Automated Identification of Brain Tumor using Image Transformation Methods

Abstract

In medical image processing, automated detection and classification of Brain Tumor Images (BTI) is very important. Tumors are nothing but the abnormal cells which grow in the brain and directly affect all the healthy cells. In young generation the effect of brain tumor is rapidly increasing. Manual detection and classification of brain tumors can cause human errors. Automated detection and classification of tumor is required as it reduces the burden of human observer and the accuracy also will not be affected due to use of large number of images. Accurate detection and classification of the tumors is required for diagnosis and subsequent treatment planning .Generally, electronic equipment is used in brain tumor diagnosis. The efficient and most popular technique used for diagnosing the brain tumor is Magnetic Resonance Imaging (MRI).This paper uses an image transformation technique named Discrete Cosine Transform (DCT) to obtain the test data results as normal or abnormal images by using the trained dataset images and calculate the percentage of accuracy, sensitivity and specificity using the confusion matrix attributes. Then the obtained abnormal images are further classified by a novel method of segmentation. In this process Otsu’s Binirization technique is used to obtain the binary transformation of an image and clustering algorithm named k-means clustering is used to segment the required area of an image. A Discrete Cosine Transform (DCT) is employed for obtaining the features of the image and these extracted features are given to kernel SVM and the Cross validation method is used for enhancement and SVM generalization to classify the Benign and Malignant tumors. These methods are helpful for early detection and also assist doctors in identifying the severity of tumor.