Prediction of cirrhosis from liver ultrasound B-mode images based on Laws´ masks analysis

In this present work, a technique for differentiation of normal and cirrhotic liver segmented regions of interest (SROIs) based on Laws´ masks analysis is reported. Thirty four B-mode ultrasound images taken from 22 normal volunteers and 12 patients suffering from liver cirrhosis were collected from Department of Radiodiagnosis and Imaging, PGIMER, Chandigarh, India. The filtered texture images are obtained by convolving the SROIs with twenty five, 2D (5×5) special filters based on laws´ masks. Metrics that can quantify the texture can be obtained by computing the statistics from these filtered texture images. Similar features are combined to remove the directional information as texture directionality is not important here. This results into 15 rotational invariant filtered texture images for each SROI. For each of the filtered images, five statistics namely, mean, standard deviation, skewness, kurtosis and energy are computed. Thus, a total of 75 Laws´ texture features (15 filtered texture images × 5 statistical features) are computed for 82 normal SROIs and 39 cirrhotic SROIs taken from 34 B-Mode ultrasound liver images. Correlation based feature selection (CFS) method is used to find the optimal subset of Laws´ texture features which can provide best discrimination between normal and cirrhotic SROIs. It has been observed that CFS method results in an optimal subset of 8 Laws´ texture features {LLmean, LLsd, LEsd, SSskewness, RRenergy, LEenergy, LSenergy and LWenegy}. The classification performance of neural network (NN) classifier is compared with support vector machine (SVM) classifier. By using all 75 Laws´ texture features the classification accuracy of 90.08% and 90.90% is obtained with NN and SVM classifier respectively. By using 8 Laws´ features selected by CFS method the classification accuracy of 91.73% and 92.56% is obtained with NN and SVM classifier respectively. From the comparison it is can be concluded that only 8 Laws´ tex- ure features namely {LLmean, LLsd, LEsd, SSskewness, RRenergy, LEenergy, LSenergy and LWenegy} can be used to build an efficient computer aided diagnostic (CAD) system for predicting of liver cirrhosis.