Algorithm for video summarization of bronchoscopy procedures.

Area of Science:

• Medical Imaging
• Endoscopy Technology
• Pulmonology

Background:

• Bronchoscopy procedure duration varies, often exceeding 20 minutes for complex cases.
• Videobronchoscopy allows full procedure recording, but manual selection of relevant clips is common.
• Poor quality frames due to patient movement and secretions are frequent in bronchoscopy videos.

Purpose of the Study:

• To develop an automated system for generating summaries of bronchoscopy video recordings.
• To enhance the utility of recorded bronchoscopy procedures for diagnostic, reference, and educational purposes.
• To enable bronchoscopists to focus on procedures rather than manual video management.

Main Methods:

• Developed a prototype system utilizing image analysis algorithms.
• Implemented models from previous research for video summarization.
• Employed methods to identify and exclude frames or sequences of limited diagnostic or educational value.

Main Results:

• The system automatically detects and excludes non-informative frames.
• Identifies and removes frames showing airway branching.
• Recognizes and excludes frames containing pathological lesions.

Conclusions:

• The developed algorithm effectively generates summaries of bronchoscopy video recordings.
• Addresses the challenge of creating valuable content from extensive video data.
• Facilitates more efficient use of recorded bronchoscopy procedures.