Non-invasive Quantification of Steatosis: A New Ultrasound based Model to Predict Fatty Liver Content

Non-alcoholic fatty liver disease (NAFLD) is becoming one of the most important causes of chronic liver disease and the quantification of the liver fat content is of great importance in diagnosing and monitoring NAFLD progression. The non-''invasive gold standard for quantifying liver steatosis is magnetic resonance imaging (MRI), but it is too expensive and not easily available. For this reason, ultrasound (US) imaging techniques are increasingly used. The main aim of this study is to develop and to validate a multiparametric system based on US able to assess the percentage of liver fat, producing a result equivalent to that computed by MRI. We have updated a previously proposed US-based image processing tool (DOI: 10.1016/j.ultrasmedbio.2018.03.011.). We tested the software on N=135 US images acquired from patients. In addition, we developed and validated a multiparametric model that combines the US-parameters extracted to predict the liver fat content, using magnetic resonance spectroscopy (MRS) and MRI proton density fat fraction (MRI- PDFF) computed values as a ground truth. We analyzed the inter-operator reproducibility between a technical and a clinical operator. We show the prediction performance of our model compared to MRI computed fat values, expressed as a correlation coefficient of 0.84. Finally, regarding the evaluation of the inter-operator variability we achieved an intraclass correlation coefficient higher than 0.9 for the final score assessed by the two different operators. The proposed system is quite inexpensive, easy to use and may be implemented on any US equipment as a tool for screening and monitoring fatty liver disease.