Purpose: In the field of medical care, urination detection for bedridden patients is particularly challenging, and existing research has focused on monitoring bladder pressure, which does not reflect the patient’s true intentions. Therefore, this study proposes a deep learning-based method for active urination detection using EEG to explore brain activities related to urination. Methods: This study introduced the FBCNet deep learning model, combined with Integrated Gradients (IG) for interpretability analysis, to explore the activity patterns of the brain in states with and without sense of urination. FBCNet was used to decode the intention to urinate from EEG, while IG was used to assess the contribution of different frequency bands and channels to the classification. Results: FBCNet achieved an accuracy of 65.83 ± 0.14% in the task of urination detection, with the highest accuracy for a single subject reaching 88.60 ± 0.06%. The analysis showed that in the state of urination, brain activity was mainly concentrated in the sensorimotor cortex under the Delta band. Conclusion: This study reveals that brain activities related to urination mainly occur in the sensorimotor cortex, a process that involves first perceiving the presence of the urination and then generating the control intention for urination motor. Additionally, it is proven that the degree of urination is determined by the feedback strength of motor intention for urination. Significance: As the first study to use EEG to detect urination, this study provides a new perspective for clinical monitoring of urination and opens up new possibilities for future clinical application.