Modelling of Pharyngeal Laryngeal Effectiveness

The process of swallowing is complex and difficult to assess because of its dynamic nature, the lack of specialists and the technical means necessary for its exploration. Thus, the majority of patients with swallowing disorders do not benefit from any specific instrumental evaluation yet essential to the diagnosis of different types of disorders. The precise analysis of the deficit and its causes makes it possible to guide the choice of adaptation or rehabilitation strategies, with the aim of preventing the risk of complications, mainly nutritional (malnutrition) and respiratory (pneumonia). The development of new, more accessible diagnostic techniques would improve access to care for this population. The purpose of this study is to collect the signals of pharyngeal laryngeal activity through five non-invasive sensors (microphone, accelerometer, surface EMG, nasal cannula and oximeter) in order to identify indicators of functional efficiency of swallowing, protection of the lower airways and phonation. Modelling the mechanisms of pharyngeal-laryngeal efficacy through the use of artificial intelligence would provide an automated assessment of disorders and prognostic risk of complications.

Primary objective: From the signals of 5 non-invasive sensors, model a multidimensional scale of pharyngeal-laryngeal efficiency measuring the 3 main functions of pharyngeal-larynx: swallowing function (evaluated by pharyngeal transport and closure of the larynx) ; airway protection function (evaluated by cough reflex and cough power); phonation function (evaluated by voice and velar efficiency)

Secondary Objectives:

• Develop a diagnostic model of different physiopathological mechanisms of swallowing
• Determine an algorithm that can predict the severity of these disorders based on pharyngeal efficacy indicators
• Determine an algorithm capable of predicting the risk of complications from swallowing disorders, based on modelling pharyngeal effectiveness (primary objective) and clinical data

Study design: An interventional research study involving the prospective, multicentric and longitudinal human person to develop a scale of pharyngeal-laryngeal effectiveness measured from signals from 5 non-invasive sensors. The signals will be collected from each participant using the sensors (microphone, accelerometer, nasal cannula, surface EMG and oximeter). Algorithms will be developed from these signals to predict different indicators of pharyngeal-laryngeal efficiency measured by baseline tests. The study population will include patients with swallowing disorders (to identify pathological signal patterns) and healthy volunteers (to identify non-pathological physiological patterns of swallowing mechanisms).

440 patients (subjects with swallowing disorders), spread over 4 centers and 80 healthy subjects spread over 2 centers will be recruited for the study.

Research procedures

For patients with swallowing disorders:

During a hospital visit as part of their care pathway, patients with the criteria will be offered to participate in the research protocol. The subjects who have given their written consent will then carry out their assessment of the swallowing performed according to the available reference examination (videofluoroscopy or nasofibroscopy). In parallel, they will carry the following sensors: microphone, accelerometer, surface electromyography, nasal cannula and pulse oximeter. For research purposes, the subjects will also perform a cough and phonation test and questionnaires. Other data used are collected as part of routine care.

At the end of the assessment, the patient will receive management advice based on the reference exams and his medical record.

Each subject (or his caregiver) will be contacted by phone once a month for 6 months to monitor the occurrence of complications, The end of study visit at 6 months will be done either at the hospital with a new examination of the swallowing, by telephone, as medically indicated.

For healthy volunteers:

Only one visit is planned for the evaluation of the swallowing with the examination of nasofibroscopy and the installation of the sensors. The procedures (tests, examinations and questionnaires) will be identical to the data collection of the initial visit of patients with swallowing disorders)

Duration of the research: The first patient will be included from 2023. The inclusion period is 18 months. Patients will be monitored for 6 months (+/- 2 months) after inclusion Healthy volunteers will have one visit (estimated time: 2H) The total duration of the clinical study is expected to be 26 months

Data analysis: The characteristics of the subjects as well as the various clinical indicators of swallowing and their evolution during monitoring will be described. Development and validation algorithms to model the various pharyngo-laryngeal effectiveness indicators and predictive algorithms of swallowing disorders will be produced using statistical approaches and also using machine learning. The signals obtained from the 5 sensors will be annotated. The signals will be standardized and synchronized to allow the modelling of the sequences corresponding to the following events: swallowing of food, saliva, cough, throat clearing, speech, wet voice. Stochastic modelling based on hidden Markov models will be used initially and followed by the implementation of deep neural networks to model indicators. For the complication's prediction algorithm, deep neural networks will also be used to evaluate signal-based methods.