EEG-Based AI Technology Accurately Diagnoses Alzheimer’s and Dementia

Symptoms often overlap with normal aging, and current diagnostic pathways can be slow, expensive, and resource-intensive. Researchers have now shown that analyzing the brain’s electrical activity can reliably separate healthy individuals from those with different forms of dementia. The findings demonstrate that artificial intelligence (AI) applied to EEG data can deliver accurate, rapid, and interpretable dementia detection.

The work was led by Örebro University (Örebro, Sweden), where researchers developed two complementary AI models for EEG-based dementia classification. The first approach combines temporal convolutional networks with long short-term memory networks to capture both short- and long-term patterns in brain signals. The system analyzes standard EEG recordings and evaluates changes across multiple frequency bands, including alpha, beta, and gamma waves.

A second model focuses on efficiency and privacy. Using a compact neural network design and federated learning, the system allows multiple healthcare providers to collaboratively train AI models without sharing sensitive patient data. Despite being under one megabyte in size, the model maintains high diagnostic performance while ensuring patient privacy and enabling deployment on portable or low-resource devices.

In the first study, the AI distinguished among healthy individuals, Alzheimer’s disease, and frontotemporal dementia with over 80 percent accuracy. An explainable AI layer highlighted which EEG signal segments most influenced the diagnosis, helping clinicians understand how decisions were made. In the second study, the privacy-focused model achieved more than 97 percent accuracy while keeping patient data decentralized. Together, the studies, published in Frontiers, demonstrate that EEG-based AI can be both highly accurate and clinically interpretable.

The findings suggest EEG combined with AI could become a rapid, low-cost, and privacy-safe screening tool for early dementia detection. Because EEG is already widely available and inexpensive, these models could extend dementia assessment beyond specialist clinics to primary care and even future home-based testing. The researchers plan to expand their work to larger and more diverse datasets and include additional dementia types such as vascular dementia and Lewy body dementia.

“Early diagnosis is crucial in order to be able to take proactive measures that slow down the progression of the disease and improve the patient's quality of life,” said Muhammad Hanif, researcher in informatics at Örebro University. “If solutions like this are fully implemented, it could ease the burden for everyone involved – patients, care staff, relatives and healthcare professionals.”

Related Links:
Örebro University