Robust correction of spike noise: Application to diffusion tensor imaging

Abstract

Echo-planar imaging (EPI) -based diffusion tensor imaging (DTI)is particularly prone to spike noise. However, existing spike noise correction methods are impractical for corrupted DTI data because the methods correct the complex MRI signal, which is not usually stored on clinical MRI systems. The present work describes a novel Outlier Detection De-spiking technique (ODD) that consists of three steps: detection, localization, and correction. Using automated outlier detection schemes, ODD exploits the data redundancy available in DTI data sets that are acquired with a minimum of six different diffusion-weighted images (DWIs) with similar signal and noise properties. A mathematical formulation, describing the effects of spike noise on magnitude images, yields appropriate measures for an outlier detection scheme used for spike detection while a normalization-dependent outlier detection scheme is used for spike localization. ODD performs accurately on diverse DTI data sets corrupted by spike noise and can be used for automated control of DTI data quality. ODD can also be extended When a large transient, localized, erroneous change in k-space signal intensity occurs during MRI data collection, it is referred to as spike noise. Despite improvements to hardware, spike noise remains a relatively common, sporadic source of image artifact on many MRI systems. Due to the Fourier transform relation between MRI k-space and image space, spike noise creates a ripple artifact in corrupted MR images (1-3). The precise cause of spike noise is difficult to troubleshoot, because the noise results from numerous system/site imperfections that interact with the application of large, fast switching magnetic field gradients during imaging. Spike noise can appear as clustered or isolated events and can vary in severity from negligible to highly detrimental. In many cases, spike noise significantly obscures the information of interest in MR images, such that the images are discarded. It is particularly problematic for pulse sequences that require high gradient amplitude and slew rate, such as echo-planar imaging (EPI) -based diffusion tensor imaging (DTI) (4).At a given slice, DTI requires a set of at least six diffusion-weighted images (DWIs), sensitized to diffusion along different orientations, as well as a single T 2 -weighted image which is not diffusion sensitized (5,6). The voxel-wise signal magnitude of the DWIs is used to solve for six independent components, D xx , D xy , D xz , D yy , D yz , D zz of a symmetric 3 ϫ 3 tensor, characterizing three-dimensional (3D) diffusion. Typically, the tensor information is condensed into a spatially invariant scalar map such that a single image per slice represents the tensor data (6,7). In the brain, a commonly adopted scalar is the fractional anisotropy (FA: values from 0 to 1), which provides good contrast for white matter (high FA) and gray matter (low FA).Other diffusion tensor-derived scalars (7), as well as tractography (8), are receiving increasing attention. Analysis of the effects o...

Brain MR Diffusion Tensor Imaging for Multi-Site Studies: A Review

“…Inclusion of poor data will lead to negative findings or false findings resulting in Type 1 or 2 errors in group comparisons. FA and diffusivity from various manufacturers may vary substantially with minor differences in pulse sequence parameters [20]. For studies involving DWI and DTI, scanning parameters need to be standardized as closely as possible and scanners calibrated.…”

Diffusion MRI and its Role in Neuropsychology

“…For this reason good dMRI practice includes evaluating image quality during data acquisition and including automated QA (quality assurance) methods that evaluate dMRI images for spike noise. For data that contain spikes, the options depend on the severity of the problem, from excluding the subject data to rejecting whole volumes or using interpolation methods to replace corrupted slices using interpolation from similar volumes (Chavez et al 2009). …”

Impact of outliers on diffusion tensor and Q‐ball imaging: Clinical implications and correction strategies

“…Corrupted images are often found to occur within volumes of acquired data, particularly when using ultrafast sequences such as echo planar imaging (EPI) (1). Among these corruptions are those characterized by large, unpredictable signal variations that may originate from patient motion (2), physiological-related fluctuations (3,4), spiking artifact (5), and magnetic gradients causing patient table vibrations (6,7).…”

“…Mean PEV was then computed by averaging all of the y from the iterations (y i ), and the 'angular error' (Erry) was computed from the dot product of y i and y original as shown in Eq. [5]:…”

“…To identify corrupted images, a visual assessment of the DW-MRI acquisition was undertaken, as per standard practice in a clinical setting. Automated detection methods are also possible (5). Scans were considered corrupted where clear losses of signal were evident in one or more slices (14).…”