Brain Activity Classification in Coma Patients Using Wavelet and CST Features with CNN-based Analysis

The classification of a patient's brain status during a coma, such as identifying whether the brain is alive, inactive, or exhibiting any activity, is a critical task in medical diagnostics. Traditional methods like Support Vector Machines (SVM) have been widely used for this purpose, but they rely heavily on manually extracted features, which may limit their effectiveness in capturing complex brain activity patterns. This study proposes a novel approach that utilizes advanced time-frequency analysis techniques, specifically wavelet transform and Continuous Stockwell Transform (CST), in combination with Convolutional Neural Networks (CNNs) for automatic classification. The wavelet transform is employed to extract multi-resolution features that capture the time-varying nature of EEG signals, while CST enhances the frequency localization. These extracted features are then used as inputs to a CNN, which learns to classify the brain status without manual feature selection. The proposed method is expected to outperform traditional SVM classifiers by providing a more comprehensive understanding of the brain's activity during a coma. Preliminary results suggest that the CNN, with its ability to automatically learn and adapt to complex data, offers superior classification accuracy. This approach holds potential for improving diagnostic accuracy and aiding clinicians in making better-informed decisions.

Keywords: coma classification, EEG signals, wavelet transform, Continuous Stock well Transform (CST), Convolutional Neural Networks (CNN), brain activity, SVM.