Image analysis – Applications – Biomedical applications
Reexamination Certificate
1998-08-10
2001-06-19
Johns, Andrew W. (Department: 2621)
Image analysis
Applications
Biomedical applications
Reexamination Certificate
active
06249595
ABSTRACT:
FIELD OF THE INVENTION
This invention is related to MRI image reconstruction and, in particular to an iterative method of reconstruction for EPI imaging.
BACKGROUND OF THE INVENTION
EPI (Echo Planar Imaging) is a fast MRI imaging sequence which enables the acquisition of an image slice in 40 milliseconds or less. In most MRI imaging techniques, including EPI, the process of image acquisition includes the following steps:
(a) placing a patient in a strong, static and homogeneous magnetic field;
(b) applying a sequence of gradients and RF pulses to the patient;
(c) detecting RF radiation emitted by portions of the patient in response to the sequence of (b);
(d) placing the detected data into an array; and
(e) applying a two-dimensional FFT to the data in the array to generate an image.
MRI imaging sequences are usually differentiated by the sequence of (b), which affects the way in which data is sampled and placed in the k-space (d).
FIG. 1
illustrates the transversal of a k-space
20
by an exemplary EPI sequence. One of the important attributes of EPI sequences is that odd lines and even lines in the k-space are sampled in opposite directions, by using gradients of opposite polarities. Thus, every other line sample needs to be reversed, before being placed in the k-space matrix. Unfortunately, EPI sequences use very fast gradients which generate eddy currents. Another source of error in EPI images is their sensitivity to main field inhomogenities. Some types of MRI sequences are less sensitive to eddy currents and the like, since all the data lines are sampled in the same direction. As a result, all the data lines are shifted and distorted in a substantially similar manner. In EPI, due to the fact that every other line is reversed, every other line is distorted in a different manner from its neighboring lines. A main type of artifact in EPI images is ghosting, which is caused by the periodic nature of this distortion. The FFT is especially sensitive to periodic distortions in the data. The different effective shifting of odd and even data lines results in the reconstructed image containing both a true image and a ghost of the true image, shifted by half of the field of view.
This ghost forms due to a disruption of the normal cancellation process for image components arising from the odd and even lines of k-space (odd and even components). Phase mismatch between odd and even components leads to interference rather than cancellation. This is evidenced in the typical “Banded” appearance of both ghost and image in EPI pictures obtained with poorly tuned MRI systems. Correcting the phase mismatch in the image domain is mathematically equivalent to correcting certain errors in the k-space data and will correct ghosts which are caused by these certain errors. It should be noted that the errors in sampling k-space arise from many different types of instrument imperfections, most of which are too difficult or too expensive to solve.
It should be appreciated that there are at least two types of phase shift between odd and even components of the image, a constant phase shift that is independent of pixel location (zero order phase shift) and linearly varying phase shifts in the readout and phase encode directions of the image (first order phase shifts). Higher order phase shifts may also be present depending on the severity of the sampling imperfections.
The zero order phase shift is unaffected by the Fourier transform. Thus, correcting the zero order phase shift is exactly equivalent to correcting its corresponding phase error in k-space.
The first order phase shift, particularly that present in the readout image direction arises from the offset between odd and even lines in k-space, due to errors in timing and gradient performance. The FFT translates such an offset into a first order phase shift in the transformed data. Unfortunately, the FFT is very sensitive to even small discrepancies between the odd lines and the even lines.
Many types of phase shift correction techniques are known in the art. A first technique is a reference scan, in which an object, in some cases a phantom and in some cases the object to be imaged, is imaged to determine the distortions of the MRI machine. A correction for these distortion is determined (usually as a template) and then applied to subsequent images.
However, a reference scan is of limited utility for moving structures. In addition, any errors, such as noise, which are acquired with the reference data, are used to “correct” the image data.
In some types of reference scans, the imaging sequence includes acquiring data for use as a reference scan. In
FIG. 1
, for example, the central echo is acquired twice. Since this echo has no phase encoding, it is more suitable for some types of reference scanning.
In some types of reference scan correction techniques, the reference scan is analyzed to determine numerical values for phase correction, usually only for the lower orders of phase correction. However, applying such a correction indiscriminately may add artifacts to an otherwise acceptable image, especially if these numerical parameters vary substantially over the entire image and/or they are not determined accurately enough.
A general limitation of all reference scan techniques is that there is no method of improving the image beyond what is suggested by the reference scan. Thus, the image quality is always limited by the quality of acquisition of the reference data.
Other artifact reduction techniques modify the EPI data acquisition process. One technique uses the data collected from cycles of the oscillating gradient that have the same gradient polarity. However, this results in a reduced data accumulation rate. Another technique modifies the phase encoding process so that data collected from paired cycles of the oscillating gradient have equal encoding covering one half of k-space. One set of data is conjugated and placed into the missing region of k-space during reconstruction. Unfortunately, this practice significantly increases geometric distortion in the final image.
In some techniques, an operator is invited to manually correct the phase. However, this type of manual correction is time consuming and complicated for the operator to perform.
U.S. Pat. No. 5,068,609 and “Image reconstruction for Echo Planar Imaging with Nonequidistant k-space Sampling”, by H. Bruder, H. Fisher, H.-E. Reinfelder and F. Schmitt, in
Magnetic resonance in Medicine,
Vol. 23, pp. 311-323 (1992), the disclosures of which are incorporated herein by reference, describe a method of ghost reduction, wherein a required phase shift between odd and even lines is determined by calculation. In addition, a user is apparently required to select a portion of an image which contains only ghost portions.
“Ghost Artifact Reduction for Echo Planar Imaging Using Image Phase Correction”, by Michael H. Buonocore and Lisheng Gao, in Magnetic resonance in Medicine Vol. 38, pp. 89-100 (1997), the disclosure of which is incorporated herein by reference, describes a method of ghost reduction whereby an operator separates an image and a ghost portion and then a phase correction is calculated from odd and even echo images. In addition, this paper, on page 99 thereof de-legitimizes determining a phase correction solely by what is necessary to entirely remove a ghost portion of the image.
“A New Phase Correction Method in NMR Imaging Based on Autocorrelation and Histogram Analysis”, by C. B. Ain and Z. H. Cho, in
IEEE Transactions on Medical Imaging
Vol. M1-6, No. 1, March 1987, pp. 32-36, the disclosure of which is incorporated herein by reference, describes a method of calculating a required phase correction from an image, using autocorrelation.
SUMMARY OF THE INVENTION
It is an object of some embodiments of the present invention to provide a method of correcting artifacts in EPI images, which method is not limited by a quality of acquisition of reference scan data.
It is an object of some embodiments of the present invention to provide a method of correcting artifacts in EPI images, with
Foxall David L.
Harvey Paul R.
Huang Jian
Colb Sanford T.
General Electric Company
Hoffman Wasson & Gitler, P.C.
Johns Andrew W.
Nakhjavan Shervin
LandOfFree
Iterative reconstruction for EPI does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Iterative reconstruction for EPI, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Iterative reconstruction for EPI will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2494977