Aiming at the problems of weak human-machine interaction on the doctor-side and insufficient visualization of patient movement information in the tele-rehabilitation system, the paper proposes a doctor-centered mixed reality tele-guidance training system, based on a self-developed doctor-patient interactive cable-driven lower limb rehabilitation robot platform. The solution constructs a high-precision virtual simulation mapping system and a mixed reality collaboration module to achieve real-time synchronization of patient movement states and multi-dimensional visual assessment. Static virtualization employs parameterized engineering models to ensure mapping accuracy, while dynamic virtualization utilizes inverse kinematics control strategies for joint motion trajectory mapping. The mixed reality collaboration module integrates Unity3D with HoloLens2 platforms to form a bidirectional closed-loop system for bioinformation acquisition, command transmission, and safety monitoring. Experimental results demonstrate that the root mean square error of trajectory tracking at the patient-side end is 20.28 mm, with rapid doctor-side interaction response. This confirms that the proposed method significantly enhances doctors’ real-time perception and dynamic adjustment capabilities during tele-rehabilitation processes, offering novel design insights for doctor-side interaction in remote rehabilitation systems.