A study published in the journal Sleep shows that deep neural network models can accurately predict the brain age of healthy test subjects based on electroencephalogram (EEG) data recorded during nighttime sleep studies. The brain age index predicted by EEG shows unique characteristics in people with different diseases.
The study found that the model's average absolute error in predicting age was only 4.6 years. Furthermore, researchers also discovered that patients with diabetes, depression, excessive daytime sleepiness, hypertension, and memory and attention problems had, on average, higher brain age indices compared to a sample of healthy individuals.
"While clinicians can only roughly estimate or quantify a patient's age based on an EEG, this study shows that AI models can predict a patient's age with high accuracy," said Yoav Nygate, a senior AI engineer at EnsoData and lead author of the paper.
Researchers created a deep neural network model to predict patients' ages using raw electroencephalogram (EEG) signals recorded in clinical sleep studies via nocturnal polysomnography. The brain age assessment method involves subtracting the actual age from the predicted age based on the EEG readings and then taking the absolute value of this variable. The analysis included factors such as gender and body mass index.
“The results of this study provide preliminary guidance for using artificial intelligence to assess patients’ brain age,” Nygate said. “We hope that, with continued investigation, research and clinical trials, the brain age index will one day become a diagnostic marker of brain health, just as high blood pressure is a risk indicator for stroke and other cardiovascular diseases.”
The study, titled "543 EEG-Based Deep Neural Network Model for Brain Age Prediction and Its Association with Patient Health Conditions," was recently published in the journal *Sleep*.
