Certain investigations on focal cortical dysplasia detection in brain mri images using recurrent neural networks

Abstract

Focal Cortical Dysplasia (FCD) was originally introduced by Taylor and has been defined as a heterogeneous cortical deformation that can be either congenital or induced. It is one of the major reasons for epilepsy in adults and children that is associated with cortical neoplasm and hippocampal sclerosis. FCD is a complex disorder that is often difficult to predict in Magnetic resonance imaging. The occurrence of epileptic seizures varies based on the site of FCD inflammation. In general, the treatment procedure for the FCD is higher for the younger population. The type II focal cortical dysplasia is most prevalent in very young children and must be treated surgically. But the Type 2 FCD is less extensive and mostly found in the temporal region, so the defect is more widespread. Therefore, pre-surgical and post-surgical imaging results play a vital role in the treatment of cortical dysplasia. In the proposed work, we have designed and developed a machine learning model that automatically evaluates and classifies the focal cortical lesion in pre-surgical FCD lesions based on MRI images. The entire process of the work is divided into two phases. In phase 1 the MRI images were acquired from the relevant source and the raw dataset is processed in the 3D wavelet transform to denoise the distorted noise and artifacts acquired during the image acquisition process. Similarly, phase 2 focus on the feature extraction and classification of MRI image based on the intensity of FCD lesions. The Weighted 3D DWT (3 Dimensional - Discrete Wavelet Transform) technique developed in phase 1 of the proposed work discuss the denoising process of the input MRI image. The 3D DWT provides sufficient contradiction to both the spatial and frequency domains newline