Objective: While the respiratory system provides the driving force for phonation, there is currently no study that modeled respiration for studying phonation biomechanics. Moreover, none of the existing computational models have incorporated the true geometry of the oral cavity. The present study proposes a lumped element model to simulate the phonation process through a holistic approach. This work models and couples the respiratory (lungs and trachea), phonatory (larynx and vocal folds), and articulatory (oral cavity) subsystems for normophonic and laryngeal edema subjects.

Methods: High-speed videoendoscopy (HSV) data were obtained from 18 adult subjects (7 normophonic males, 7 normophonic females, and 2 males and 2 females with edema). The data were collected during production of sustained vowel /i/ at habitual pitch and loudness using a Phantom v7.1 camera at 16,000 fps, 320 _ 320 resolution, and 12-bit depth. The respiratory subsystem is modeled by a mass-spring-damper governed piston-cylinder system, mimicking the expansion and contraction of the lungs. The phonatory subsystem is modeled as a single mass-spring-damper system and the contribution of articulatory subsystem to the flow losses is integrated as a lumped element considering no volume variability. Later, the glottal area variation, extracted from the HSV data using deep learning, was integrated into the model to drive the temporal evolution of the flow dynamics using the Particle Swarm Optimization algorithm.

Results: The proposed model successfully calculates the spatially averaged temporal variations in lungs’ pleural pressure and lung volume during sustained phonation across different subjects. Additionally, the subglottal and oral cavity flow characteristics for each subject were evaluated and compared to distinguish between the normophonic subjects and those with laryngeal edema. Unlike minimal flow resistances during respiration, higher flow velocity during phonation accentuates the losses that are quantified as additional lung, collapsible airway, and small airway resistances, reported here for the first time.

Conclusions: The present mechanical and electrical computational modeling frameworks offer a patient-specific tool adaptable to individual vocal fold conditions to correctly predict the flow and lung dynamics, which cannot be measured directly. Therefore, this work can provide in-depth information for laryngeal edema characterization, capturing individual differences and potentially assisting with patient-specific diagnosis and treatment.