Automated Breast Tumour Detection in Ultrasound Images Using Support Vector Machine and Ensemble Classification

Breast cancer is the second leading cause of death in women after heart diseases. A well-known statement in cancer society is “Early detection means better chances of survival”. In the past few years several techniques were developed to detect breast tumors in early stages. A proposed system is designed for breast tumors detection using ultrasound images. Ultrasound is used because it is less expensive and less invasive than X-rays used in mammography and computerized tomography. It can provide a second opinion for a physician to detect breast tumors. The proposed system consists of three main steps: preprocessing, feature extraction and classification. Gaussian blurring, anisotropic diffusion and histogram equalization are used to reduce additive noise, speckle noise and to enhance the image quality respectively. The second step is feature extraction and dimensionality reduction. PCA is used to reduce the dimensions of the feature vector. The third and final step is the classification step. A comparison is conducted between support vector machine and bagging ensemble classifier as different classification techniques. The third step is deployed to classify the images into image with/without tumors.