Detection of small bowel tumors in capsule endoscopy frames using texture analysis based on the discrete wavelet...

Area of Science:

• Medical Imaging
• Gastroenterology
• Computer Vision

Background:

• Capsule endoscopy is crucial for diagnosing small bowel tumors.
• Image color and texture contain vital diagnostic information.
• Accurate tumor detection relies on advanced image analysis techniques.

Purpose of the Study:

• To develop and validate an automated method for tumor lesion detection in capsule endoscopy images.
• To leverage textural analysis and wavelet transforms for enhanced image feature extraction.
• To improve the diagnostic accuracy of capsule endoscopy through computational methods.

Main Methods:

• Textural analysis of color channels in capsule endoscopic images.
• Wavelet transform for selecting significant texture information bands.
• Image synthesis using inverse wavelet transform.
• Second-order textural feature extraction.
• Classification using a multilayer perceptron neural network.

Main Results:

• The proposed algorithm achieved 98.7% sensitivity in detecting tumor lesions.
• The method demonstrated 96.6% specificity in real-world capsule endoscopy data.
• The results confirm the algorithm's high diagnostic performance.

Conclusions:

• The developed algorithm is feasible and effective for diagnosing small bowel tumors.
• Texture analysis combined with wavelet transform significantly enhances capsule endoscopy diagnostics.
• This computational approach offers a promising tool for clinical gastroenterology.