Study Of Skull-brain CT Image Segmentation Technique

One of the most important constituent parts of modern medicine is medical imageSegmentation technique. As an indispensable link of medical image segmentation, the resultof skull-brain segmentation helps to reconstruct 3-D, to provide reliable evidence for plasticoperation, trauma and repair, and to satisfy the requirements of pathology research andclinical diagnosis.Through the analysis of the background, status quo of medical image segmentation and thecharacteristics of algorithmic research, combining with the practical application of skull-brainsegmentation, this paper dissects three methods commonly used in bones segmentation:Threshold Method, Region Growing Method and Active Contour Model Method. The paper,furthermore, raises a modified scheme to improve the adaptability of CT image segmentationand weaken the limitations of above methods.This paper mostly managers to improve two optimum threshold auto-selected algorithms:Otsu Threshold Method and Adaptive Iterative Threshold Method, Which basically solve thefamiliar problem that threshold segmentation will be invalidated when the target proportion iscomparatively small. The modified Region Growing algorithms maximally reduced regiongrowth's dependence on the characteristics ofIt seeds and best strengthen the segmentationeffect. It also strengthen the smoothing effect and are applied into practical skull-brainsegmentation. In the research on skull-brain segmentation technique, this paper focuses on thealgorithm of Active Contour Modeling Method whose insufficiency has been modified afterseriously studying the original model theory. Moreover, combining the segmentation resultsof Region Growing Method with of Active Contour Model Method, the paper raises an activecontour model skull-brain CT segmentation algorithm which combines with region growth.This newborn algorithm largely improves the performance of skull-brain segmentation.This paper elaborates on the theoretical basis of algorithms and segmentation theory, andvalidates different algorithms via correlative programs designed by Visual C++ language andMatlab language. Experiments show that, the modified algorithms raised by author in thispaper are all optimized in performance. Compared with different segmentation results ofmodified Threshold Method, Region Growing Method and Active Contour Model Method, ithas been known that, the best effect is procured via the Active Contour Modeling Methodcombining with region growth. These methods have certain significance on practicalapplication to satisfy the requirements of skull-brain segmentation.