Correction of vibration artifacts in DTI using phase-encoding reversal (COVIPER).
Bottom Line: We refined the model of vibration-induced echo shifts, showing that asymmetric k-space coverage in widely used Partial Fourier acquisitions results in locally differing signal loss in images acquired with reversed phase encoding direction (blip-up/blip-down).COVIPER was validated against low vibration reference data, resulting in an error reduction of about 72% in fractional anisotropy maps.COVIPER can be combined with other corrections based on phase encoding reversal, providing a comprehensive correction for eddy currents, susceptibility-related distortions and vibration artifact reduction.
Affiliation: Wellcome Trust Centre for Neuroimaging, UCL Institute of Neurology, University College London, United Kingdom. firstname.lastname@example.orgShow MeSH
Related in: MedlinePlus
Mentions: Figure 3 shows the FA and the root-mean-square of the error of the tensor fit, and Fig 4 shows the quantified bias in FA. For dataset DTI±1 the vibration-induced bias in FA was visible in at least one dataset (blip-up or blip-down) of each subject (Fig 3a,b, arrows), while the extent of the bias varied between individuals ( = 0.38; = 0.29; = 0.39; Fig 4). The artifact manifested itself in different regions for the blip-up (Fig 3a, yellow arrows) relative to the blip-down data (Fig 3b, red arrows). Averaged over subjects the standard arithmetic mean combination of blip-up and blip-down data reduced the vibration-induced bias in FA by 49% (from = 0.35 to = 0.18, Fig 4). In contrast, the proposed COVIPER correction based on weighted-sum combination of blip-up and blip-down data reduced the error in FA by 72% = 0.1) and the resulting maps showed better correspondence to the reference FA maps (Fig 3d,e). The contribution of miscellaneous effects that could not be attributed to vibration effects in the COVIPER method was about 6% (Fig 4). Note that the arithmetic-mean combination affects the tensor estimate for all regions showing vibration artifacts in the original blip-up or blip-down data, i.e., the union of affected regions in the blip-up and blip-down data. As a result the arithmetic-mean combination may show more widely spread artifacts (see regions highlighted by yellow and red arrows in Fig 3a–c), than any of the two original datasets (although usually at a lower level).
Affiliation: Wellcome Trust Centre for Neuroimaging, UCL Institute of Neurology, University College London, United Kingdom. email@example.com