Subjective and Objective Quality Assessment of Colonoscopy Videos.

Area of Science:

• Medical Imaging
• Computer Vision
• Signal Processing

Background:

• Colonoscopy videos often have real-world distortions (blur, low brightness, occlusion) hindering interpretation.
• Existing methods primarily use synthesized distortions, which differ significantly from actual video degradation.

Purpose of the Study:

• To conduct an in-depth study of colonoscopy Video Quality Assessment (VQA).
• To develop both subjective and objective solutions for VQA in colonoscopy.

Main Methods:

• Collected 1,000 colonoscopy videos with practical quality degradation.
• Constructed a multi-attribute VQA database with subjective annotations for five distortion attributes and overall quality.
• Proposed Distortion Attribute Reasoning Network (DARNet) with spatial and temporal feature extraction streams.
• Utilized multi-attribute association and behavior-guided reasoning modules for attribute-aware feature fusion.

Main Results:

• DARNet demonstrated strong correlation with subjective quality ratings.
• The proposed method outperformed nine state-of-the-art VQA techniques.
• The developed database serves as a benchmark for colonoscopy VQA research.

Conclusions:

• DARNet offers an effective objective solution for colonoscopy VQA.
• Addressing real-world distortions is crucial for accurate video quality assessment.
• The study advances the field of medical video quality evaluation.