Motor Imagery is a cognitive simulation of motor behavior and an important research area in brain computer interface technology. The principle of this technology is to convert motor consciousness into computer instructions, thereby achieving direct control of external devices by the brain. However, due to the sensitivity of non-invasive electroencephalography (EEG) to noise, different categories of feature information often have significant interference. In addition, people's understanding of the brain is still in the preliminary research stage, and it is difficult to accurately extract the discriminative information contained in EEG. In response to this issue, this paper proposes a robust phase slope index brain network detection algorithm. This algorithm eliminates the possible discrete factors in continuous cross frequency phase, while preserving the causal information between sequences and minimizing the influence of abnormal interference. It maximizes the extraction of discriminative information in the brain network and provides the possibility to improve the classification performance of motor imagery tasks. To test the effectiveness of the algorithm, this paper validated it using a publicly available motor imagery dataset. The experimental results show that the method proposed in this paper can effectively improve the separability of classification features, thereby enhancing the classification accuracy and generalization ability of the classification model. This method provides a new approach for brain computer interface technology based on motor imagery.