Automation of the timed water swallow test (TWST) using high-resolution cervical auscultation signals - Archive ouverte HAL Accéder directement au contenu
Poster De Conférence Année : 2022

Automation of the timed water swallow test (TWST) using high-resolution cervical auscultation signals

Résumé

Purpose: The Timed Water Swallow Test (TWST) is a reliable and validated protocol to evaluate the swallowing functions (Hughes & Wiles 1996). Clinicians measure the time taken by their patient to consume 100mL or 150mL of water and count the number of swallows required. The average volume per swallow (V/S), average time per swallow (T/S) and swallowing capacity (V/T) are then calculated to perform a swallowing assessment. This preliminary study aims to automate time measurement and swallow count by analyzing high-resolution cervical auscultation signals. Method: 41 healthy subjects bellow 60 years old were asked to consume a 100 ml glass of water, without interruption, while wearing the Swallis DSA , a high-resolution cervical auscultation device including a microphone and a tri-axis accelerometer. We developed and trained a recurrent neural network to automatically identify and localize multiple successive swallows in the Swallis DSA signals. The TWST duration is estimated by the time between the start of the first detected swallow and the end of the last one. The average duration of the test is 5.82s. Results: When applied on the 41 recordings, 173 of the 205 (84%) swallows have been correctly detected. The average error in the duration estimation is 0.67s. Despite some missed detections, the mean computed TWST values (V/S, T/S and V/T) are in the expected range for the tested population. Conclusions: These results are encouraging in the perspective of the development of an automatic TWST. Multiple successive swallows differ from isolated swallows as it does not always imply a total repositioning of the larynx. This can lead to weaker signals through the sensors, and deeper work must be done in terms of signal processing to improve the reliability of the system. The non-invasiveness of this examination combined with a more efficient algorithm, could allow clinicians to quickly obtain the TWST indices in a routine swallowing disorders screening or to monitor the evolution of a patient.
Fichier principal
Vignette du fichier
PosterSession2_Gravellier.Lila.pdf (1.09 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04285442 , version 1 (14-11-2023)

Identifiants

  • HAL Id : hal-04285442 , version 1

Citer

Lila Gravellier, Maxime Le Coz, Jérôme Farinas, Julien Pinquier. Automation of the timed water swallow test (TWST) using high-resolution cervical auscultation signals. 31st Dysphagia Research Society Annual Meeting (DRS 2022), Mar 2022, San Francisco, CA, United States. . ⟨hal-04285442⟩
57 Consultations
15 Téléchargements

Partager

Gmail Facebook X LinkedIn More