Characterization of Liver Lesion in Computed Tomographic Images Using Enhancement Pattern and Texture Analysis

Abstract

The main objective of this study was to characterize liver lesion in computed tomographic images using enhancement pattern and texture analysis. A sample size of 100 patients was used, their ages were from (24 – 88) years old, both genderwere included. All were diagnosed to have liver lesions. This study was conducted in Radiology Department in Wad Medani Hospital and ELsoni Clinic Center- Wad Medani-Sudan, during the period from January 2015 up to January 2018.The result of this study showed that The mean attenuation measurement in Hounsfield units (HU) of cysts in the hepatic arterial phase was 35 HU±69.98(range, -17–272 HU), and it increased in the portal venous phase to a mean of 39.92HU±72.40(range,-7–283 HU). Cysts and abscess showed HU of -705.00±134.35 in the arterial phase then showed an increasing at the venous phase to be -550.00±70.71 then reduced at the equilibrium phase to be-805.00HU±134.35.Haemangiomas was 44.42HU ±17.38(range ,18.50 to83.60HU) in the arterial phase and increased in the venous phase 48.19HU±25.73 then decreased in the equilibrium phase 47.27HU±22.78.On average, the Hepatocellular carcinoma (HCC) was 52.3 HU±25.90 (range, –18 to136 HU) less than the adjacent liver parenchyma in the arterial phase,57.31HU±26.44(range,-11.0to112.00HU) in the venous phase and 54.15HU±24.06(range -9.00 to102.00HU) at the equilibrium phase. Metastases also showed an increase in the attenuation pattern after contrast enhancement from 73.54HU±63.62(range,1.00 - 236.50HU) in arterial phase to be increased in the venous phase80.94HU±67.50(range ,2.00 -243HU) then it decreased to be 75.31 HU±57.30( range,7.00-229.0HU) at the equilibrium phase .The study showed that there is a significant differences of the HU between all the detected liver lesions in all scanning phase at p= 0.000.The presence of liver enhancing lesions were in association with liver and spleen texture changes. Ascites was found in most of the cases i.e. (47%) of the cases. The result of textural analysis showed that the overall classification accuracy of abdomen parts was 95.1%, where the classification accuracy of cyst was 91.5 %, liver 95.3 %, the spleen 95.3%, while the spine showed a classification accuracy 100%.The model generated from these relationships can be used to automatically annotate new CT images appropriately. The study was dedicate that texture analysis is superior to visual perceptions system where texture reveal the change and difference of the image pattern objectively in respect to the ground truth, without the application of several phases in CT and hence save the patient from more radiation.