Deep learning framework for EEG-based diagnosis of dementias (www.frontiersin.org)

A new deep learning framework has been unveiled for the diagnosis of Alzheimer's Disease and Frontotemporal Dementia (FTD) using electroencephalogram (EEG) data. Developed by a collaborative team from institutions in Pakistan, the UK, and Saudi Arabia, this explainable and lightweight framework employs temporal convolutional networks and long short-term memory networks to classify FTD, Alzheimer's, and healthy controls. Utilizing modified Relative Band Power (RBP) analysis for feature engineering, the model has demonstrated impressive classification accuracies—99.7% for binary tasks and 80.34% for multi-class tasks—marking a significant advancement over traditional machine learning models that often lack interpretability. This development is crucial for the AI/ML community as it highlights the potential of combining deep learning with EEG data for early and accurate dementia diagnosis, a pressing need given the projected rise in dementia cases globally. By incorporating SHAP (SHapley Additive exPlanations), the framework enhances transparency and interpretability, crucial aspects for gaining trust in healthcare AI applications. The approach not only promises improved diagnostic capabilities but could also lead to better treatment outcomes through earlier intervention, making it a noteworthy contribution to the intersection of AI and neurological health.