This study introduces a mixed reality (MR)-enhanced pneumatic glove rehabilitation system that applies the PDCA cycle evaluation method to ensure continuous design optimization, enhancing usability, effectiveness, and patient compliance, where a systematic evaluation on six stroke patients (three male, three female, ages 28–68) demonstrated a 33.1% reduction in task completion time (from 19.8 s to 13.2 s on average), an 18.5% increase in movement accuracy (from 70.7% to 89.2%), and a 96.7% training completion rate, while 100% of participants reported enhanced motivation, 93.3% expressed willingness to continue rehabilitation with the device, and the system achieved a response latency of less than 45ms, ensuring seamless interaction in gamified training scenarios, proving that MR-assisted rehabilitation, combined with PDCA-driven iterative refinement, offers a clinically viable, engaging, and data-driven solution for stroke rehabilitation, with future research recommendations including the expansion of sample size, the integration of AI-driven adaptive rehabilitation, and the exploration of home-based therapy applications to further enhance patient outcomes, rehabilitation accessibility, and long-term effectiveness.