Pen-based Interactive Segmentation Of Abdominal Organs

CT and other medical imaging technologies which is the product of scientific and technological progress provide a reliable image data for the medical inspection, diagnosis, treatment. Because the analysis of acquired image data is vital to diagnosis, so how to segment CT accurately and precisely has become an urgent problem in medical image processing. At present, GVF Snake Model is used widely in many computer-aided systems for medical image segmentation.Initial outline of image segmentation should be as close as possible to the real contours because GVF Snake model is sensitive to the initial position. For this purpose, it is required that the user should input the initial contour manually. The traditional CT image segmentation system adopts WIMP interactions, which require the user to use the mouse and keyboard to interact with the system. WIMP interactions which are based on desktop metaphor process word conveniently and efficiently. However, experienced people know that using the mouse to tag and map is extremely inconvenient. In this case, using the mouse to initialize GVF Snake model is lack of validity, accuracy and efficiency. Moreover, when doctors check the CT images, they may record something. But with WIMP interactive mode, they have to switch the keyboard and the mouse frequently, which may reduce the user's work efficiency. Besides, commands are always in menu form with WIMP interactive mode, so more screen resources may be taken up, which can reduce the image display capacity.For the shortcomings of CT image segmentation software based on GVF Snake, this paper introduces pen-based interactive segmentation of abdominal organs through the research of the pen-based user interface. Pen and paper is a metaphor of Pen-based user interface, so this interaction can meet people who get used to pens and papers since childhood, and the initialization of the GVF Snake model is in the "pen and paper" work environment. Therefore, the use of pen interaction may support convenience for uses to initialize the model; pen input support for character recognition and medical text does not have too much, so he can write directly by hand, which can reduce the switching between the mouse and keyboard; pen interaction techniques support gesture recognition, which improve the naturalness of interaction. While reducing the occupancy of the screen command control, valuable screen resources can be saved.Finally, through a set of examples, the validity, naturalness which put forward in this paper will be confirmed.