AI-Enhanced EEG Analysis Improves Early Dementia Identification

AI-powered analysis of routine EEG scans is now capable of distinguishing Alzheimer’s disease from frontotemporal dementia while also estimating disease severity. This offers a faster and more affordable pathway to diagnosis. Alzheimer’s and frontotemporal dementia are the two most common causes of dementia, often presenting with overlapping symptoms that make accurate diagnosis difficult in early stages. Misclassification can delay appropriate treatment, distort prognosis, and prevent timely care planning. Recent advances in AI have renewed interest in EEG as a front-line screening tool, opening opportunities for accessible and scalable dementia evaluation. Researchers developed a deep learning framework that combined a Convolutional Neural Network with an attention-based Long Short Term Memory network to extract both temporal and spectral EEG features.

The AI system achieved over 90% accuracy in distinguishing Alzheimer’s disease and frontotemporal dementia from cognitively normal individuals. It predicted disease severity with relative errors of less than 35% for Alzheimer’s disease and approximately 15.5% for frontotemporal dementia. Differentiating the two dementias proved more challenging, with initial specificity of 26%, but application of a feature selection procedure improved this to 65%. A two-stage classification approach then enabled simultaneous identification of Alzheimer’s disease, cognitively normal status and frontotemporal dementia, delivering an overall accuracy of 84% in separating the three groups. These findings indicate that AI enhanced EEG assessment could become a valuable front line triage tool for memory services, supporting faster referral decisions and more targeted investigations.

*Transparency Disclosure: This analysis was prepared by an AI language model to provide an executive summary of the provided news article. While the AI strives for accuracy and objectivity, its analysis should be considered as one perspective among many. Readers are encouraged to consult the original source and other expert opinions before making decisions based on this information.*