Characterization of Hepatocellular carcinoma in CT images using Higher Order Statistics and Daubechies Wavelet Coefficient

Abstract

The aim of study to characterize Hepatocellular carcinoma (HCC) in CT images using higher order statistic and Daubechies wavelet based on texture analysis. for classification and delineation of the HCC and normal liver, spine and ribs, and it’s a method to improve the accuracy of the diagnosis and to reduce the number of required invasive procedures. This study was conducted at five hospitals Darelaj specialized hospital, Alnilein Medical Diagnostic Center, Modern Medical Center and Royal care international hospital in Khartoum state during the period May 2014 to September 2016. The study sample included 180 patients with Hepatocellular (HCC) carcinoma underwent abdominal CT Images, from both gender. For higher order statistic the texture reveals a different underlying pattern of the HCC compared to the liver and other abdominal tissues with classification sensitivity 98.8%, and the combination of the texture features throughout the different triple phase image phases provides the highest predictive overall accuracy of 85.4 % using stepwise linear discriminant analysis. The Daubechies wavelet measures the gray level variations in a CT images, and it complements the coefficient of Daubechies wavelet Features extracted from the coefficient can be used to estimate the size distribution of the sub patterns. The Daubechies wavelet and its features seem very useful in texture classification. The classification accuracy of hepatocellular carcinoma 97.1 %, liver accuracy 91.7 %, While the spine and ribs showed a classification accuracy of 97.1, 91.2 % respectively. This study proposed that texture analysis is superior to visual perception system where texture revealed that change and the difference of the image pattern objectively in respect to the ground truth.