Upper Esophageal Sphincter Opening Segmentation With Convolutional Recurrent Neural Networks in High Resolution Cervical Auscultation

Abstract

Upper esophageal sphincter is an important anatomical landmark of the swallowing process commonly observed through the kinematic analysis of radiographic examinations that are vulnerable to subjectivity and clinical feasibility issues. Acting as the doorway of esophagus, upper esophageal sphincter allows the transition of ingested materials from pharyngeal into esophageal stages of swallowing and a reduced duration of opening can lead to penetration/aspiration and/or pharyngeal residue. Therefore, in this study we consider a non-invasive high resolution cervical auscultation-based screening tool to approximate the human ratings of upper esophageal sphincter opening and closure. Swallows were collected from 116 patients and a deep neural network was trained to produce a mask that demarcates the duration of upper esophageal sphincter opening. The proposed method achieved more than 90% accuracy and similar values of sensitivity and specificity when compared to human ratings even when tested over swallows from an independent clinical experiment. Moreover, the predicted opening and closure moments surprisingly fell within an inter-human comparable error of their human rated counterparts which demonstrates the clinical significance of high resolution cervical auscultation in replacing ionizing radiation-based evaluation of swallowing kinematics.

Artificial Intelligence and Dysphagia: Novel Solutions to Old Problems

“…Our results demonstrated that we can detect these two events noninvasively (i.e. with HRCA) with over 90% accuracy (7) .…”

“…We call these recordings high-resolution cervical auscultation (HRCA) recordings. To determine the UES opening and closure through HRCA, we trained a deep neural network (7,8) . Neural networks are meant to simulate activity in the human brain.…”

Characterizing Effortful Swallows from Healthy Community Dwelling Adults Across the Lifespan Using High-Resolution Cervical Auscultation Signals and MBSImP Scores: A Preliminary Study

“…Afterward, motion artifacts and low frequency noise, such as head movement, were removed using fourth-order splines. Finally, wavelet denoising was used to eliminate any additional noise that might exist in the signals [ 17 , 20 , 21 ]. The onset and offset of swallows were taken from the segmented videos after applying the proper sampling mapping between videos and signals.…”

“…HRCA is a method of characterizing swallow function that integrates information from acoustic and vibratory signals from non-invasive sensors (contact microphone, tri-axial accelerometer) attached to the anterior laryngeal framework during swallowing. Following collection of HRCA signals, HRCA signal features are extracted using advanced signal processing techniques to use the HRCA signal features as input to machine learning algorithms to provide insight into swallowing physiology using human ratings of VF images as the “ground truth.” HRCA has demonstrated promise as a dysphagia screening method and potential diagnostic adjunct to VF by classifying safe and unsafe swallows (as measured by the penetration-aspiration scale) [ 11 – 17 ], tracking hyoid bone displacement in healthy adults and patients with suspected dysphagia [ 18 , 19 ], annotating temporal swallow kinematic events in healthy adults and patients with suspected dysphagia (e.g., durations of upper esophageal sphincter opening and laryngeal vestibule closure) [ 20 – 22 ], categorizing swallows between healthy participants and different patient populations [ 23 , 24 ], and detecting clinical ratings of swallow physiology in patients with suspected dysphagia using the Modified Barium Swallow Impairment Profile (MBSImP) [ 25 ] with a high degree of accuracy [ 19 , 21 ]. However, the utility of HRCA’s capabilities to noninvasively characterize these physiologic events, many of which are targets of behavioral augmentation via compensatory swallowing maneuvers (e.g., effortful swallow, Mendelsohn maneuver), and differentiate between swallows in which they are accurately deployed without imaging verification, has yet to be investigated.…”

A review of Hidden Markov models and Recurrent Neural Networks for event detection and localization in biomedical signals

“…In swallowing, acceleration signals have been used for the detection of pharyngeal swallowing activity via maximum likelihood methods with minimum description length in [16] and using short time Fourier transform and neural networks in [14]. RNNs were also employed for event detection in swallowing acceleration signals including the upper esophageal sphincter opening in [15,190], laryngeal vestibule closure [191], and hyoid bone motion during swallowing [192]. In gait analysis, HMMs were used for recognition and extraction in multiple occasions [193][194][195][196] as well as RNNs [197][198][199].…”