Computer-aided detection of bleeding regions for capsule endoscopy images

Area of Science:

• Medical Imaging
• Computer-Aided Diagnosis
• Gastroenterology

Background:

• Capsule endoscopy (CE) generates numerous images, posing a significant burden on physicians for manual inspection.
• Automated methods are crucial for efficient analysis of CE data to diagnose digestive tract diseases.
• Bleeding detection in CE images is essential for timely patient management.

Purpose of the Study:

• To develop a computer-aided system for accurate detection of bleeding regions in capsule endoscopy images.
• To introduce and evaluate a novel color texture feature for improved bleeding identification.
• To enhance the diagnostic efficiency of capsule endoscopy through automated image analysis.

Main Methods:

• Proposed a novel computer-aided system for bleeding region detection in CE images.
• Introduced a new chrominance moment feature, utilizing Tchebichef polynomials and illumination invariant color spaces.
• Combined chrominance moment with uniform local binary patterns for color texture feature extraction.
• Employed a multilayer perceptron neural network for classification of normal and bleeding regions.

Main Results:

• The proposed color texture features demonstrated effectiveness in discriminating between normal and bleeding regions.
• Experimental results on a dedicated bleeding image dataset showed promising performance.
• The system showed potential for accurate and efficient detection of gastrointestinal bleeding from CE images.

Conclusions:

• The developed computer-aided system shows significant promise for detecting bleeding in capsule endoscopy images.
• The novel chrominance moment feature contributes effectively to color texture analysis for medical imaging.
• This approach can alleviate physician burden and improve diagnostic accuracy in capsule endoscopy.