Electromagnetic Based Fatty Liver Detection Using Machine Learning

Fatty Liver Disease (FLD) is becoming prevalent disease in nowadays lifestyle while not being restricted to individuals with uncontrolled alcohol intake. Early diagnosis can prevent advanced irreversible liver disease, liver failure and ultimately can save lives. Computational methods reduce the operator dependability and hence improve diagnostic reliability. Most of these methods are based on automated analysis of Computed Tomography (CT), Ultrasound (US), and Magnetic Resonance (MRI) images. Besides the high costs and harmful radiation involved in the conventional imaging tools, the outcome of the automated imaging strongly depends on the image quality. To address the shortcomings of current tools, we propose an electromagnetic system, including an antenna operating across the band 0.4-1 GHz as a data acquisition device and a supervised Machine Learning (ML) framework to learn an inferring model for FLD directly from collected data. This paper reports the system configuration, ML problem setup and the obtained results, which show an accuracy of more than 97% for the simulated torso model.