CwE-T: A Channel-wise Encoder with Transformer for EEG Abnormality Detection

Electroencephalogram (EEG) signals are critical for detecting abnormal brain activity, but their high dimensionality and complexity pose significant challenges for effective analysis. In this paper, we propose CwE-T, a novel framework that combines a channel-wise convolutional neural network (CNN)-based encoder with a single-head transformer classifier for efficient EEG abnormality detection. The channel-wise encoder compresses raw EEG signals while preserving channel independence, reducing computational costs and retaining biologically meaningful features. CwE-T was evaluated using two public datasets. For the TUH Abnormal EEG Corpus, the proposed model achieved 85.0% accuracy, 76.2% sensitivity, and 91.2% specificity at per-case level, outperforming baseline models such as EEGNet, Deep4Conv, and FusionCNN. For the CHB-MIT dataset, the proposed model achieved 85.4% sensitivity and 90.0% specificity for per-signal evaluation. Furthermore, CwE-T requires only 202M FLOPs and 2.9M parameters, making it significantly more efficient than transformer-based alternatives. The framework incorporates a channelwise design that provides potential for interpretability, offering promising directions for future research in neuroscience and clinical applications. The source code is available at https://github.com/YossiZhao/CwE-T.