In a recent breakthrough, researchers at Massachusetts General Hospital developed an AI-based method for accurately detecting Alzheimer's disease using routinely collected clinical brain images. The study, published in PLOS ONE, utilized deep learning to create a model trained on brain MRI data from patients with and without Alzheimer's. This model achieved a remarkable 90.2% accuracy in detecting Alzheimer's risk across multiple datasets, including 11,103 images from 2,348 at-risk patients and 26,892 images from 8,456 patients without the disease.
The deep learning model was designed to be "blind" to age-related brain features, allowing for accurate detection of early-onset cases. The model also utilized an uncertainty metric to identify patient data that differed significantly from its training set, ensuring successful predictions even in real-world clinical settings. This study represents a significant step towards the practical application of AI-driven diagnostic technology in detecting Alzheimer's disease.
Reference: Chase, B. (2023, March 3). Using AI to target Alzheimer's. Harvard Gazette. https://news.harvard.edu/gazette/story/2023/03/using-ai-to-target-alzheimers/