Noninvasive and automatic diagnosis of patients at high risk of swallowing aspiration

Abstract

Swallowing aspiration is known as the most clinically significant symptom of swallowing disorders (dysphagia). Noninvasive methods for detection of aspiration (the entry of food into airway due to dysphagia) are of great interest as they will lead to better management of dysphagia; thus, the risk of pneumonia, length of hospital stay and overall health care expenses can be reduced. The risk of aspiration is much higher in severely dysphagic patients. Normally, aspiration is detected by an imaging technique during swallowing, which is time consuming, costly and requires the patient's cooperation. In this study, we investigated the application of acoustical analysis of breathing and swallowing sounds for identifying patients at high risk of aspiration. We propose a novel method based on phase-space analysis of breathing sounds immediately after the swallow followed by support vector machine classification for use as a diagnostic aid for identifying patients with high risk of aspiration. We evaluated the method using breath and swallowing sounds recorded from 50 dysphagic individuals, 27 of which demonstrated silent aspiration (without cough or throat clearance) during either fiberoptic endoscopic evaluation of swallowing (FEES) or videofluoroscopic swallowing (VFS) assessment. The classification result of the proposed method was compared with those of the FEES/VFS assessment provided by speech-language pathologists; it showed 91% sensitivity and 85% specificity in detection of patients with severe aspiration (high risk dysphagia). The result is promising to suggest the proposed phase-space acoustical analysis method as a quick and noninvasive screening clinical tool to detect patients developing severe aspiration.