Machine Learning Uncovers Dementia Subtypes With Implication for Drug Trials

A study from University College London (UCL) researchers published in Nature Communications found machine learning could help uncover new treatments for dementia, reports Medical Xpress . The team describes an algorithm that can automatically extract different patterns of disease progression in persons with a range of dementias, including Alzheimer's, to allow individuals to be identified that may respond best to different therapies. The Subtype and Stage Inference (SuStaIn) algorithm was applied to routine magnetic resonance imaging (MRI) scans from subjects with dementia. SuStaIn employs medical imaging, which lets doctors see how disease is progressing, examine at the specific sites of protein buildup within the brain, and infer which parts are degenerating. SuStaIn successfully classified three separate subtypes of Alzheimer's, which broadly correlate with those seen in postmortem brain tissue analysis, as well as several different subtypes of frontotemporal dementia. However, the researchers think this subtyping could be done on living subjects via brain scanning, very early in the disease process. "This work shows that it is possible to tease out different disease patterns — some hitherto unknown — from single MRI scans taken from persons with a range of different dementias," says UCL Professor Jonathan Schott. "As well as providing new insights into dementia, this work demonstrates the huge potential of SuStaIn to delineate disease subtypes in a range of other medical contexts."