Non-invasive quantitative diagnosis of liver fibrosis with an artificial neural network

Hepatic fibrosis is the body’s response to chronic liver disorders caused by various causes. Ultrasonic examination using an intelligent algorithm is increasingly important for the diagnosis of hepatic fibrosis. The purpose of this study was to analyse the preliminary application of artificial neural networks (ANNs) combined with ultrasonic quantitative diagnosis of hepatic fibrosis. In this study, 93 patients with hepatic fibrosis in our hospital were enrolled. The ultrasound image data of patients with chronic liver disease and the normal control group were collected, and 10 ultrasound examination indexes were quantified. The liver fibrosis grade of patients with chronic liver disease was confirmed by ultrasound-guided liver biopsy. After statistical analysis of all the ultrasonic data, the meaningful indexes were selected as the input layer, and the pathological grading results of liver fibrosis were selected as the output layer. The results showed that grades 2 and 4 hepatic fibrosis were diagnosed by ANN. The sensitivity, specificity and accuracy of the artificial neural network model were 95.4%, 96.2% and 95.8%, respectively. The preliminary artificial neural network model of ultrasound diagnosis of early liver cirrhosis has high sensitivity and specificity. In this study, the accuracy and specificity of ultrasound diagnosis of hepatic fibrosis were greatly improved. It is concluded that this is very helpful for the treatment of the disease.