Abnormal pattern detection in Wireless Capsule Endoscopy images using nonlinear analysis in RGB color space.

Area of Science:

• Medical Imaging
• Gastroenterology
• Computer Vision

Background:

• Wireless Capsule Endoscopy (WCE) allows non-invasive gastrointestinal tract visualization.
• Reviewing extensive WCE image data is time-consuming.
• Ulcers are a common gastrointestinal disease requiring accurate detection.

Purpose of the Study:

• To develop a novel technique for distinguishing pathogenic endoscopic images related to gastrointestinal ulcers.
• To automate the analysis of WCE images for ulcer detection.

Main Methods:

• Applied Bidimensional Ensemble Empirical Mode Decomposition to WCE RGB images to extract Intrinsic Mode Functions (IMFs).
• Utilized lacunarity analysis to quantify texture patterns for discriminating ulcer regions from normal mucosa.
• Employed a classification approach to differentiate abnormal from normal WCE images.

Main Results:

• The proposed method successfully extracted structural differences using IMFs across various scales.
• Lacunarity analysis effectively quantified texture variations between ulcerous and normal tissues.
• Achieved promising classification accuracy exceeding 95% in distinguishing ulcer images.

Conclusions:

• The developed technique shows high potential for automated WCE image analysis.
• This approach can significantly reduce the time required for reviewing WCE data.
• Offers a promising tool for improving the diagnosis of gastrointestinal ulcers.