Deep Learning-Assisted Evaluation of Laryngeal Mobility in a Rat Model

Area of Science:

• Laryngology and Neuroscience
• Biomedical Engineering
• Animal Models in Research

Background:

• Vocal fold mobility is crucial for laryngeal function.
• Recurrent laryngeal nerve (RLN) and superior laryngeal nerve injuries can impair vocal fold mobility.
• Quantitative assessment methods are needed to evaluate these impairments.

Purpose of the Study:

• To quantitatively evaluate laryngeal mobility in a rat model following unilateral recurrent laryngeal nerve (RLN) injury.
• To establish a precise method for assessing laryngeal symmetry using computer vision.
• To provide a reliable tool for studying the effects of RLN injury on vocal fold function.

Main Methods:

• Utilized a rat model of unilateral RLN injury.
• Performed direct laryngoscopy before and after injury.
• Employed the Social LEAP Estimates Animal Poses (SLEAP) deep learning framework for high-resolution video analysis.
• Tracked laryngeal landmarks and measured arytenoid process displacement frame-by-frame.

Main Results:

• Developed a quantitative method to assess laryngeal mobility and symmetry.
• Established a mean difference threshold of 0.42 to differentiate laryngeal asymmetry from symmetry.
• Demonstrated the capability of computer vision techniques to precisely measure laryngeal landmark displacement.

Conclusions:

• Advanced computer vision, specifically SLEAP, offers a quantitative and accurate method for assessing laryngeal symmetry in a rat model.
• This technique is valuable for evaluating vocal fold mobility deficits following RLN injury.
• The findings support the use of this method in future research on laryngeal nerve injuries and their functional consequences.