Unsupervised machine-learning approach shows potential for diagnosing Alzheimer’s disease

  • Researchers have developed a new method combining statistical and unsupervised machine-learning approaches that could be used to differentiate between magnetic resonance imaging (MRI) scans from people with Alzheimer’s disease (AD) and cognitively unimpaired individuals.
  • The method was tested on brain structural MRI scans from 198 people with AD and 231 controls, and had an accuracy of 84% for discriminating between the two groups.
  • The study authors highlight that existing algorithms for AD diagnosis use a supervised learning approach that requires a large volume of labelled MRI scans, which is difficult to do, but their new method only requires a limited number of labelled scans.