Non-invasive Assessment of Swallowing and Phonation using High-density Electromyography

Swallowing and phonation are essential neuromuscular functions critical for hydration, digestion, and communication. Phonation requires precise vocal fold coordination to produce sound, while swallowing involves coordinated movements to transport food and liquids safely. Dysfunctions in these processes, dysphagia and dysphonia, affect many individuals and significantly impact quality of life and health outcomes. This creates a need for effective diagnostic and rehabilitative tools. Unlike conventional assessments that focus on biomechanics and structure, high-density surface electromyography (sEMG) provides detailed insights into muscle physiology and motor activity. However, the rapid, dynamic contractions and complex musculature involved present unique challenges for the application, processing, and analysis of high-density sEMG. This thesis developed two flexible high-density electrode arrays with 128 and 256 channels that non-invasively cover the muscles involved in swallowing and phonation. The arrays were designed for consistent placement across individuals despite anatomical variability. A conductive paste applicator was integrated to allow efficient application within 10 minutes. A triggering system with temporal accuracy around 0.5 ms was created to enable precise comparisons between trials, subjects, and biomechanics. An automated signal processing framework was developed to efficiently process dynamic contractions across numerous muscle activations, minimising manual steps and reducing subjectivity and human error in handling large high-density datasets. Development of the framework involved optimising filtering and windowing, alongside a comprehensive comparison of four algorithms for detecting low-quality channels. Among these, a density-based local anomaly detection method performed exceptionally well, achieving an F1 score of 0.97. For data segmentation, the framework integrated audio recording to identify audio-based functional fiducial markers. For phonation events, this involved segmenting vocalisations, while for swallowing events, it involved segmenting the distinctive sound of the Eustachian tube opening. This approach improved temporal accuracy tenfold to approximately 0.025 s. The framework incorporated visualisation of temporal myoelectrical signatures via spatial averaging of regionally defined signals, with morphological metrics quantifying temporal activation differences. Additionally, spatial activation patterns were displayed on 2D maps and 3D patient-specific anatomy to illustrate muscle activity distribution. Using a test–retest design across ten sessions, high-density sEMG demonstrated highly repeatable measurements of neuromuscular activation, with extracted features showing excellent intersession reliability (ICCs > 0.90; CoV < 10%) and stable temporal waveform morphology (mean cross-correlation coefficients > 0.90). This repeatability is likely attributable to the enhanced spatial resolution of the high-density arrays combined with regional spatial averaging, which reduces variability from electrode placement and signal crosstalk. Distinct spatial and temporal biomarkers of healthy swallowing and phonation were identified. Swallowing tasks exhibited high-amplitude, low-frequency activity localised to the suprahyoid region with consistent regional coordination, whereas phonation was characterised by diffuse, lower-amplitude, higher-frequency activation. Temporal morphological metric analysis revealed task-specific neuromuscular strategies in swallow tasks, with effortful 5 ml swallows showing steeper activation slopes and greater amplitude, saliva swallows exhibiting broader, prolonged activation profiles, and effortful 10 ml swallows demonstrating lower amplitude and shorter duration, highlighting differences in volitional control and motor coordination demands across swallowing behaviours. Non-invasive high-density sEMG measurements were validated by performing concurrent recordings with established assessment modalities including videofluoroscopy, manometry, and endoscopy. This established that the audio-based functional fiducial accurately correlated with manometric data of oropharyngeal swallow onset, allowing subsequent sEMG investigations to non-invasively align with functional onset. The integration of endoscopy, audio, and high-density sEMG data, enhanced physiological and functional understanding of phonation. Using Granger causality, it was found that extrinsic laryngeal muscle activity has a stronger influence on audio frequency than on audio magnitude, with a significantly higher cG-max value (p-value < 0.001) and increased causality at 0.5 s post-vocalisation onset as pitch rises. Disorder-specific biomarkers for muscle tension dysphonia (MTD) were discovered, with MTD patients showing altered sEMG activation patterns. Normalising spatial maps by subtracting baseline activation revealed that MTD participants had low amplitude levels, indicating minor activation above an elevated baseline. These findings elucidate the previously unverified impact of extrinsic sEMG activation on phonatory biomechanics, clarifying conflicting literature and demonstrating altered relationships between extrinsic laryngeal muscle activation and intrinsic movements. This research extends high-density sEMG technique to dynamic neuromuscular coordination, demonstrating its effectiveness in providing clinically relevant details and offering insights that could advance techniques and improve strategies for diagnosing and managing swallowing and phonation disorders.

Name thesis:degree_name

PhD

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Bioengineering

Grantor dc:publisher

ResearchSpace@Auckland

Year dc:date.issued

2024

Author dc:creator

• Miller, Kiara JW

Advisors dc:contributor.advisor

• Cheng, Leo
• Avci, Recep
• Sands, Gregory
• Huckabee, Maggie-lee

• Sensors
• Swallow
• Phonation
• High-density electromyography
• Signal processing
• Neuromuscular coordination
• Muscle tension dysphonia
• Functional biomarkers
• Multimodal integration
• Non-invasive assessment

Statement dc:rights

• Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Licence dc:rights.uri

• https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htm

Handle dc:identifier.uri

https://hdl.handle.net/2292/72919

OAI identifier oai:identifier

oai:researchspace.auckland.ac.nz:2292/72919