Calculation of upper esophageal sphincter restitution time from high resolution manometry data using machine learning

Area of Science:

• Gastroenterology
• Physiology
• Medical Technology

Background:

• The upper esophageal sphincter (UES) requires time to return to its resting pressure after swallowing.
• Impaired UES function during this restitution phase can cause dysphagia and globus sensation.
• Current high-resolution manometry (HRM) methods struggle to precisely determine UES restitution time (RT).

Purpose of the Study:

• To develop a Machine Learning (ML) model for objective determination of UES RT.
• To analyze physiological data related to UES function post-swallowing.
• To improve the accuracy and reproducibility of RT measurements in HRM studies.

Main Methods:

• Utilized HRM data from 15 healthy participants (10 swallows each).
• Manually annotated RT intervals by two swallowing experts.
• Applied a sequence labeling ML model based on logistic regression for RT calculation.
• Compared ML-generated RT values with expert annotations.

Main Results:

• The ML model generated RT estimates for 150 swallows.
• Mean RT values ranged from 8.91s±3.71 to 10.87s±4.68, compared to expert-annotated means of 11.16s±5.7 and 10.04s±5.74.
• Correlation scores between model predictions and expert annotations ranged from 0.63 to 0.76.

Conclusions:

• RT is a crucial, previously overlooked, physiological parameter in UES HRM studies.
• UES requires approximately 9-11 seconds to rest post-swallowing, suggesting a 25-30 second interval between swallows.
• The developed ML model offers a reproducible method for RT determination, with potential for future integration with other swallowing parameters for automated HRM analysis.