Effect of artificial intelligence-based video-game system on dysphagia in patients with stroke: A randomized controlled trial

Background and aims: Post-stroke dysphagia is highly prevalent and causes complication. While video games have demonstrated potential to increase patient engagement in rehabilitation, their efficacy in stroke patients with dysphagia remains unclear. This aim of this study was to explore the effectiveness of the artificial intelligence-based video-game (AI-VG) intervention in improving swallowing function among stroke patients with dysphagia. Methods: A prospective, single blinded, randomized controlled trial (RCT) was conducted in a rehabilitation hospital from October 2023 to July 2024. Participants were randomly assigned to either the AI-VG system group or the usual care group. All participants received training for 30 min per session per day, five times per week for 4 weeks. The primary outcome was change in swallowing function from baseline (T0) to post-intervention (T1) and 1 month follow-up (T2). Secondary outcomes included changes in laryngeal function, oral intake function, nutritional status, and swallowing-related quality of life. The adherence, satisfaction, and acceptance of the two groups were evaluated. Changes in outcomes over time were compared using generalized estimating equation modeling. Results: A total of 84 participants (AI-VG group = 42, age = 65.00 +/- 9.70 years, 57.10 % male; usual care group = 42, age = 66.40 +/- 13.10 years, 69.00 % male) were included in this study. Compared with the usual care group, the AI-VG group showed significantly improved swallowing function, with a mean group difference of 4.02 (95 % CI = from -6.16 to -1.89, P < 0.001) at T1 and 4.14 (95 % CI = from -6.16 to 2.12, P < 0.001) at T2. Oral intake function, nutritional status, and swallowing-related quality of life improved significantly (P < 0.001 for overall group x time interaction). Adherence was significantly higher in the intervention group than in the control group (median [IQR], 18.00 [17.00-20.00] vs. 16.00 [15.00-17.00], P < 0.001). The intervention group had higher levels of acceptance and satisfaction of AI-VG (103.00 [100.50-104.00] and 73.00 [72.00-74.00], respectively). No significant difference was observed in laryngeal function (P > 0.05). Conclusions: The AI-VG intervention is an effective method for enhancing swallowing function in patients with post-stroke dysphagia. Future validation through multicenter, large-sample RCTs are warranted to confirm the findings and explore broader clinical applications.