CFSCNet: A Coarse-Fine Stream Conformer Neural Network for Swallow Segmentation.

Area of Science:

• Biomedical Engineering
• Clinical Diagnostics
• Signal Processing

Background:

• Dysphagia, or difficulty swallowing, significantly impacts patient quality of life and requires accurate diagnostic tools.
• Current methods for swallowing event segmentation often struggle with complex cases and multimodal data integration.

Purpose of the Study:

• To develop an efficient and accurate method for segmenting complex swallowing events in patients with dysphagia.
• To enhance swallowing event segmentation by integrating multimodal signals: audio, nasal airflow, and tri-axial accelerometry.

Main Methods:

• Proposed CFSCNet, a novel dual-stream model integrating multimodal signals for swallowing event segmentation.
• Employed an exponential moving average-based cost-sensitive weighting to address data imbalance.
• Introduced a Swallowing State Machine-Driven Post-Processing Method for stable clinical predictions.

Main Results:

• CFSCNet achieved state-of-the-art performance on two benchmark datasets.
• Achieved an AUC of 89.83% on the HRCA dataset and 94.10% on the SMSD dataset.
• Significantly outperformed existing swallowing event segmentation methods.

Conclusions:

• The proposed framework offers a comprehensive and novel solution for swallowing event segmentation with high accuracy and reliability.
• CFSCNet demonstrates potential for clinical translation, advancing high-precision dysphagia detection and diagnosis.
• This work supports more effective swallowing rehabilitation and medical interventions for dysphagia patients.