An Automated Framework to segment sthe tumor part of Multimodal images using Cuckoo search with Kernel Fuzzy C means Clustering

Abstract

In recent past many people in the world are highly affected by cancer-related diseases such as brain tumor, breast cancer, and skin cancer, etc. Most researchers focused on soft computing and machine learning techniques to analyze the tumor with the support of radiologist and classify the tumor part with minimum duration. An automated method is highly required for the radiologist to prevent the severity of cancer and helpful for early diagnosis, we proposed an automated approach for segmenting the tumor of MRI brain and MRI  breast images using cuckoo search optimization with kernel  fuzzy C means clustering. Contrast Limited Adaptive Histogram Equalization preprocessing is used to remove the noise information using and improve the contrast for better understanding, Combination of cuckoo search optimization with Kernel Fuzzy C-Means clustering is to segment the tumor part which also produces best optimal solution among all other optimization techniques.we collect MR brain and Breast images from publicly available online resources like  Harvard, BRATS,  Reference Image Database to Evaluate Therapy Response,Breast Imaging Reporting and Database System and a few from clinical datasets for validating and ensuring the efficiency of the proposed method. Performance measures like Mean Square Error, Peak Signal to Noise Ratio, Jaccard index. Dice coefficient, Precision, Recall and Accuracy and computation time are computed to ensure the tumor segmentation accuracy. Proposed approach yields an average of 48.3% PSNR value and an average of 8ms computational time with 98.40% segmentation accuracy which is far better than state of art techniques