Titulo Estágio
Swallow Detection and Analysis
Área Tecnológica
Informática Médica
Local do Estágio
DEI-FCTUC
Enquadramento
Swallowing is a complex act that involves non-trivial coordination of various muscles with the simultaneous closing of the epiglottis and soft palate. As with other symptoms, Multiple-Sclerosis (MS) can damage the nerves as well as the brain areas involved in the swallowing coordination, leading to dysphagia. Dysphagia has a relatively high incidence and prevalence among MS patients. It is estimated that between 30 and 40 percent of people with MS experience swallowing problems at some time. Like many other symptoms of MS, dysphagia might be temporary and more accentuated during relapses and improve, or disappear completely, over time. Patients with dysphagia might inhale food or liquids into the trachea. Once in the lungs, the inhaled food or liquids can cause pneumonia or abscesses. Because the food or drink is not reaching the stomach, a person may also be at risk for malnutrition or dehydration. Another potential danger of dysphagia associated to food aspiration is choking. The currently considered gold standard method for detecting swallowing events is videofluoroscopy and EMG. Both approaches are inadequate for ambulatory and continuous applications. Videofluoroscopy requires bulky and potentially unsafe equipment for continuous use while EMG relies on frequently subcutaneous placement of electrodes in the masseter, suprahyoid and infrahyoid muscles to avoid interference of the muscles of the neck. Manual annotation by one or more trained observers (mainly based on sound signals) is another common approach followed in medical practice. Automatic swallowing event detection has been attempted based on several types of sensors such as strain sensors, accelerometers and audio. Most of existing methods for automatic swallowing analysis rely on audio analysis. This approach has widely explored in contexts such as food intake monitoring for obesity management and dysphagia diagnosis. From the signal processing point of view two main problems have to be addressed: (i) discrimination of swallowing sounds from respiration sounds, ambient noise and speech, and (ii) identification of the swallowing origin or dysfunction (e.g. dysphagia or normal). Discrimination of swallowing sounds from other sound sources has been attempted using different linear and nonlinear signal processing methods by exploring signal energy, spectral content, time-frequency analysis (e.g. wavelet analysis), the intrinsic non-stationarity of the signal (e.g. by the fractal variance dimension) and non-linear dynamics (e.g. by the embedding dimension and time delay analysis). Regarding dysphagic swallowing discrimination from normal swallowing, usually algorithms rely on the implicit or explicit detection of the three swallowing phases, commonly referred to as oral, pharyngeal, and esophageal. It is observed that current methods still exhibit low accuracy in dysphagic swallowing discrimination.
Objetivo
The goal of the proposed thesis is to study, implement, evaluate and improve current algorithms for swallowing detection in multiple real-life contexts. The following tasks will be addressed: i) discrimination of swallowing sounds from respiration sounds, ambient noise and speech, (ii) identification of the swallowing origin or dysfunction and (iii) discrimination of the type of food (i.e. liquid, mild solid, etc.) swallowed.
Plano de Trabalhos - Semestre 1
-Analysis of the State of the Art in Swallowing detection and discrimination
-Formulation of Hypothesis
-Collection and annotation of a clinical database on swallowing
-Writing of the first semester report
Plano de Trabalhos - Semestre 2
-Implementation, evaluation and improvement of swallowing detection algorithms
-Implementation, evaluation and improvement of swallowing origin and dysfunction detection methods
-Implementation, evaluation and improvement of algorithms for the detection of food intake
-Writing of the dissertation
Condições
Good knowledge in data analysis
Good programming skills
Orientador
Paulo de Carvalho
carvalho@dei.uc.pt 📩