Background
Vascular Dementia (VaD) and Progressive Supranuclear Palsy (PSP) are diseases that
partially overlap regarding their clinical phenotypes and pose challenges in clinical practices and differential diagnoses. The Magnetic Resonance Parkinsonism Index (MRPI) has been developed and being used to support PSP clinical diagnosis even for different clinical variants of PSP [15]. The aims of this study are first, to validate a research application that automates the measurement of MRPI components and then, to identify two sub-phenotypes of VaD based on the presence of a positive or negative MRPI sign, with respect to cerebrovascular lesion load including white matter changes and presence of microbleeds (MBs).
Methods
A cohort of 169 consecutive patients with a diagnosis of VaD (77.83± 6.80 y.o, 45%
females) assessed at the Leenaards Memory Center of the Lausanne University Hospital between August 2013 and February 2023 were retrospectively selected. Inclusion criteria were VaD diagnosis and availability of an MRI, and exclusion criteria was a history of stroke. Seventeen patients were excluded from further analyses because of low MRI quality. Cerebrovascular lesion load was assessed on 152 patients through semi quantitative scales of white matter lesions (Fazekas, ARWMC and MARS scales) [9, 10, 11]. The MRPI sign was calculated using 2 indexes MRPI1.0 and MRPI2.0 based on segmented brain components with a fully automated pipeline using a convolutional neural network (CNN). The cohort was divided into two groups, one with a positive (MRPI+) and the other one with a negative (MRPI-) MRPI sign.
Results
The patients in the MRPI+ sign were older than those in the MRPI- sign, while their global
cognition was similar. The spatial distribution of white matter lesion load was similar between the MRPI- and the MRPI+ groups, with the frontal and parieto-occipital regions being the most affected. However, the basal ganglia was more impacted in the MRPI- group. The MBs location, but not the MBs amount, was key to characterize the two groups. Although the MBs were preferentially located in the lobes for both groups, the MBs occurrence in the cerebellum was almost four times higher in the MRPI+ group than in the MRPI- group.
Conclusions
The MRPI+ group presented an higher MB occurrence in the cerebellum and by less white
matter lesion load in the basal ganglia than the MRPI- group. Although additional research is needed to consolidate our findings, our results suggest that MARS scale for assessing MBs load, ARWMC scale for assessing white matter lesion and MRPI may serve as significant neuroimaging markers offering an added value for VaD/PSP differential diagnosis.