Research on SVM-based MRI image segmentation

Based on the gray symbiotic matrix and gray characteristic values, an extraction feature method is proposed for medical image of magnetic resonance imaging (MRI), which can reduce the dimensions of characteristic values, and improve the operation rate as well as classification accuracy. For a single image, by selecting its outline as characteristic values, support vector machine (SVM) is trained by using interval samples to realize the segmentation of other similar images. The problem of low computation efficiency due to large amount of samples is also avoided by reducing the number of samples. The result shows that, by taking the main feature of one picture as training sample, the identification of the same picture can reach 90% and the identification of other similar images can reach 80%. This method can reduce the lesion area and improve processing speed. Meanwhile, the image compress ratio can reach 1/4 when PSNR is 20.49% by utilizing compressed sensing technology on the background of the image based on this image segmentation method.