Machine Learning-Based Classification of Abnormal Liver Tissues Using Relative Permittivity.

Area of Science:

• Biomedical Engineering
• Medical Diagnostics
• Computational Biology

Background:

• Non-invasive diagnostic tools are crucial for early disease detection.
• Dielectric properties of biological tissues vary between healthy and diseased states.
• Microwave-based dielectric measurements offer unique insights into tissue characteristics.

Purpose of the Study:

• To characterize dielectric properties of liver tissues across different disease models.
• To explore the potential of machine learning for classifying liver tissue types based on dielectric data.
• To assess the feasibility of developing non-invasive diagnostic tools for liver diseases.

Main Methods:

• Dielectric spectroscopy was used to measure dielectric properties (relative permittivity and conductivity) of liver tissues.
• Mouse models included healthy, non-alcoholic steatohepatitis (NASH) induced by two diets, and liver fibrosis.
• Multi-class classification machine learning models, including Support Vector Machine (SVM), were employed.

Main Results:

• Dielectric properties showed significant variations between healthy and diseased liver tissues.
• The SVM model achieved up to 90% accuracy in differentiating between the four liver tissue groups.
• The study successfully demonstrated the classification of liver disease states using dielectric measurements.

Conclusions:

• Dielectric spectroscopy combined with machine learning shows promise for non-invasive liver disease diagnosis.
• This approach can differentiate between healthy, NASH, and fibrotic liver tissues.
• The developed technology pipeline has the potential for next-generation diagnostic tools.