Automatic hyoid bone detection in fluoroscopic images using deep learning

Zhenwei Zhang, James L. Coyle, Ervin Sejdić
2018 Scientific Reports  
The displacement of the hyoid bone is one of the key components evaluated in the swallow study, as its motion during swallowing is related to overall swallowing integrity. In daily research settings, experts visually detect the hyoid bone in the video frames and manually plot hyoid bone position frame by frame. This study aims to develop an automatic method to localize the location of the hyoid bone in the video sequence. To automatically detect the location of the hyoid bone in a frame, we
more » ... osed a single shot multibox detector, a deep convolutional neural network, which is employed to detect and classify the location of the hyoid bone. We also evaluated the performance of two other state-of-art detection methods for comparison. The experimental results clearly showed that the single shot multibox detector can detect the hyoid bone with an average precision of 89.14% and outperform other autodetection algorithms. We conclude that this automatic hyoid bone tracking system is accurate enough to be widely applied as a pre-processing step for image processing in dysphagia research, as well as a promising development that may be useful in the diagnosis of dysphagia. Dysphagia, a common condition among older individuals, is defined as an impairment in swallowing function during eating and drinking 1 . Dysphagia causes subjective discomfort and objective difficulty in the formation or transportation of a bolus from mouth to stomach, and prevention of errant entry of swallowed material into the airway. Dysphagia is a frequent clinical sign in patients with stroke, head and neck cancer and a variety of other medical conditions 2-4 . The prevalence of dysphagia is very high: stroke is the most commonly reported etiology with over 50% of patients exhibiting dysphagia in the immediate post-onset stage of recovery, diminishing to a lower prevalence of around 11% within 6 months of onset 5 . Additionally, chronic dysphagia affects 7.2% of people with other neurological diseases, and 4.9% of patients treated for head and neck cancer 6 . Up to 40% of people over 65 years old and more than 60% of adults in nursing home 7 suffer from dysphagia. It is estimated that 25-50% of Americans over 60 2 and 17% of citizens over 65 in Europe 8 will suffer from dysphagia, leading to increased risk of poor nutrition or dehydration. The variation in estimation may be due to different definitions of dysphagia, the method of swallowing assessment and the number of patients investigated. As a more immediate clinical consequence, dysphagia may lead to misdirection of food and colonized saliva into the airway, possibly causing pneumonia and chronic lung disease. In many cases aspiration occurs without any obvious clinical signs of dysphagia (silent aspiration), postponing early identification and preventive treatment therefore lowering patient survival 9 . Efforts to accurately evaluate swallowing function early after the onset of conditions leading to dysphagia can mitigate many of these health risks 10 . The videofluoroscopic swallowing studies (VFSS), also known as modified barium swallow study, is the gold standard test for dysphagia evaluation 11-14 . VFSS, unlike bedside clinical examination, enables the examiner to visualize oral, pharyngeal and upper esophageal structure and function during patient swallowing. VFSS also evaluate errors of biomechanical coordination that lead to bolus misdirection. Patients with dysphagia may not exhibit overt signs of swallowing problems at the bedside. VFSS excels at allowing clinicians to identify occult disorders in airway protection and biomechanical errors leading to impaired airway protection and transfer of food to the digestive system. Airway closure and upper esophageal sphincter opening are largely influenced by the timing and displacement of the hyolaryngeal complex during the pharyngeal stage of swallowing. During VFSS, the hyoid bone is the most salient anatomic structure for detecting hyolaryngeal motion 15 . Hyolaryngeal excursion is
doi:10.1038/s41598-018-30182-6 pmid:30120314 pmcid:PMC6097989 fatcat:ri4sk23ogjelxeavz2lj55jgqm