A knowledge-based framework for robust segmentation of high-resolution impedance manometry catheters in...

Area of Science:

• Medical Imaging
• Gastroenterology
• Biomedical Engineering

Background:

• Simultaneous high-resolution impedance manometry (HRIM) and videofluoroscopic swallow studies (VFSS) offer comprehensive swallowing analysis but require independent interpretation.
• Clinician workload is increased by analyzing HRIM and VFSS separately.
• Defining manometric regions on VFSS images remains a challenge.

Purpose of the Study:

• To develop and validate an automated algorithm for detecting the HRIM catheter within VFSS images.
• To enable direct spatial registration of HRIM data onto VFSS frames.
• To reduce the analytical burden on clinicians by integrating HRIM and VFSS data.

Main Methods:

• A template-free, knowledge-based algorithm was developed to automatically localize the HRIM catheter centerline in VFSS frames.
• The algorithm identifies the catheter by recursively adding segments based on proximity and directional alignment.
• Validation was performed on 122 single-swallow VFSS videos from head and neck cancer patients.

Main Results:

• The catheter segmentation module achieved high performance metrics.
• Precision was 93.8%, recall was 83.8%, and the F1-score was 88.5% with a 1.77 mm tolerance.
• The algorithm demonstrated robust performance across varied anatomies and imaging conditions.

Conclusions:

• The developed algorithm successfully automates HRIM catheter detection in VFSS, facilitating direct manometric region delineation on imaging data.
• The knowledge-based approach, relying on geometric priors, provides consistent and interpretable results without extensive annotated datasets.
• This framework has the potential to streamline swallow analysis and can be adapted for other clinical applications involving radiopaque structures.