A Novel Hybrid R-CNN-TPOT Model for Predicting Brain Age from EEG Data

With advancing age, significant transformations in electroencephalogram (EEG) signals are evident, which can be defined as brain age. One can identify discrepancies from conventional aging patterns by juxtaposing chronological age with brain age. Although Machine Learning (ML) techniques have been broadly applied in predicting brain age through brain imaging, EEG-based strategies have not been sufficiently explored, particularly about the Tree-based Pipeline Optimization Tool (TPOT). This study proposes a new hybrid ML technique for estimating brain age utilizing EEG data to fill this void. The approach integrates various features, including spectral, statistical, frequency-domain, and decomposition-domain attributes. Moreover, a novel model called Regression-based Convolutional Neural Network-TPOT (R-CNN-TPOT) has been developed for this analysis. This model merges the mathematical framework of CNN with the TPOT classifier via regression modelling. The performance of R-CNN-TPOT was measured using metrics such as Mean Absolute Error (MAE), Mean Squared Error (MSE), R-squared, and Root MSE (RMSE), achieving values of 0.033, 0.063, 0.095, and 0.251, respectively.