Edge density based automatic detection of inflammation in colonoscopy videos

Area of Science:

• Medical Imaging
• Computational Pathology
• Gastroenterology

Background:

• Early detection of colon cancer significantly improves patient survival rates.
• Inflammation and polyps are key indicators of early-stage colon cancer.
• Current diagnostic tools rely on colonoscopy or capsule endoscopy videos.

Purpose of the Study:

• To investigate a simple, model-based texture analysis approach for detecting inflamed tissue in colon videos.
• To develop a real-time, parallel-processing compatible method for anomaly detection.
• To improve the accuracy of automated colon cancer screening.

Main Methods:

• Utilized texture analysis to differentiate between inflamed and healthy colon tissues.
• Developed and implemented a specific filter kernel to detect characteristic inflamed tissue textures.
• Enhanced the method to mitigate the interference of blood vessels in video analysis.

Main Results:

• The proposed texture analysis method achieved real-time detection of inflamed regions with over 84% accuracy.
• Specific video frame segments containing inflammation were detected with an accuracy exceeding 90%.
• The method demonstrated efficient parallel processing capabilities for real-time applications.

Conclusions:

• Texture analysis offers a viable and efficient alternative for automated detection of inflamed tissues in colon videos.
• The developed method shows promise for enhancing early colon cancer detection through improved video analysis.
• Further development could lead to more accurate and accessible colon cancer screening tools.