Brain tumor detection and classification using SIFT in MRI images
Mohammed Sahib Mahdi Altaei, Sura Yarub Kamil · AIP conference proceedings · 2020
MRI images are the most important tool for early detection of brain tumor. Tumor and cancer are a harmful and death-defying disease for human life. In this paper a proposed system deals with medical MRI for classifying input digital image into normal or abnormal tumors, also the type of abnormal case that refers to the existence of brain tumors is also diagnosed into benign tumor or malignant tumor. The proposed brain tumor classification system is based on using SIFT descriptor for extracting useful MRI features for diagnosis medical MRI images. The benefits of using SIFT is nevertheless of the image brightness or rotation of the MRI image, it also provides huge number of strong features that can be prepared well to be suitable for MRI classification. Two classification levels are adopted: the first is uses Naïve Bays classifier to detect the tumor and examine the considered case if it is normal or abnormal (i.e. tumor). Whereas, the second level uses J48 classifier to diagnosis the abnormal case detected in the first level into Malignant cancer or Benign. Frequent tests based on using cross validation technique are carried out, results of applying the first Naïve Bays classifier showed 98.9% accuracy for detecting the tumor, while the second J48 classifier showed an accuracy of 100%. This indicates that the developed tumor detection and classification is a robust due to the use of SIFT features when applied on a set of training images representing the typical tumor cases, which ensure the success of the classification system and correct path of the computations.