Establishment of Voice Analysis Cohort for Development of Monitoring Technology for Dysphagia
Overview
- Phase
- Not Applicable
- Intervention
- Not specified
- Conditions
- Deglutition Disorders
- Sponsor
- Seoul National University Hospital
- Enrollment
- 300
- Locations
- 1
- Primary Endpoint
- Accuracy of machine learning prediction model using voice change before and after dietary intake
- Status
- Recruiting
- Last Updated
- 2 years ago
Overview
Brief Summary
Collection of basic data to develop a technique for monitoring the state of dysphagia using voice analysis.
Detailed Description
* Design: Prospective study * Inclusion criteria of the patient group * Patients scheduled for VFSS examination and normal person (without dysphagia) capable of recording voice (selected as a control group for comparison of voice indicators with patients with dysphagia) * Patients who can record voices such as "Ah for 5 seconds", "Ah. Ah. Ah.", "umm\~\~\~" * Inclusion criteria of the control group: Patients unable to speak, Patients who cannot follow along, If the VFSS test is a retest * Setting: Hospital rehabilitation department * Intervention: After obtaining the consent form for the patient scheduled for the VFSS test, "Ah for 5 seconds", after clearing the throat, "Ah for 5 seconds", briefly cut with a high-pitched sound, "Ah. Ah. Ah", close your lips lightly and make a "ummm\~\~\~\~" sound, and record 2 times each.
Investigators
Ju Seok Ryu
Professor
Seoul National University Hospital
Eligibility Criteria
Inclusion Criteria
- •Patients with dysphagia and scheduled for VFSS testing
- •Patients who can record voice such as "Ah for 5 seconds", "Ah. ah. ah", or "Um\~\~"
- •Normal people (without dysphagia symptoms) who can record voice (additionally recruited for comparison of voice indicators with patients with dysphagia)
Exclusion Criteria
- •Patients who cannot speak.
- •Patients who cannot speak according to the researcher's instructions.
- •Patients whose VFSS test was reexamined
Outcomes
Primary Outcomes
Accuracy of machine learning prediction model using voice change before and after dietary intake
Time Frame: day 1
Accuracy measures how well machine learning predicts three groups ('Normal', 'Residue', 'Aspiration') according to voice changes before and after dietary intake.
Secondary Outcomes
- AUC (Area Under the ROC curve) of machine learning prediction model using voice change before and after dietary intake.(day 1)
- mAP (mean Average Precision) of machine learning prediction model using voice change before and after dietary intake(day 1)
- Recall of machine learning prediction model using voice change before and after dietary intake.(day 1)
- Accuracy of machine learning prediction model using only voice after dietary intake.(day 1)
- mAP (mean Average Precision) of machine learning prediction model using only voice after dietary intake.(day 1)
- AUC (Area Under the ROC curve) of machine learning prediction model using only voice after dietary intake.(day 1)
- Recall of machine learning prediction model using only voice after dietary intake.(day 1)