Computer-Aided Image Enhanced Endoscopy Automated System to Boost Polyp and Adenoma Detection Accuracy

Area of Science:

• Gastroenterology
• Medical Imaging
• Computer Vision

Background:

• Colonoscopy is crucial for early colon polyp detection, but the polyp miss rate remains high.
• Narrowed-band imaging (NBI) enhances polyp visualization but has usability limitations.
• Novel image processing techniques are needed to improve polyp detection and characterization during colonoscopy.

Purpose of the Study:

• To simulate the Narrowed-band imaging (NBI) system using image processing techniques.
• To evaluate the effectiveness of simulated NBI methods in colon polyp identification and classification.
• To compare the performance of different color transfer algorithms for enhancing polyp detection.

Main Methods:

• Three image processing methods were selected to simulate NBI: Color Transfer with Mean Shift (CTMS), Multi-scale Retinex with Color Restoration (MSRCR), and Gamma and Sigmoid Conversions (GSC).
• These methods were applied to colonoscopy images to enhance polyp features.
• Performance was evaluated using classification accuracy and mean Average Precision (mAP).

Main Results:

• All tested color transfer methods outperformed the original colonoscopy images in polyp classification.
• The Multi-scale Retinex with Color Restoration (MSRCR) method demonstrated excellent performance.
• Color transfer significantly positively impacted polyp identification and classification tasks.

Conclusions:

• Simulating NBI with color transfer techniques, particularly MSRCR, can enhance colon polyp detection and classification accuracy.
• These findings suggest that advanced image processing can help reduce the colon polyp miss rate.
• Color transfer methods offer a promising approach to improve colonoscopy effectiveness in colon cancer prevention.