Advanced Machine Learning Tools to Monitor Biomarkers of Dysphagia: A Wearable Sensor Proof-of-Concept Study

Abstract

<b><i>Introduction:</i></b> Difficulty swallowing (dysphagia) occurs frequently in patients with neurological disorders and can lead to aspiration, choking, and malnutrition. Dysphagia is typically diagnosed using costly, invasive imaging procedures or subjective, qualitative bedside examinations. Wearable sensors are a promising alternative to noninvasively and objectively measure physiological signals relevant to swallowing. An ongoing challenge with this approach is consolidating these complex signals into sensitive, clinically meaningful metrics of swallowing performance. To address this gap, we propose 2 novel, digital monitoring tools to evaluate swallows using wearable sensor data and machine learning. <b><i>Methods:</i></b> Biometric swallowing and respiration signals from wearable, mechano-acoustic sensors were compared between patients with poststroke dysphagia and nondysphagic controls while swallowing foods and liquids of different consistencies, in accordance with the Mann Assessment of Swallowing Ability (MASA). Two machine learning approaches were developed to (1) classify the severity of impairment for each swallow, with model confidence ratings for transparent clinical decision support, and (2) compute a similarity measure of each swallow to nondysphagic performance. Task-specific models were trained using swallow kinematics and respiratory features from 505 swallows (321 from patients and 184 from controls). <b><i>Results:</i></b> These models provide sensitive metrics to gauge impairment on a per-swallow basis. Both approaches demonstrate intrasubject swallow variability and patient-specific changes which were not captured by the MASA alone. Sensor measures encoding respiratory-swallow coordination were important features relating to dysphagia presence and severity. Puree swallows exhibited greater differences from controls than saliva swallows or liquid sips (<i>p</i> < 0.037). <b><i>Discussion:</i></b> Developing interpretable tools is critical to optimize the clinical utility of novel, sensor-based measurement techniques. The proof-of-concept models proposed here provide concrete, communicable evidence to track dysphagia recovery over time. With refined training schemes and real-world validation, these tools can be deployed to automatically measure and monitor swallowing in the clinic and community for patients across the impairment spectrum.

AI-Based Detection of Aspiration for Video-Endoscopy with Visual Aids in Meaningful Frames to Interpret the Model Outcome

“…Regarding the aspiration detection task, the current model performance might be comparable to that of an untrained human examiner [ 21 ]. Hence, at present, our model does not lead to better results than comparable non-endoscopic/non-radiologic approaches [ 28 , 33 , 35 , 37 , 62 , 63 ]; but in clear contrast to them, our model outcomes, as well as the false positives and negatives, are fully interpretable and can therefore be corrected by an experienced examiner. This becomes particularly easy, because the examiner can perform the correct assignment by jumping to the respective point in the timeline of the video sequence.…”

“…Therefore, the final explanation provided by the system is effective and acceptable. This goes beyond most existing approaches for this task (except the VFSS approach [ 25 ]), because they only provide predictions or classifications without providing proper interpretable information for the diagnostician [ 28 , 35 , 37 , 42 , 43 , 62 , 63 ]. As this lack of transparency conflicts with EU GDPR, which prohibits decisions based solely on automated processing [ 44 , 45 ], a subsequent FEES or VFSS would become necessary, in any event, before critical decisions—such as abstinence from food, insertion of a nasogastric tube, or even re-intubation and tracheotomy—could be made.…”

Enhancing Nutrition Care Through Real-Time, Sensor-Based Capture of Eating Occasions: A Scoping Review

“…Unlike traditional event-contingent EMDA, passive sensor-based devices do not require participants to initiate the self-reporting of each eating event ( 19 ), reducing the number of missed occasions resulting from when users forget or decline to report ( 97 ). However, these devices will only be effective if eating or drinking gestures are detected instantly at the start of a main meal, snack, or beverage or even predicted before the event ( 74 ).…”

“…Nutrition Diagnosis involves the naming of a specific nutrition problem based on data collected during nutrition assessment. Sensors may be able to detect the signs and symptoms of nutrition problems such as excessive or inadequate energy intake based on the correlation between energy intake and HTM motions ( 18 ); swallowing or chewing difficulties using devices that measure these gestures to detect eating ( 19 ); and irregular eating patterns based on the detected start time of eating occasions and the consistency of these times across multiple days. This type of information collected by the sensor-based devices may assist dietitians in establishing a nutrition diagnosis more efficiently.…”

Soft skin-interfaced mechano-acoustic sensors for real-time monitoring and patient feedback on respiratory and swallowing biomechanics

“…These results demonstrate the feasibility in monitoring breathing and swallowing in a wireless, noninvasive manner, continuously and in nearly any scenario, even during natural daily activities. Such features are unavailable in any other technology 17,[35][36][37] , to the best of our knowledge.…”