In-vivo MRI-based prediction of cerebral amyloid angiopathy in community-based older adults

The National Institute on Aging awarded Rush University Medical Center $656,503 under CFDA 93.866 (Aging Research) on September 1, 2026, to develop and validate an in-vivo MRI-based classifier for detecting cerebral amyloid angiopathy in community-dwelling older adults.

The project combines multimodal MRI imaging with neuropathological data to create a machine-learning classifier capable of identifying cerebral amyloid angiopathy—a common age-related pathology affecting up to 80 percent of older persons at autopsy—without requiring postmortem diagnosis. The work addresses a significant clinical gap: while moderate-to-severe cerebral amyloid angiopathy affects 30 percent or more of older adults and associates with hemorrhagic stroke, microinfarcts, and accelerated cognitive decline, current clinical detection methods lack sensitivity in community-based populations. The classifier will be trained using ex-vivo MRI measurements from brain tissue specimens, translated to in-vivo imaging protocols, and tested longitudinally using clinical data, in-vivo MRI, and pathology findings from older adults enrolled without dementia. The award supports research performed in Chicago, Illinois, with a period of performance from September 1, 2026, through May 31, 2031. This assistance type is a Project Grant, the standard mechanism for investigator-initiated research under the National Institute on Aging's aging and dementia research portfolio.

Abstract

Cerebral amyloid angiopathy (CAA) is a common age-related pathology, detected at autopsy in up to 80% of older persons, up to 30% diagnosed as moderate or severe. CAA involves deposition of β-amyloid (AΒ) in the walls of cortical and leptomeningeal blood vessels and, when moderate to severe, is associated with a plethora of adverse outcomes including hemorrhagic stroke, microbleeds, convexity subarachnoid hemorrhage, cortical superficial siderosis, microinfarcts, enlarged perivascular spaces, and white matter hyperintensities. CAA is the most common vascular comorbidity of alzheimer's disease (ad) and is associated with an increased odds of dementia and faster cognitive decline. Recent evidence suggests that CAA increases the risk of amyloid-related imaging abnormalities (aria) in ad patients undergoing anti-amyloid treatments. In spite of its high frequency and deleterious effects, definitive diagnosis of CAA can only be conducted at autopsy. While the boston criteria have good sensitivity and specificity for CAA in select patients with advanced clinical symptoms, most community-based older adults with moderate or severe CAA remain undetected. Therefore, a method that provides accurate information in-vivo about the presence of CAA in community-based older adults is needed. The overall goal of the proposed project is to develop and test an in-vivo MRI-based classifier of CAA pathology for use in community-based older adults by combining multimodal MRI and pathology in the same older persons from large community cohorts. Specifically, we propose to a) train a CAA classifier using machine learning based on EX-vivo MRI measurements of brain characteristics of older persons, b) translate the classifier in-vivo, and c) test IT in-vivo using longitudinal clinical, in-vivo MRI, and pathology data on older adults enrolled without dementia. We have generated a unique EX-vivo and in-vivo multimodal MRI-pathology database within the infrastructure of the rush memory and aging project (map), minority aging research study (mars), religious orders study (ROS), african american clinical core (aacc) and latino core (latc) of the rush alzheimer's disease research center, five longitudinal, epidemiologic clinical-pathologic cohort studies of aging that recruit individuals without dementia and have high follow-up and autopsy rates. Using our database, we have produced compelling results in support of our aims. First, we have linked EX-vivo and in-vivo brain MRI data. Second, we have shown that CAA severity is related to specific brain MRI characteristics independent of other neuropathologies. Third, we have developed a preliminary CAA classifier exhibiting high sensitivity and specificity in community-based persons EX-vivo. Fourth, we have successfully developed novel, automated, in- vivo, MRI-based classifiers of other pathologies following the same strategy. We propose to further develop and test our promising CAA classifier for use in community-based older adults.