Dementia affects individuals differently, causing a variety of symptoms that progress at varying rates, which makes uniform treatment challenging. Recent research published in PLOS One, led by PhD candidate Sayantan Kumar and Philip Payne, PhD, along with Inez Oh, PhD, Suzanne Schindler, MD, PhD, Nupur Ghoshal, MD, PhD, and Zachary Abrams, PhD, utilized a novel data-driven unsupervised machine learning method to identify subtypes within a clinical dementia cohort. This method groups patients based on changes in their cognitive abilities over time.
By using readily accessible cognitive assessment scores across various cognitive and functional domains, rather than costly imaging biomarkers like MRI and PET scans, the study demonstrates that patients with very mild or mild dementia exhibit the greatest variability in cognitive symptoms and future risk of progression. This research marks an exciting advancement in understanding cognitive heterogeneity in dementia.