Silent aspiration detection by breath and swallowing sound analysis

Abstract

Detecting aspiration after swallows (the entry of bolus into trachea) is often a difficult task particularly when the patient does not cough; those are called silent aspiration. In this study, the application of acoustical analysis in detecting silent aspiration is investigated. We recorded the swallowing and the breath sounds of 10 individuals with swallowing disorders, who demonstrated silent aspiration during the fiberoptic endoscopic evaluation of swallowing (FEES) assessment. We analyzed the power spectral density (PSD) of the breath sound signals following each swallow; the PSD showed higher magnitude at low frequencies for the breath sounds following an aspiration. Therefore, we divided the frequency range below 300 Hz into 3 sub-bands, over which we calculated the average power as the characteristic features for the classification purpose. Then, the fuzzy k-means unsupervised classification method was deployed to find the two clusters in the data set: the aspirated and non-aspirated groups. The results were evaluated using the FEES assessments provided by the speech language pathologists. The results show 82.3% accuracy in detecting swallows with silent aspiration. Although the proposed method should be verified on a larger dataset, the results are promising for the use of acoustical analysis as a clinical tool to detect silent aspiration.

A review on intelligent aid diagnosis for dysphagia using swallowing sounds

“…Time-domain analysis, the most commonly analyzed signal feature, directly extracts temporal, amplitude, or morphologic features of the swallowing sound signal, including the mean, variance, kurtosis, root mean square of the swallowing sounds, and total swallowing duration 16,24,25,36,37,[41][42][43][44][45][46][47][48] . Some studies have analyzed various frequency-domain features of the swallowing sound signal, such as peak frequency, mean frequency, spectral centroid, and bandwidth, by extracting the spectral features after the Fourier transform of the swallowing sound signal or the power spectral density 13,20,24,25,29,36,43,[49][50][51][52][53][54][55][56][57] . Some studies have utilized the short-time Fourier transform or continuous wavelet transform to obtain the time-frequency joint features of swallowing sound signals.…”

Analysis of COVID-19 Heavy Cough Sounds Using Bark Wavelet Cepstral Coefficients

“…V. Swarnkar et al used automatic algorithm to classify wet and dry cough sounds [16]. Sarraf Shirazi and Z. Moussavi detected silence aspiration by breath [17]. T. Drugman et al used sensors to detect cough automatically [18].…”