Portable Electromagnetic Device for Steatotic Liver Detection Using Blind Source Separation and Shannon Wavelet Entropy

Hepatic steatosis is the most common cause of chronic liver disease worldwide. Symmetric placement of sensors is the main challenge of previous electromagnetic hepatic steatosis techniques as they require signals from the left side of the torso as a healthy reference. To tackle this issue, this study proposes an electromagnetic technique to assist in observation and detection of hepatic steatosis only by measuring electromagnetic signals from the right side of the torso where liver is located. The combination of blind source separation and wavelet transform techniques is used for detection. Firstly, the principal component analysis is applied to the time-domain converted electromagnetic signals to sphere the measured data. Secondly, the sphered signals are modelled as a linear mixture of different components and separated utilizing independent component analysis technique. Thirdly, the liver response signal is isolated based on its lowest cross-correlation with the average of input signals. Finally, Shannon wavelet entropy and wavelet energy of the estimated liver signal is evaluated by considering the excitation electromagnetic Gaussian pulse as the mother wavelet. Healthy livers can be differentiated from the steatotic ones based on their higher level of wavelet entropy. The method is validated by realistic numerical models and human measurements. In both scenarios, the wavelet entropy of estimated liver response is significantly lower and distinguishable in steatotic livers. The proposed method can detect HS cases with more than 16% diffused fat. Compared to previous techniques, it requires significantly less scanning time, is independent of antennas symmetrical placement, and is thus more reliable.

Automated summary

This study proposes an electromagnetic technique to assist in observation and detection of hepatic steatosis only by measuring electromagnetic signals from the right side of the torso. The method uses blind source separation and wavelet transform techniques for detection, and can differentiate healthy livers from steatotic livers.