DTI-ALPS) has emerged as a candidate method for this purpose. Reduced ALPS index has been suggested to indicate impaired glymphatic function. However, the potential impact of crossing fibres on the ALPS index has not been assessed, which was the aim of this cross-sectional study. For this purpose, we used DTI-ALPS in a cohort with three groups: Parkinson's disease (PD) (n = 60, mean age 63.3 ± 1.5, 33 males), progressive supranuclear palsy (PSP) (n = 17, mean age 70.9 ± 1.5, 9 males) and healthy controls (n = 41, mean age 64.5 ± 8.4, 15 males). The ALPS index was calculated blinded to diagnosis, by manually placing two sets of regions of interest (ROI) on the projection and association fibres of each hemisphere. Annotation was performed twice: once on conventional diffusion-encoded colour maps weighted by fractional anisotropy and once on maps with weights adjusted for high incidence of crossing fibres. PSP patients had significantly lower conventional ALPS indices compared with both healthy controls (right hemisphere: P = 0.009; left hemisphere: P < 0.001) and PD patients (right hemisphere: P = 0.024; left hemisphere: P < 0.001). There were no differences between healthy controls and PD patients. After adjusting the ROI to avoid regions of crossing fibres, the ALPS index significantly decreased in healthy controls (right hemisphere: P < 0.001; left hemisphere: P < 0.001) and PD (right hemisphere: P < 0.001; left hemisphere: P < 0.001). In PSP, the adjusted ALPS index was lower compared with the conventional one only in the right hemisphere (P = 0.047). Overall, this adjustment led to less significant differences among diagnostic groups. Specifically, with the adjusted ALPS index, PSP patients showed significantly lower ALPS index compared with healthy controls (right hemisphere: P = 0.044; left hemisphere: P = 0.029) and PD patients (P = 0.003 for the left hemisphere only). Our results suggest that crossing fibres significantly inflate the ALPS index and should be considered a critical pitfall of this method. This factor could partly explain the variability observed in previous studies. Unlike previous research, we observed no differences between PD and healthy controls, likely because most patients in our cohort were in the early phase of the disease. Thus, the ALPS index may not be a sensitive indicator of glymphatic function at least in the initial stages of neurodegeneration in PD.