Development of an image processing Pipeline for the extraction of white Matter from mri

Abstract

The loss of White Matter (WM) volume quantified from the newlineMagnetic Resonance (MR) images is an important biomarker for diagnosing a newlinewide vairety of neurological diseases. Precise segmentation of the WM region newlinefrom the MRI slices is crucial in quantifying its volumetric loss. Noise and newlinelow-tissue contrast are two major challenges because of which extraction of newlinestructures like the WM from the MRI slices become complex. Traditional newlineimage denoising filters blur edges and an additional sharpening step is often newlinerequired to restore the strength of the edges that got faded during the newlinedenoising. Using separate denoising and sharpening filters cause newlinecomputational overhead. To solve this issue, a computationally fast algorithm newlinebased on Locally-Adaptive Steering Kernel Regression (LASKR), which can newlinejointly perform denoising and sharpening operations on MR imagery is newlineintroduced. The majority of the state-of-the-art image contrast enhancement newlinealgorithms do not consider the pixel statistics and suffer from one or more of newlineprocessing-induced distortions like contrast overshoot, noise amplification, newlinemean brightness shift, blockiness , and dynamic range compression. newlineTo address these limitations, a Feature-preserving Contrast Enhancement newlineTransform (FCET) in which the intensity transformation is performed based newlineon the statistical distribution of pixels, exclusively free from processinginduced newlinedistortions, for improving the contrast of MRI is instigated. newlineTo facilitate newline