Brain Tumor Segmentation of Magnetic Resonance Imaging Based on Improved Support Vector Machines

Background: There are many methods to diagnose brain tumors, one, commonly used detection methods is imaging techniques. In the traditional method of examining, magnetic resonance imaging (MRI), the images are not only inefficient, but also depends on the expertise of the diagnostician. In this paper, an improved support vector machine (SVM) artificial intelligence algorithm is used to segment brain tumor MRI. Methods: The MRI images of 100 patients with brain tumors were collected before the experiment, and 100 cases of patients with brain images were divided into 4 groups of 25. The 4 sets of images were numbered: A1, A2,...,A25; B1, B2,..., B25; C1, C2,..., C25; D1, D2,..., D25. Images of Group A were segmented by an artificial method, images of group B were processed by general image segmentation algorithm, images of group C were treated with traditional SVM, and images of group D were processed with improved SVM algorithm. Results: According to the two indices of processing accuracy and efficiency of MRI segmentation of brain tumor, the experimental results found that: Processing accuracy of the artificial method is the highest of the 4 methods, close to 100%, but the efficiency of this method is the lowest. Processing accuracy and efficiency of the general image processing methods and the traditional SVM are not prominent. Only the improved SVM method is superior to the two indices in the brain tumor image segmentation processing accuracies and efficiencies. Conclusion: The experimental results show that the MRI image segmentation method, based on the improved SVM, is superior to the traditional methods.