The study published online in the Communications Medicine revealed that Alzheimer′s Disease Neuroimaging Initiative (ADNI, a machine learning system can accurately predict whether a person has Alzheimer′s disease (AD) based on a single MRI scan.

The researchers created a predictive model that uses T1-weighted MRI images to compute multi-regional statistical morpho-functional mesoscopic features, with or without cognitive scores. A biomarker known as "Alzheimer′s Predictive Vector" (ApV) was created for each patient utilizing a two-stage least absolute shrinkage and selection operator (LASSO). They divided the brain into 115 areas and assigned 660 distinct features to each region to assess each region, such as size, shape, and texture, then built the model to find out where changes in these aspects may accurately predict AD.

The brains of 400 patients with early and later stage Alzheimer′s disease, healthy controls, and patients with other neurologic diseases such as frontotemporal dementia and Parkinson′s disease were scanned using data from the Alzheimer′s Disease Neuroimaging Initiative (ADNI). In 98 % of cases, the MRI-based machine learning program could reliably identify whether a person had Alzheimer′s disease, and in 79 % of cases, it could distinguish between early- and late-stage AD.

The study found that this novel data analytic method has the potential to improve Alzheimer diagnosis accuracy, and additional research is needed before it can be used in clinical practice and adjusted for the clinical context.

(Medscape - Jun 21, 2022. Commun Med 2, 70 (2022). https://doi.org/10.1038/s43856-022-00133-4