SVM-Based Bone Tumor Detection by Using the Texture Features of X-Ray Image

Bone tumor is a kind of harmful tumor which mostly occurs in adolescents. In this paper, we present a support vector machine (SVM) based bone tumor detector by using the texture feature of x-ray images. Due to the low incidence of bone tumor, it is hard to acquire dataset on a large scale, we use linear kernel function of SVM and cross validation to reach a more stable result. According to the characteristic of bone tumor x-ray images, we extract the texture features such as the angular second moment, correlation, entropy, homogeneity, contrast, dissimilarity from the x-ray images based on gray level co-occurrence matrix (GLCM). These features are used as input for the support vector machine classifier. And according to the scale of the dataset, a 5-fold cross validation test is performed in this paper. The highest accuracy of this detector can reach 99%.