Segmentation and Rigid Registration of Liver Dynamic Computed Tomography Images for Diagnostic Assessment of Fatty Liver Disease
Authors
Kyoyeong Koo ; Jeongjin Lee ; Jiwon Hwang Hanwha Vision ; Taeyong Park ; Heeryeol Jeong ; Seungwoo Khang ; Jongmyoung Lee ; Hyuk Kwon ; Seungwon Na ; Sunyoung Lee ; Kyoung Won Kim ; Kyung Won Kim
Citation
Journal of Computing Science and Engineering, Vol.17(3) : 117-126, 2023-09
This study presents a method for diagnosing fatty liver disease by using time-difference liver computed tomography (CT) images of the same patient to perform segmentation and rigid registration on liver regions, excluding the vascular regions. The proposed method comprises three main steps. First, the liver region is segmented in the precontrast phase, and the liver and liver vessel regions are segmented in the portal phase. Second, rigid registration is performed between the liver regions to align the liver positions affected by the patient"s posture or breathing. Finally, fatty liver diagnosis is performed with the average Hounsfield unit (HU) value calculated using only the area removed from the vessel area segmented in the portal phase after registration in the precontrast liver area. The mean distance error between the points corresponding to the liver boundary was 3.136 mm and the mean error between the anatomic landmarks was 4.166 mm. A fatty liver diagnosis was confirmed in a total of 18 cases, and the results were identical to the histology results. This technique may be valuable in clinically diagnosing fatty liver using liver CT imaging, which is widely available and more commonly used than abdominal magnetic resonance.