Fig. 5.

The decision tree of rehabilitation mode. A The decision tree. The first layer (also the bottom layer) was used to monitor the active motion state for the BCI. When the patient’s EEG score does not reach the threshold through analysis and decoding, it prompts the patient; at the same time, when the patient’s interruption time exceeds a certain time, it stops the exercise training. On the second layer, when the patient has no muscle strength, the patient used the BCI control mode. The rotation speed depends on the patient’s Mscore. The third layer, patient’s muscle strength is greater than grade 1. Through real-time analysis of EEG signals, when Mscore is greater than the threshold, the system control robot will provide patients with the lowest speed of rehabilitation training. A torque sensor was installed at the motor output shaft to detect the patient’s active force and control the speed change by the force control algorithm to provide assistance and resistance. B Three modes were available for the patients. For AIS A patients, the training generally switched from mode 1 to mode 2; for AIS B patients, the training generally switched from mode 1 to mode 3; and for AIS C patients, the training generally switched from mode 2 to mode 3