Using Non-Invasive EEG Signals to Identify Urinary Urgency and Voiding Intentions

Urinary incontinence is closely associated with mobility limitations. In understaffed nursing homes, timely toileting support is often lacking, making reliable monitoring of voiding intention essential for reducing dependence and improving the quality of life. This study explores the feasibility of using non-invasive electroencephalography (EEG) signals to identify urinary urgency and voiding intentions. Ten participants were recruited, and EEG signals were recorded with a 64-channel EEG device during both non-urgent and urgent states. The SEResNet34 deep learning model, which incorporates ResNet34 and Squeeze-and-Excitation (SE) components with temporal aggregated pooling (TAP), was proposed, and the performance of the model was evaluated on the test set. The result showed that the improved SEResNet34 model can effectively classify EEG signals for urgency, achieving an accuracy of 0.9208, a precision of 0.9130, a recall of 0.9292, and an F1-score of 0.9211. The model outperformed ResNet18; however, the additional increase in layers beyond ResNet34 was not significantly beneficial. The ablation study showed that integrating the SE mechanism and the TAP function into the ResNet architecture can improve its accuracy. The proposed model surpasses the other models, indicating that it is better suited for the task of urgency detection. EEG shows significant potential in predicting urinary urgency and voiding intention, offering a new direction for the development of non-invasive monitoring and intelligent rehabilitation devices, with the potential to drive technological advancement in related fields.