A texture classification method for diffused liver diseases using Gabor wavelets

Area of Science:

• Medical Imaging
• Signal Processing
• Biomedical Engineering

Background:

• Accurate classification of diffused liver diseases is crucial for effective treatment.
• Existing texture analysis methods have limitations in precision and efficiency.
• Gabor wavelets offer optimal joint space-frequency resolution for texture feature extraction.

Purpose of the Study:

• To develop and evaluate an efficient Gabor wavelet-based method for classifying liver diseases.
• To compare the performance of Gabor wavelets against dyadic wavelets and statistical texture methods.
• To assess the method's effectiveness in discriminating between normal liver, hepatitis, and cirrhosis.

Main Methods:

• Utilized Gabor wavelets for texture extraction from ultrasonic liver images.
• Developed a classification algorithm leveraging the space-frequency properties of Gabor filters.
• Compared the proposed method with dyadic wavelets and statistical texture analysis techniques.
• Tested the algorithm on 45 biopsy-proven ultrasonic liver images for each of three states: normal, hepatitis, and cirrhosis.

Main Results:

• The Gabor wavelet method achieved higher classification accuracy than dyadic wavelets and statistical methods.
• Achieved 85% sensitivity in distinguishing normal from hepatitis liver images.
• Achieved 86% sensitivity in distinguishing normal from cirrhosis liver images.
• The proposed method generates a small feature vector, enhancing retrieval speed and robustness to image shifts.

Conclusions:

• Gabor wavelets are more appropriate for liver disease texture classification than dyadic wavelets and statistical methods.
• The proposed Gabor wavelet-based algorithm demonstrates high accuracy and efficiency for diagnosing liver conditions.
• This method shows promise for improving the non-invasive diagnosis of diffused liver diseases using ultrasound imaging.