Liscano Yamil, Bernal Lina Marcela, Díaz Vallejo Jhony Alejandro
Grupo de Investigación en Salud Integral (GISI), Departamento Facultad de Salud, Universidad Santiago de Cali, Cali 760035, Colombia.
Grupo de Investigación en Fonoaudiología y Psicología, Facultad de Salud, Universidad Santiago de Cali, Cali 760035, Colombia.
Brain Sci. 2025 Sep 18;15(9):1007. doi: 10.3390/brainsci15091007.
Traditional aphasia therapy is often limited by insufficient dosage, a barrier that AI-assisted digital therapies are poised to overcome. However, it remains unclear whether gains on specific tasks translate to functional, real-world communication. This systematic review evaluates the effectiveness of these novel interventions and investigates the potential for a "generalization gap" when compared to conventional treatments for post-stroke aphasia rehabilitation. Following PRISMA guidelines, we systematically reviewed randomized controlled trials (2010-2024) from six databases. We included studies examining AI-powered digital platforms for adults with chronic post-stroke apha-sia that reported standardized language outcomes. Our analysis of five trials ( = 366) shows that AI-assisted therapies successfully deliver high-dose interventions, leading to significant improvements in trained language skills, including word retrieval (up to 16.4% gain) and auditory comprehension. However, a critical "generalization gap" was consistently identified: these impairment-level gains rarely transferred to functional, real-world communication. AI-assisted digital therapies effectively solve the dosage problem in aphasia care and improve specific linguistic deficits. Their primary limitation is the failure to generalize skills to everyday use. Future platforms must therefore be strategically redesigned to incorporate therapeutic principles that explicitly target the transfer of skills, bridging the gap between clinical improvement and functional communication.
传统的失语症治疗往往受到剂量不足的限制，而人工智能辅助的数字疗法有望克服这一障碍。然而，尚不清楚在特定任务上取得的进展是否能转化为功能性的现实世界交流能力。本系统评价评估了这些新型干预措施的有效性，并研究了与中风后失语症康复的传统治疗方法相比，是否存在“泛化差距”的可能性。按照PRISMA指南，我们系统地审查了来自六个数据库的随机对照试验（2010 - 2024年）。我们纳入了研究为患有慢性中风后失语症的成年人提供人工智能驱动数字平台的研究，这些研究报告了标准化的语言结果。我们对五项试验（n = 366）的分析表明，人工智能辅助治疗成功地提供了高剂量干预，使训练后的语言技能有显著改善，包括单词检索（提高了16.4%）和听觉理解。然而，一个关键的“泛化差距”一直存在：这些在损伤水平上的改善很少能转化为功能性的现实世界交流能力。人工智能辅助的数字疗法有效地解决了失语症治疗中的剂量问题，并改善了特定的语言缺陷。其主要局限性在于未能将技能泛化到日常使用中。因此，未来的平台必须进行战略性重新设计，纳入明确针对技能转移的治疗原则，弥合临床改善与功能性交流之间的差距。
