Related Experiment Video
Updated: Jun 12, 2026
08:32
Adapting Human Videofluoroscopic Swallow Study Methods to Detect and Characterize Dysphagia in Murine Disease Models
Published on: March 1, 2015
Human-AI collaboration for dysphagia rehabilitation from effectiveness to implementation complexity: a systematic
Wenwen Yang1, Sufang Li1, Yifei Du1
1The Affiliated Cancer Hospital of Zhengzhou University & Henan Cancer Hospital, Zhengzhou, China.
NPJ Digital Medicine
|June 10, 2026
Summary
AI-assisted swallowing therapy offers short-term benefits for oropharyngeal dysphagia patients but faces significant adoption barriers. Long-term effectiveness and adherence without clinician support remain unproven, limiting widespread application.
Area of Science:
- Neurology
- Oncology
- Rehabilitation Medicine
- Artificial Intelligence
Background:
- Oropharyngeal dysphagia impacts over 50% of neurological and oncological patients.
- Rehabilitation efforts are hampered by a global shortage of speech-language therapists.
- Existing human-AI collaboration models have not demonstrably solved this therapist deficit.
Purpose of the Study:
- To systematically review the evidence on AI-augmented swallowing rehabilitation for adults with oropharyngeal dysphagia.
- To synthesize findings based on etiological factors and collaboration modes.
- To assess the risk of bias and certainty of evidence (GRADE) for AI interventions.
Main Methods:
- Systematic review of 31 studies involving 1012 participants (PROSPERO: CRD420251115997).
- Synthesis of findings stratified by etiology and collaboration type.
- Evaluation of risk of bias and certainty of evidence using GRADE criteria.
- Analysis using the NASSS framework to identify implementation barriers.
Main Results:
- AI interventions show short-term improvements in oral intake and physiological measures (GRADE: moderate/low certainty).
- Benefits diminish weeks after cessation; adherence drops significantly without clinician supervision.
- Adopter domain issues (digital literacy, usability) are major barriers (61.3% rated high).
- AI algorithm performance has very low certainty, with limited validation in healthy volunteers.
Conclusions:
- AI-augmented therapy shows promise for short-term gains, particularly in supervised post-stroke rehabilitation settings.
- Evidence for sustained outcomes, unsupervised use, and diverse etiologies is currently insufficient.
- Addressing adopter domain barriers is crucial for successful implementation in patient populations with the greatest need.
