AI-Based Dysphagia Rehabilitation Program: Development and Validation

Summary

The goal of this clinical study is to learn if an artificial intelligence (AI)-assisted swallowing rehabilitation program can improve swallowing function in adults with dysphagia. It will also learn whether this approach can better detect aspiration (food or liquid entering the airway) and improve eating ability. The main questions it aims to answer are:

Can AI-assisted analysis help identify swallowing treatment parameters that improve swallowing right away?

Does an optimized AI-assisted swallowing treatment work better than standard treatment parameters over 6 weeks?

Can AI using data from sensors detect aspiration accurately?

Researchers will compare swallowing function before and after treatment. They will also compare optimized treatment parameters selected with AI and FEES (a camera-based swallowing test) with standard treatment parameters. In addition, researchers will compare AI model results with FEES findings to see how well the sensors can identify aspiration.

Participants will:

Complete swallowing assessments, including FEES, drinking tests, and swallowing questionnaires.

Have small sensors placed to record muscle activity, sound, oxygen levels, and breathing while swallowing.

Receive swallowing rehabilitation over 6 weeks, with treatment given in 1-week periods. During each period, treatment will be randomly assigned as either optimized parameters or standard parameters.

Complete swallowing assessments after treatment periods to measure changes in swallowing function and eating ability.

Conditions

• Dysphagia

Interventions

DEVICE

Neuromuscular electrical stimulation for swallowing

Participants receive neuromuscular electrical stimulation (NMES) applied to swallowing-related muscles as part of swallowing rehabilitation. The stimulation parameters are individualized using an AI-assisted optimization process. The AI model analyzes swallowing assessment data, including FEES findings and multi-sensor signals, to select parameters expected to improve swallowing. The selected parameters are then applied. This intervention differs from conventional methods because the stimulation parameters are not selected solely by clinician experience, but are optimized by an AI-assisted system based on the participant's swallowing assessment results.

DIAGNOSTIC_TEST

Multi-sensor swallowing screening for dysphagia

Participants undergo a multi-sensor screening test for dysphagia. During swallowing, signals are collected from multiple sensors, such as surface electromyography, acoustic sensors, oxygen saturation, and respiratory sensors. The signals are analyzed using AI-assisted algorithms to identify patterns associated with swallowing impairment and aspiration risk. This approach provides objective, quantitative swallowing assessment and may be compared with FEES. It differs from conventional bedside screening and FEES by combining multiple physiological signals.

Sponsors & Collaborators

• National Research Center for Rehabilitation Technical Aids

  lead OTHER

Principal Investigators

• Pengxu Wei · National Research Center for Rehabilitation Techincal Aids

Study Design

Allocation

RANDOMIZED

Purpose

TREATMENT

Masking

SINGLE

Model

CROSSOVER

Eligibility

Min Age

18 Years

Sex

ALL

Healthy Volunteers

Yes

Timeline & Regulatory

Start

2026-08-31

Primary Completion

2027-12-31

Completion

2027-12-31