Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Automated Dysphagia Screening Using Noninvasive Neck Acoustic Sensing

Jade Chng, Rong Xing, Yunfei Luo, Kristen Linnemeyer-Risser, Tauhidur Rahman, Andrew Yousef, Philip A Weissbrod

TL;DR

This work tackles dysphagia screening by developing an automated, noninvasive neck acoustic sensing system collected during FEES. It combines domain-informed acoustic features with pretrained audio representations and evaluates them on five patient-level splits, achieving an AUC-ROC of $0.904$ for abnormality detection. Domain-informed features outperform pretrained embeddings, with segmentation strategy and aggregation affecting performance and upper bounds approaching human-segmented benchmarks. The approach demonstrates feasibility for scalable, real-time pharyngeal health monitoring, while highlighting the need for larger, more diverse datasets and at-home validation.

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.

Automated Dysphagia Screening Using Noninvasive Neck Acoustic Sensing

TL;DR

This work tackles dysphagia screening by developing an automated, noninvasive neck acoustic sensing system collected during FEES. It combines domain-informed acoustic features with pretrained audio representations and evaluates them on five patient-level splits, achieving an AUC-ROC of for abnormality detection. Domain-informed features outperform pretrained embeddings, with segmentation strategy and aggregation affecting performance and upper bounds approaching human-segmented benchmarks. The approach demonstrates feasibility for scalable, real-time pharyngeal health monitoring, while highlighting the need for larger, more diverse datasets and at-home validation.

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.
Paper Structure (20 sections, 2 equations, 2 figures, 3 tables)

This paper contains 20 sections, 2 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Preliminary Experiments
  5. Data Annotation
  6. Feature Extraction and Modeling
  7. Frequency
  8. Amplitude
  9. Area Under the Curve
  10. Representations from Pretrained Audio Models.
  11. Segmentation and Aggregation Analysis
  12. Fixed Parameter Segmentation
  13. Sliding Window Segmentation
  14. Evaluation Protocol
  15. Results
  16. ...and 5 more sections

Figures (2)

  • Figure 1: Overview of our proposed automated system. (A) Data collected from participants during this study. (B) Demonstration of our data annotation process. (C) Modeling procedure to explore the relationship between acoustic signal and the swallow abnormalities. (D) Presentation of the empirical results.
  • Figure 2: SHAP Summary Plot of Top 8 Features from Performance on Human Segmented Swallows