NIH expands USC-led Alzheimer’s imaging AI initiative

The National Institutes of Health has awarded $12.6M to advance AI4AD2, the second phase of the Artificial Intelligence for Alzheimer’s Disease initiative. The new award brings total NIH investment in AI4AD to $30.7M, according to Keck School of Medicine of USC.

Paul M. Thompson, PhD, associate director of the USC Mark and Mary Stevens Neuroimaging and Informatics Institute, is leading the multi-institutional project. AI4AD2 includes 10 investigators and 23 co-investigators from 10 institutions.

Researchers will analyze large-scale datasets that include whole-genome sequencing, brain imaging, cognitive testing, and other biological data. The goal is to develop AI tools that can help uncover biological causes of Alzheimer’s disease and related dementias, improve prediction of disease progression, and support more precise treatment research.

AI4AD began in 2020. The first phase trained tools on 80,000 brain scans and developed AI methods that identified Alzheimer’s-related features on brain images with more than 90% accuracy, according to USC.

The second phase has 4 research goals. One will use AI to identify meaningful subtypes of Alzheimer’s disease and related dementias by combining brain scans, cognitive data, neuropathology, and genetics.

Another arm will develop genomic language models to analyze DNA sequences across data from more than 58,000 participants in 57 cohorts. The project will also adapt disease-classification tools for global and multi-ancestry populations, including African, Indian, Korean, and U.S. cohorts.

AI4AD2 will also use PreSiBO, an AI-based drug discovery tool developed during the first phase, to identify subtype-specific drug targets. The project’s work will include amyloid, tau, vascular injury, and inflammation-related disease mechanisms, according to USC.

“With AI4AD2, we are launching a program of genome-guided drug discovery,” Thompson said. He said the project is intended to help identify drugs that target specific types of dementia, including rarer subtypes.

The USC Stevens Neuroimaging and Informatics Institute will serve as the coordinating hub for the consortium. Software tools, analysis pipelines, training workshops, and project resources are expected to be shared publicly.