Human Brain Tumor Detection and Classification by Medical Image Processing

The Human Brain Tumor recognition and elimination be the medical issues that residue problem in the field of medical. Before imaging methods are pneumo-encephalography and cerebral angiography has a disadvantage of keep invasive and so the CT scan and MRI scan acquisition methods useful for surgeons in giving a improved vision. This paper brain tumor recognition involves three stages namely image pre processing, image segmentation and image morphological function. Acquisition of brain tumor input image, it pre processed by changing the input image to gray scale then apply the high pass filter for image noise reduction and median filter for image quality improvement [1]. Proposed concept analyzed with transform of wavelet to acquire the image features of tumor, and then apply PCA method to minimize a dimensions of acquired image features. A minimized feature was analyzed with kernel support vector machine (KSVM). The goal of K-fold cross verification is to be use to improve parameters of KSVM. In this paper, choosen seven general brain diseases like Alzheimer's disease, Visual agnosia, glioma, Huntington's probelm meningioma, sarcoma, and Pick's disease as abnormal human brains, and acquired 240 MRI scan huamnbrain images (60 good image and 180 problem image) [2,3]. Our proposed methods for three dissimilar kernels, and obtain the GTB kernel got the maximum classification accuracy as 98.17%. The HPOL and LIN kernel achieves 94.7%, and 96.22%. Our proposed method compared with literature and results displayed our PCA+DWT+KSVM along with GTB kernel given the good accurate classification and feature extraction output. The normal image analyzing time for a 512x512 size input image on an desktop computer with 2.72 GHz & 4GB RAM is 0.0511.