Trying to provide ultrasonic image-aid measures for quantitative diagnosis and dynamic monitoring of liver fibrosis, we propose two scoring methods for liver fibrosis tissue in vivo , based on ultrasound radio frequency (RF) time series in this paper. Firstly, RF echo signals of human liver were recorded in this study. Then one of the recorded frame RF data was demodulated to be B model image. After that, a region of interest (ROI) in the B model image was selected. For each point in the ROI, its all frame data were acquired so that RF time series were formed. An SMR (size measure relationship) fractal dimension and six spectral features were extracted from RF time series in the ROI. With relative deviation and Fisher's discriminant ratio, seven features were weighted and summed so that the liver tissues' scores were obtained, Score-rd and Score-fisher, respectively. Area under ROC curve (AUC) and a support vector machine (SVM) were used to evaluate whether these scoring methods would be useful in distinguishing normal and cirrhosis tissues. Experimental results are shown as follows: Score-rd's AUC was 0.843, while Score-fisher was 0.816, SVM classification accuracies were both up to 87.5%. This proved that our proposed scoring methods were effective in distinguishing normal and cirrhosis tissues. Score-rd and Score-fisher have potential for clinical applications. They can also provide quantitative references for liver fibrosis diagnosis.
Area Under Curve ; Fractals ; Humans ; Liver Cirrhosis ; classification ; diagnostic imaging ; Radio Waves ; Support Vector Machine ; Ultrasonography