If geriatricians and primary care doctors could know which of their aging patients are at risk for Alzheimer’s disease, they could help these patients and their families prepare for what’s to come.
Meanwhile the same capability could help researchers pick the right patients for clinical trials testing new treatments and preventions.
Both clinicians and researchers will turn those coulds into cans if a new machine-learning model developed at MIT translates from lab to clinic.
MIT News reported Aug. 1 that the model can predict declining cognitive capacity up to two years before impairment symptoms become pronounced.
The model’s developers are set to present their work at the 2019 Machine Learning for Healthcare conference, which will be held at the University in Michigan Aug. 8 to 10.
To develop the technique, the team first trained a population model on neuropsychological test scores and biometric data from Alzheimer’s patients as well as from healthy individuals.
Then they developed a second model that was personalized for each patient and updated cognitive score predictions each time a patient visited the doctor.
The system they came up with is “like a model on top of a model that acts as a selector, trained using metaknowledge to decide which model is better to deploy,” explains MIT researcher Oggi Rudovic, PhD.
The MIT News article describes the work in some detail. Click to read: