Scientists have developed a new AI system that can help diagnose Alzheimer's disease by analyzing just 8 seconds of brain wave data. The method, which uses a deep learning model called LaBraM, achieves remarkable accuracy in distinguishing dementia patients from healthy individuals.
The Research
Researchers at the University of California San Diego, led by Maggie Lin and Tzyy-Ping Jung, trained a large AI model on over 2,500 hours of EEG data. EEG (electroencephalography) records the brain's electrical activity through sensors placed on the scalp. The model, called LaBraM, learns complex patterns from this massive dataset. Then, they used an algorithm called Random Forest to classify whether a person has Alzheimer's.
In their study, published on the arXiv preprint server, they tested the system on a separate group of dementia patients and healthy controls. Using only 8-second EEG clips, the AI achieved an ROC-AUC of 89.36%, a balanced accuracy of 82.44%, and a precision-recall AUC of 81.45%. This outperforms traditional methods that analyze brain wave frequency bands.
The AI's decisions were also interpretable. By occluding (masking) parts of the EEG data, the researchers found the model relied on well-known Alzheimer's biomarkers: degradation of alpha and theta rhythms in occipital and frontal regions. They also found that higher predicted dementia probability correlated with worse cognitive performance, greater clinical severity, and specific changes in brain wave patterns, such as increased theta and alpha relative power and a higher aperiodic exponent—a measure of neural noise.
Why It Matters
Alzheimer's is biologically heterogeneous, making it tough to diagnose early. Traditional linear methods miss non-linear brain dynamics. This AI approach captures those complexities, potentially enabling earlier and more precise diagnosis. Because EEG is non-invasive, relatively cheap, and portable, this could make screening for cognitive decline more accessible. For the public, understanding that advanced AI can pull clinically relevant signals from noisy brain data is exciting—it hints that our brain activity holds subtle clues to our cognitive health.
What You Can Do
While this AI is not yet available for clinical use, you can proactively monitor your cognitive health. Regular mental stimulation, physical exercise, and adequate sleep support brain health. If you notice memory or thinking changes, consult a healthcare professional. You can also test your own cognitive abilities with validated online tools, but always seek professional advice for medical concerns.
Source: arXiv q-bio.NC
Curious about your own brain? Take our free adaptive IQ test or try 306 brain training levels.