Development of Cyst Detection System for Liver Ultrasound Images

Abstract

Cystic liver diseases are relatively common and most identified incidentally on routine imaging. Ultrasound is used to detect liver cysts. Speckle artifact affects ultrasound images. Low-quality ultrasound images make diagnosing the cyst difficult. This paper proposes using ultrasound images and image processing to detect simple hepatic cysts. Current image segmentation techniques were evaluated to determine the best method for segmenting ultrasound images of simple hepatic cysts. This paper develops the system using MATLAB in which the GUI detects cysts. This graphical user interface (GUI) lets users select liver images from a database and check for cysts. The user can use the GUI segmentation method to detect the cyst's area in the pixel unit. The system was evaluated by measuring the area of detected liver cyst images and comparing them to the gold standard, manual segmentation. In boundary segmentation, the cyst was detected with 80% accuracy, in active contour with 100%, and in the watershed with 70%. This system can identify liver cyst ultrasound samples with 84% accuracy. Radiologists can use this system to analyse ultrasound images and detect cysts.