Certain investigations on segmentation and compression of brain tumor image for implementation of medical decision support system

Abstract

Advanced Neuroimaging technology plays a great role in Medical diagnosis The most important need of today s medical field is the requirement of methods to diagnose the disease or pathology In order to analyze medical image data efficient methods are required Different imaging techniques like X-ray CT scan and MRI have different information pattern MRI scan is used for diagnosis of tumor since it does not use any radiation The automatic brain tissue methods has problems in classification of brain tissues The efficiency of the feature extraction process reduces due to artifacts Especially in clinical applications the accuracy level of segmentation process is low and diagnosis needs manual intervention The features are sufficient enough to do the diagnosis but for transmission the data size is to be reduced without information loss Since existing methods do not show a considerable PSNR and compression ratio efficient methods are required The objective of the work is to detect the impulse noises in the brain image and to remove them and to detect brain tumor using Adaptive Neuro Fuzzy Inference System ANFIS classifier The compression of data is to be done to accommodate efficient transmission Certain investigation of preprocessing methods were presented in this work and directional difference for denoising the impulse noises was chosen to remove the noises The method removes the noises in and around the edges of the brain image The quality of the image is improved by magnitude image construction using Gabor transform The performance comparisons of brain image denoising in terms of PSNR have been evaluated for different methods From the observation it was found that the proposed method shows an improvement in PSNR for about 49.9% over the Discrete Wavelet Transform method The medical image segmentation seems a complex and challenging process for MRI image with noises The accuracy level of the tumor segmentation process is less and diagnosis needs manual intervention newline