Researchers have published a description of the Knight Alzheimer Research Imaging (KARI) dataset in Nature Communications, a long-running collection of brain scans and clinical data that spans healthy aging, preclinical Alzheimer's disease and symptomatic disease. The resource was developed by the Knight Alzheimer Disease Research Center at Washington University in St. Louis, and the paper's authors say its main purpose is to give other scientists the context and quality-assurance information they need to use it for their own work.
The numbers are substantial. The dataset includes 1,645 participants aged 42 to 97, who together contributed 6,217 neuroimaging sessions, and more than 800 of them were imaged more than once, with some followed for over 15 years. Scans include structural, functional and diffusion MRI, plus PET imaging of amyloid and tau, alongside cognitive and clinical assessments, genetic information and fluid biomarkers. Most participants had no signs of cognitive impairment, which matters because the researchers are particularly interested in the years before symptoms appear, when amyloid and tau are thought to build up.
Within the cohort, the imaging measures tracked together with cognition in the expected direction. Greater cortical thickness in an Alzheimer's-sensitive brain signature was linked to better performance on a preclinical cognitive composite, while higher amyloid and higher tau burdens were linked to worse performance. The authors also reported that tau spread across the brain grew progressively from unimpaired participants to those with more severe clinical ratings, and that vascular features such as white matter hyperintensities and cerebral microbleeds were more common in impaired participants. The team also showed the dataset could be used to sort participants into an existing four-stage biological framework based on amyloid and tau patterns.
For researchers, the value of a single-institution dataset like this lies in deep, consistently collected data on the same people over long periods, which is hard to assemble otherwise. The authors note that some hardware variation over time is unavoidable. The paper's broader argument is that age and clinical status alone cannot reliably tell who has Alzheimer's pathology, since some people with amyloid or tau abnormalities remain cognitively normal, and resources like KARI are meant to help identify what separates healthy aging from preclinical disease.