Investigations on brain magnetic resonance images for diagnosis of focal cortical dysplasia using spatial incorporated fuzzy c means segmentation

Abstract

Neuro imaging is a vast area of research which provokes on the diagnosis and treatment of neurological disorders. These neurological disorders are variety in number in which epilepsy is one of the most common diseases. Epilepsy is a one of the most persistent neurological disorder which is characterized by seizures. Though there are several reasons for epilepsy, the most known cause is Focal Cortical Dysplasia (FCD). Focal Cortical Dysplasia is a kind of cortical malformation which remains undetected through normal inspection or requires an expertise for detection. Magnetic Resonance Imaging (MRI) plays an important role for detecting and analyzing dysplastic lesions for treatment and pre-surgical evaluations of FCD patients. The FCD lesions are too subtle to be diagnosed by normal segmentation procedures. There are certain segmentation techniques which could aid the detection of lesional FCD. But, there are cases where the segmentation algorithms fail to detect MRI negative or non-lesional FCD. Hence, a technological assistance for the early and fast verdict of presence of epilepsy has evoked to work on the prognosis of epilepsy. MRI is mostly associated with artifacts which introduce noise in the image produced. For precise diagnosis of diseases the noise in the image needs to be removed without any loss of anatomical features. Thus, image denoising plays a prime role as a preprocessing stage in the field of image processing. Image denoising of Magnetic Resonance (MR) images by manifold approach focuses on the removal of heterogeneous Gaussian and Rician noise without blurring the fine anatomical structures. Among the various denoising techniques found in literature, the Non Local Means (NLM) filter is a popular denoising technique with superior level of image smoothing. But, it is found that NLM does not hold well for edge preservation of images with low Signal to Noise Ratio (SNR). newline