Caracterización de sonidos deglutorios adquiridos mediante auscultación cervical en sujetos sanos y con disfagia.

Swallowing is an act of high neuromuscular complexity due to the involvement of more than 30 muscle pairs and 5 cranial pairs that occurs in a short period of time. This activity is divided into 4 stages: pre-oral, oral, oropharyngeal and esophageal, an alteration in the normal development of this process can develop a symptom secondary to neuromuscular and neurogenic diseases that is known as dysphagia. Dysphagia can bring many difficulties for those who suffer from it including bronchial pneumonia, malnutrition, dehydration or even death by asphyxia. The identification of characteristics that help recognize this symptom, in addition to correctly describing the swallowing process is of great importance since the existing methods are invasive. Cervical auscultation is a technique by which information about the gothic closure can be obtained in the swallowing process using audio signals, and which can be analyzed offline. The aim of this study is to evaluate different methods of characterization of cervical auscultation signals and develop a machine learning model with manually segmented swallowing event sounds to classify between control subjects and patients with oropharyngeal dysphagia. The results showed, with a maximum accuracy of 75% that by means of signs is cervical auscultation it is possible to identify dysphageal subjects. In the same way it was possible to identify that the average potency of the swallowing segments was the feature with the best score (ROC curve) and that exists a different distribution between classes in this characteristic according to the U-Mann-whitney test to discriminate between healthy and pathologic subjects during different swallowing activities. [ABSTRACT FROM AUTHOR]