Associations of automatically segmented, enlarged perivascular spaces with neuropathology and cognition in community‐based older adults. (December 2021)

• Record Type: Journal Article
• Authors: Javierre‐Petit, Carles; Tamhane, Ashish A; Evia, Arnold M; Kontzialis, Marinos; Makkinejad, Nazanin; Agam, Gady; Bennett, David A.; Schneider, Julie A.; Arfanakis, Konstantinos
• Is Part Of: Alzheimer's & dementia. Volume 17(2021)Supplement 3
• Page Start: n/a
• Page End: n/a
• Publication Date: 2021-12
• Subjects: Alzheimer's disease -- Periodicals; Alzheimer Disease -- Periodicals; Dementia -- Periodicals; Démence; Maladie d'Alzheimer; Périodique électronique (Descripteur de forme); Ressource Internet (Descripteur de forme); 616.83
• DOI: 10.1002/alz.050294
• Languages: English
• ISSNs: 1552-5260

Abstract

Abstract: Background: Previous studies on the associations between enlarged perivascular spaces (EPVS) and pathology have used manual assessment of EPVS burden, with limitations on cohorts and access to pathology. In this work, we first developed a deep learning model to automatically segment EPVS on MRI. We then combined total and regional EPVS measurements with detailed neuropathology and longitudinal cognitive assessment on 817 community‐based older adults to investigate EPVS links to neuropathology and cognition (Fig. 1). Method: Data: This work included 817 participants from three longitudinal cohort studies of aging. All participants underwent annual cognitive assessment, ex‐vivo T2 ‐weighted imaging (0.6x0.6x1.5 mm) using clinical 3T MRI, and detailed neuropathologic examination by a board‐certified neuropathologist blinded to all data (Fig. 2). Pathologies assessed: gross and microscopic infarcts, atherosclerosis, arteriolosclerosis, cerebral amyloid angiopathy (CAA), amyloid plaques, neurofibrillary tangles, Lewy bodies, and TDP‐43. Segmentation: Data from 10 participants and manually segmented EPVS were used to train convolutional neural networks (Fig. 3) to automatically segment EPVS. The trained networks were used to segment and quantify total and regional EPVS clusters in the ex‐vivo MRI data of all participants. Model performance validation: Using 100 randomly generated ROIs, EPVS were 1) identified by an expert neuroradiologist, and 2) manually segmented by an experienced observer. Neuroradiologist and observer were blind to each other. Model segmentation accuracy and sensitivity were evaluated. Statistical analysis: Associations with neuropathologies, controlling for demographics, were investigated using elastic‐net regularized linear regression. Linear mixed‐effects models were used to investigate independent association of EPVS with cognitive decline independent of neuropathologies and demographics. Result: Model validation showed 68% average sensitivity, DSC=0.66 segmentation accuracy, and high correlation (Pearson, ρ=0.91) in segmentation volume per ROI (Fig. 4). EPVS number was associated with gross infarcts, microscopic infarcts, and diabetes in multiple brain regions, and with CAA in occipital and temporal lobes (Fig. 5). Total and frontal lobe EPVS were associated with faster decline in visuospatial ability, (‐0.0079, p=0.036) and (‐0.0071, p=0.018) respectively, independent of neuropathologies, diabetes and demographics. Conclusion: This is the largest investigation on neuropathologic and cognitive correlates of EPVS conducted to date, generating robust evidence on EPVS associations with infarcts, CAA and diabetes, and independent contributions on cognitive decline.