HepNet: Deep Neural Network for Classification of Early-Stage Hepatic Steatosis Using Microwave Signals.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Biomedical Engineering

Background:

• Hepatic steatosis, or fatty liver disease, is a prevalent condition often diagnosed late.
• Early detection of hepatic steatosis is crucial for managing chronic liver diseases effectively.
• Current diagnostic methods for hepatic steatosis can be invasive or lack sensitivity for early stages.

Purpose of the Study:

• To develop and validate a novel deep learning-based classifier for early detection of hepatic steatosis.
• To utilize microwave technology for non-invasive assessment of liver fat content.
• To improve the accuracy and reliability of hepatic steatosis diagnosis through advanced computational methods.

Main Methods:

• Development of a deep learning model, HepNet, incorporating convolutional layers with skip connections.
• Generation of extensive simulation data using 3D electromagnetic tools for model training and validation.
• Application of transfer learning techniques to adapt the model for clinical data with limited patient samples.
• Validation against 1H-MRS (proton magnetic resonance spectroscopy) as the gold standard in clinical trials.

Main Results:

• HepNet achieved a high F1-score of 0.91 in simulations, outperforming traditional models like LeNet and ResNet.
• The model demonstrated robust performance with a confidence level of 0.97 for classifications with low entropy.
• Clinical validation showed high F1-scores of 0.95 and 0.88 on patient datasets of 94 and 158 samples, respectively.

Conclusions:

• The developed HepNet model shows significant potential for accurate and reliable early detection of hepatic steatosis.
• Microwave technology combined with deep learning offers a promising non-invasive approach for liver disease diagnosis.
• The study highlights the clinical applicability and robustness of the proposed method for hepatic steatosis assessment.