Multimodal framework for swallow detection in video-fluoroscopic swallow studies using manometric pressure...

Area of Science:

• Biomedical Engineering
• Clinical Diagnostics
• Swallowing Disorders

Background:

• Oropharyngeal dysphagia is common in head and neck cancer (HNC) patients, affecting up to 50%.
• Comprehensive swallowing assessment often involves simultaneous video-fluoroscopic swallow studies (VFSS) and high-resolution impedance manometry (HRIM).
• Current methods face challenges due to high clinical workload and complex data analysis.

Purpose of the Study:

• To develop an automated framework for swallow detection in simultaneous VFSS-HRIM examinations for HNC patients.
• To address the limitations of existing software in clinical data collection and analysis.
• To reduce clinician workload while maintaining high accuracy in swallowing assessment.

Main Methods:

• Proposed a framework for automatic swallow detection using VFSS videos and HRIM data.
• Employed an optimized double-sweep optical flow algorithm for identifying candidate swallow intervals.
• Utilized a pressure-based swallow template with features from upper esophageal sphincter sensors for classification.

Main Results:

• Evaluated on 97 swallows from 12 HNC patients.
• Achieved 95% Recall and 92% F1-score for swallow detection.
• Reduced the need for HRIM annotations by 63%, significantly decreasing clinician workload.

Conclusions:

• The developed framework enables accurate, low-input swallow detection in HNC patients undergoing VFSS-HRIM.
• It overcomes current software limitations, facilitating scalable, objective, and multimodal swallowing assessments.
• This technology lays the foundation for improved diagnostic capabilities in HNC-related dysphagia.