Improving the Classification of Cirrhotic Liver by using Texture Features.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Hepatology

Background:

• Liver cirrhosis diagnosis relies on imaging and clinical data.
• Accurate and early detection of cirrhosis is crucial for patient outcomes.
• Developing automated diagnostic tools can improve efficiency and consistency.

Purpose of the Study:

• To develop and evaluate a computer-aided diagnosis (CAD) system for distinguishing liver cirrhosis in MRI scans.
• To integrate shape and texture features for improved diagnostic accuracy.
• To assess the effectiveness of an artificial neural network (ANN) in classifying cirrhosis.

Main Methods:

• Utilized MRI images for liver cirrhosis detection.
• Calculated two shape features from segmented liver regions.
• Quantified seven texture features using the grey level difference method (GLDM) in regions-of-interest (ROIs).
• Integrated features into a three-layer feed-forward artificial neural network (ANN).
• Defined cirrhosis presence based on the degree of ROIs exceeding a threshold.

Main Results:

• The ANN successfully learned patterns from training data.
• The system correctly identified 82% of cirrhosis cases.
• The system achieved 100% accuracy in identifying normal cases.
• 18 test cases were evaluated in total.

Conclusions:

• The proposed CAD system demonstrates effectiveness in predicting liver cirrhosis from MRI.
• The combination of shape and texture analysis with ANN shows promise for automated diagnosis.
• Further validation is warranted for clinical application.