Assessment of oropharyngeal dysphagia using multimodal biosignals

ABSTRACT: Dysphagia is a swallowing impairment that affects the food, liquid, or saliva transit from the mouth to the stomach. Dysphagia leads to malnutrition, dehydration, and aspiration of the bolus into the respiratory system, which can lead to pneumonia with subsequent death. Dysphagia is produced by a set of neurogenic and neuromuscular conditions with variable incidence and prevalence. This condition is under-recognized and under-diagnosed. However, physical, economic, social, and psychological burdens have been clearly identified. The clinically accepted methods for dysphagia diagnosis and follow-up are invasive, uncomfortable, expensive, and experience-dependent. Furthermore, the reliability of some methods is still discussed. In this way, biosignals-based approaches that try to solve the aforementioned problems have been proposed, but no conclusive and hardly reproducible results have been achieved. Otherwise, such strategies generally ignore some physical aspects of the swallowing process. Therefore, this work explored non-invasive strategies to objectively assess dysphagia. To evaluate different physical aspects of the swallowing process, a multi-modal asynchronous analysis was performed with three biosignals: surface electromyography, accelerometry-based cervical auscultation, and speech. Such biosignals contributed to analyzing the swallowing-related phenomena in electrophysiological, mechanical, and acoustic dimensions. This thesis was focused on understanding oral and pharyngeal phases of the swallowing process by the use of the aforementioned signals. The following methodological steps were proposed to develop the dysphagia assessment scheme: 1) design of an acquisition protocol for the three biosignals in patients with dysphagia and healthy controls; 2) characterization of such biosignals in different mathematical domains, leading to the proposal of interpretable biomarkers; 3) construction of representation spaces and modeling of the swallowing patterns; and 4) evaluation of the multi-modal approach as a reliable method for swallowing assessment. All signals demonstrated their suitability for dysphagia screening by themselves, but bi- and tri-modal scenarios with Support Vector Machines, Extreme Gradient Boosting, k-Nearest Neighbors, and Gated Multimodal Units outperformed the uni-modal classification results. Specific configurations retrieved outstanding results, i.e. all performance measures obtained values ≥ 0,95. This thesis contributes to reducing the knowledge gap about swallowing-related phenomena and alterations from non-invasive and multi-modal points of view, with high potential to transfer and implement in clinical practice. It also contributes to objectively assessing dysphagia in the consulting room, helping with the diagnosis, follow-up, and rehabilitation of patients with dysphagia.

• Autores: Roldán Vasco, Sebastián
• Tipo de recurso: Doctoral thesis
• Fecha de publicación: 2023
• Institución: Universidad de Antioquia
• Repositorio: Repositorio UdeA
• Idioma: eng
• OAI Identifier: oai:bibliotecadigital.udea.edu.co:10495/36438
• Acceso en línea: https://hdl.handle.net/10495/36438
• Palabra clave: Trastornos de deglución; Deglutition Disorders; Deglución; Deglutition; Machine learning; Aprendizaje Automático; Procesamiento de señales; Signal processing; Dysphagia; Biomedical Signal Processing; Swallowing
• Rights: embargoedAccess
• License: https://creativecommons.org/licenses/by-nc-sa/4.0/