Hybrid Automated Brain Tumor Detection by Using FKM, KFCM Algorithm with Skull Stripping

Brain tumor detection from MRI images is a time consuming and precarious task due to irregular characteristics of tumor tissue image segmentation. In MR images permit convincing evidence and play a decisive part in diagnosing the different kinds of tumors. The segmentation recognition and extraction of tumor area from (MRI) magnetic resonance image are an initial interest. The clinical or radiologist specialists performed a time-consuming and tedious task but their precision relies on their experience. Therefore, the usage of computer-aided expertise becomes mandatory to overcome that limitation. A sophisticated fully automated tumor recognition system is proposed to have the maximum accuracy, specificity and sensitivity with a minimum error rate, computational time and competently extract tumor from MRI images. The current study emphases on tumor and edema segmentation that is built on kernel-based fuzzy C-means and skull stripping method. The clustering method amended by merging multiple kernels established on spatial information. Furthermore, once the acquired image is de-noised the automated brain tumor recognition algorithm stripped the outer boundaries of the irrelevant tissue and then the segmentation algorithm is applied to extract the tumor area precisely. For analysis and recording of the experimental result, hundred MRI images are used. The algorithm in the current study is compared and after the experimental result, the algorithms certify having the detection of brain tumor with accuracy i.e. 98.7%, specificity 90.0%, sensitivity 92.8% with minimum error rate 0.002% given by the improved algorithm KFCM while the minimum computation time i.e. 1.64 seconds achieved by Fuzzy K-means (FKM).