Analysis of Brain Tumor using Novel Image Processing Approach

Magnetic Resonance Imaging (MRI) is known to be one of the commonly used non-invasive imaging methods that primarily helps in the diagnosis of brain tumors. The early detection as well as the cure of brain tumors, is considered to be a pivotal factor for the degradation of death rate. Radiologists have to sort MR brain images into normal as well as tumor images manually, which takes a long time. Henceforth, this project brings forth the novel image processing based approach for the diagnosis of brain tumor. Initially, the image undergoes the pre-processing function using adaptive noise fuzzy switching median filter. It removes the black and white noise in the input image while retaining the fine details as well as textures, and generates a noise-free image. The filtered image is segmented using the OTS U method which generates an appropriate threshold value for data segmentation. In OTS U method, the image after filtration is digitalized. Subsequently, the process of extracting the features of the digital image is carried out by Gray Level Run Length Matrix (GLRLM) which extracts the texture features from images. Finally, Probabilistic Neural Network (PNN) architecture which is a feed forward neural network pre-trained with the broad database of images of Image Net, is introduced to perform classification for the detection of tumor in brain images. The obtained classification outputs indicate the detection of brain tumor with improved efficiency and accuracy.