Identifying incipient dementia individuals using machine learning and amyloid imaging

• DOI: 10.1016/j.neurobiolaging.2017.06.027
• Publication Date: 2017-07-11T09:10:27Z
• Authors: Sulantha Mathotaa...; Tharick A. Pascoal; Monica Shin; Andrea L. Benedet; Min Su Kang; Thomas Beaudry; Vladimir S. Fonov; Serge Gauthier; Pedro Rosa-Neto
• Keywords: Amyloid (mycology)

Identifying individuals destined to develop Alzheimer's dementia within time frames acceptable for clinical trials constitutes an important challenge design studies test emerging disease-modifying therapies. Although amyloid-β protein is the core pathologic feature of disease, biomarkers neuronal degeneration are only ones believed provide satisfactory predictions progression short frames. Here, we propose a machine learning–based probabilistic method designed assess 24 months, based on regional information from single amyloid positron emission tomography scan. Importantly, proposed was overcome inherent adverse imbalance proportions between stable and progressive mild cognitive impairment observation period. The novel algorithm obtained accuracy 84% under-receiver operating characteristic curve 0.91, outperforming existing algorithms using same biomarker measures previous multiple modalities. With its high accuracy, this has immediate applications population enrichment in therapies aiming mitigate disease dementia.