Extracting tumor in MR brain and breast image with Kapur's entropy based Cuckoo Search Optimization and morphological reconstruction filters

Magnetic Resonance Imaging (MRI) scanners are used to determine the presence of tumors in human bodies. In clinical oncology, algorithms are heavily used to analyze and identify the tumor region in the slice images produced by the MRI scanners. This article presents an unique algorithm which is developed based on Kapur's Entropy-based Cuckoo Search Optimization and Morphological Reconstruction Filters. The former is used to locate and segment the boundary of tumors, while the later to remove unwanted pixels in the slice images. The proposed method yields 97% accuracy in the identification of the exact topographical location of tumor region. It requires less computational time (about 3 milliseconds, on average) for processing. Thus the proposed method can help radiologists quickly detect the exact topographical location of tumor regions even when there are severe intensity variations and poor boundaries. The method fares well in terms also of other standard comparison metrics like entropy, eccentricity, Jaccard Index, Hausdorff distance, MSE, PSNR, precision, recall and accuracy, when compared to the existing methods including Fuzzy C Means clustering and PSO. Above all, the algorithm developed can detect the tumor regions in the MR images of both brain and breast. The method is validated using various types of MR images (T1, T2 for MRI brain, and T1 post contrast and post processed images for breast) available in the online datasets of BRATS, RIDER and Harvard.