Development of an automated liver-outline-extraction method for EOB-enhanced MRI images

Abstract

The irregular form of surface of liver is one of the important signs in diagnosing fibrosis on images. A pathologist performs the definitive confirmed diagnosis of fibrosis by evaluating F-Grade with in the hepatic tissue of a liver biopsy. In this study, we developed a technique of automatically extracting a liver outline on EOB-enhanced hepatocyte-phase MRI images. After enhancing the liver edges using the unsharp-masking filter, the proposed method extracts the liver region automatically using the p-tile method. Next, the liver outline is automatically delineated by outline-determination processing. The F-Grade classification can be performed by the mean SD of the differences in positions between an outline and its polynomial fitting curve. The proposed technique was applied to 64 cases (F0: 9, F1: 15, F2: 12, F3: 11, and F4: 17) . The average error of automatic outlines to the manual outlines extracted by a radiologist was 0.78 mm (F0: 0.70 mm, F1: 0.77 mm, F2: 0.78 mm, F3: 0.71 mm, and F4: 0.86 mm) . The mean SD was significantly higher in the F3 group than in the F2 group (p<0.05, Tukey test) . In conclusion, the proposed technique may be useful for automatic classification of F-Grade on EOB enhanced MRI images.