The Research And Implementation Of Algorithm Of Medical Image Segmentation

Medical image segmentation is one of the key aspects of the medical image processing. It is an interdisciplinary technology that covers digital image processing, pattern recognition, computer vision and biomedical. Medical image segmentation can replace the cumbersome manual outlining and reduce the computational complexity of the subsequent processing.Current research in medical image processing tends following directions: further reducing the dependence on artificial and developing more intelligent algorithms; expanding the scope of the traditional segmentation algorithms; developing appropriate algorithm for images obtained by advanced device; designing particular segmentation strategy for particular needs; reducing the computational complexity and improving the efficient.In this paper, we discuss some of directions above and finish some research, including:1. Based on the classic Otsu method, we proposed an interleaved Otsu segmentation algorithm. In the iterative process of the algorithm, Otsu segmentation and bias field correction are conducted by turns, and the segmentation results and bias correction image are achieved simultaneously. Compare to the original Otsu algorithm, the proposed method is more robust to intensity inhomogeneous.2. Benefit to the idea of information fusion, we extend the single-channel segmentation to multi-channel. The multi-channel segmentation algorithm aims to the images obtained by MRI under very strong magnetic field, which contains serious bias field. Several MR images of different modalities are segmented simultaneously, promoting each other, reaching a better segmentation results.3. A novel ventricle segmentation algorithm is designed for multi-channel lesion segmentation. The algorithm first adjusts parameters to extend the area of white matter in the segmentation result, and then make a mask trough the connectivity, at last recover the ventricle based on the mask and normal segmentation.