Automated Dysphagia Screening Using Noninvasive Neck Acoustic Sensing
Abstract
Pharyngeal health plays a vital role in essential human functions such as breathing, swallowing, and vocalization. Early detection of swallowing abnormalities, also known as dysphagia, is crucial for timely intervention. However, current diagnostic methods often rely on radiographic imaging or invasive procedures. In this study, we propose an automated framework for detecting dysphagia using portable and noninvasive acoustic sensing coupled with applied machine learning. By capturing subtle acoustic signals from the neck during swallowing tasks, we aim to identify patterns associated with abnormal physiological conditions. Our approach achieves promising test-time abnormality detection performance, with an AUC-ROC of 0.904 under 5 independent train-test splits. This work demonstrates the feasibility of using noninvasive acoustic sensing as a practical and scalable tool for pharyngeal health monitoring.
The following analysis is generated by AI and may contain errors. It is separate from the authors’ abstract and is not peer review.
Research summary
This paper presents a novel automated framework for dysphagia screening using noninvasive acoustic sensing, showcasing promising results and a strong methodological foundation. The study's contributions are significant, particularly in advancing the use of machine learning in clinical diagnostics, although further validation and resource availability are necessary for broader application.
Methodology
The proposed methodology leverages noninvasive acoustic sensing to detect dysphagia, which is innovative in its approach to using audio signals from the neck during swallowing tasks. The authors employed a comprehensive data collection strategy, utilizing fiberoptic endoscopic evaluation of swallowing (FEES) to gather a robust dataset of swallow events. The use of multiple machine learning models, including established audio embedding models, and the incorporation of domain-informed features demonstrates a thoughtful approach to feature extraction. However, the methodology could benefit from further details on the model training process and hyperparameter tuning.
Results & evaluation
The experiments are well-structured, with a clear focus on patient-level train-test splits to prevent overfitting. The reported AUC-ROC scores indicate strong classification performance, particularly in the binary classification task. The authors also explore various segmentation strategies, which is crucial for real-world applications. However, the reliance on a moderate-sized dataset (617 swallow events) limits the generalizability of the findings, particularly in multiclass settings.
Reproducibility
The paper provides a reasonable level of detail regarding the data collection and preprocessing steps, which aids in reproducibility. However, the absence of publicly available code or data limits the ability for other researchers to replicate the study fully. Including a link to a repository with the code and data would significantly enhance reproducibility.
Limitations
The study acknowledges limitations such as the moderate dataset size, which may hinder the model's ability to generalize across diverse populations. Additionally, the segmentation methods could be further refined to improve accuracy in real-time applications. The authors also note that the performance in multiclass classification was lower, indicating a need for larger, more balanced datasets.
Broader impact
This research has significant implications for the field of digital health, particularly in developing low-cost, noninvasive diagnostic tools for dysphagia. The potential for real-time monitoring of swallowing health could lead to improved patient outcomes and greater accessibility to care. The approach could also inspire further research into acoustic sensing applications in other areas of health monitoring. This paper presents a novel automated framework for dysphagia screening using noninvasive acoustic sensing, showcasing promising results and a strong methodological foundation. The study's contributions are significant, particularly in advancing the use of machine learning in clinical diagnostics, although further validation and resource availability are necessary for broader application.
Author of this paper? Send a correction.