Liver fat analysis using optimized support vector machine with support vector regression.

Area of Science:

• Medical Imaging Analysis
• Machine Learning in Healthcare
• Liver Disease Diagnostics

Background:

• Fatty liver disease, including alcoholic and non-alcoholic forms (NAFLD), affects a significant portion of the population.
• Epidemiological studies indicate prevalence rates of 9% to 32% in India, particularly among overweight individuals.
• Accurate diagnosis and quantification of liver fat are crucial for patient management.

Purpose of the Study:

• To propose an Optimized Support Vector Machine with Support Vector Regression (LFA-OSVM-SVR) model for evaluating liver fat volume.
• To leverage image analysis techniques for non-invasive fatty liver assessment.
• To enhance the accuracy of fatty liver detection and classification through advanced machine learning.

Main Methods:

• Utilized computed tomography (CT) liver images from Chennai liver foundation and Liver Segmentation (LiTS) datasets.
• Pre-processed images using Gaussian smoothing and bypass filters to reduce noise and enhance intensity.
• Implemented U-Net for liver segmentation, Optimized Support Vector Machine for classification, and Support Vector Regression for fat percentage analysis.

Main Results:

• The LFA-OSVM-SVR model demonstrated effective analysis of liver fat from CT scan images.
• The proposed approach, implemented in Python, showed significant performance improvements.
• Quantitative analysis of efficiency was conducted using established performance metrics.

Conclusions:

• The LFA-OSVM-SVR method achieved superior accuracy and reduced error rates in classifying both fatty and normal liver images compared to existing methods like CNN-FDE, FCN-NMF, and DL-FCN.
• The model offers a promising advancement in the automated diagnosis and assessment of fatty liver disease.
• This AI-driven approach provides a more efficient and accurate alternative for liver fat evaluation.