Detection Of Small Liver Cancers In Enhanced Liver Scan Images

Medical image examination is an important part of modern medical technology. With the development of medicine and computer science, people begin to put their attentions to automatic analysis of medical images with computer, which include automatic detection of small liver cancers in CT images. Liver cancer's death rate is very high and cured small liver cancer's death rate is relatively low, so the detection of small liver cancers is very important. Accordingly, the study, which detects small liver cancers in CT images, is of great significance and can be widely used in many fields.In this thesis, the characters of small liver cancers and common methods of image segmentation are analyzed. Then a method of detection is proposed, which include three phases: liver region extraction, shadow detection and small liver cancer detection. In each phase a corresponding image processing method, which is based on organic dissection characters, is developed.In the phase of liver region extraction, a region growth method is proposed. This region growth method gets seeds form pre-process of threshold method and morphological method. The seed region get by this process is bigger and can efficiently decrease the growth area and time. The growth principles and ending conditions are proposed based on the liver's dissection characters. By these methods, a liver region is acquired. In order to get the whole liver region from a series CT images, the 2-demetion region growth method is developed into 3-demention.A sign distance function's initialization method is proposed by analyzing the evolvement of CV model. This method can efficiently accelerate the convergence speed of CV model and is used to detect the shadow in liver. By analyzing the "energy" of CV model, the CV model is connected with threshold method. Then a method using threshold and morphological median is development to detect the shadow and get a perfect result finally. The detection of small liver cancers needs the comparisons of different phases. So some characters of matching the same location in different phases are put forward. Then an algorithm based on the characters of small liver cancers is proposed to detect the small liver cancers.The experiments prove: liver region extraction can efficiently reduce the region and time of detection; shadow detection in liver region can detect the shadows of liver distinctly and can further reduce the region of detection; small liver cancer detection can compare the CT images with same position in different phases, therefore it can detect the small liver cancers with "fast in and fast out"' character.