A swallowing assessment system based on dual-position multi-source sensing fusion

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Signal Processing

Background:

• Traditional swallowing disorder tests require specialized clinical settings, limiting early detection.
• Intelligent swallowing research using multi-source sensing fusion offers a potential solution for daily pre-screening.

Purpose of the Study:

• To develop and validate a novel swallowing assessment system integrating wireless sensing and multi-source data fusion.
• To evaluate the efficacy of decision-level and feature-level fusion strategies for swallowing disorder detection.

Main Methods:

• A wireless platform collecting 6-axis inertial measurement unit (IMU) and dual-channel photoplethysmography (PPG) signals from two body positions.
• Implementation of both decision-level fusion (weighting models) and feature-level fusion (stacking features) using sensor data.
• Validation on 21,024 data segments and 219 Water Swallowing Test recordings.

Main Results:

• The proposed system achieved average F1 scores of 81.1% (decision-level) and 80.8% (feature-level) fusion, outperforming single-sensor or single-position models.
• Sensor and position comparisons revealed distinct information contributions, with IMU and central positioning showing advantages.
• A simplified two-level classification for the Water Swallowing Test yielded accuracies of 81.0% and 71.3%.

Conclusions:

• The developed multi-source sensing fusion system effectively monitors swallowing function and enables daily recognition.
• The system provides a promising approach for early pre-screening of swallowing disorders, particularly for the Water Swallowing Test.