IES20170145A2 - Magnetic resonance imaging - Google Patents

Abstract

In a magnetic resonance imaging involving a Propeller-type acquisition sequence, a reconstruction based on an inverse problem approach is used, T2-blurring effect is included as an additive structural error term, and in the inverse problem reconstruction k-space profiles less subject to the error are preferentially weighted.

Description

The invention relates to a method and device for magnetic resonance imaging, and in particular to magnetic resonance imaging involving a Propeller-type acquisition sequence.

BACKGROUND OF THE INVENTION Magnetic Resonance Imaging (MRI) is performed by applying and subsequently measuring time varying magnetic fields. A subject of interest (for example a patient undergoing examination), is placed in a static homogeneous magnetic field. After the application of a radio frequency magnetic field, and time varying gradients of magnetic field, a radio frequency signal is measured.

This signal is subsequently stored with a set of other measured signals and processed to create an image. Conventional Cartesian acquisition requires the acquisition of signals laying on a Cartesian grid. Non Cartesian acquisition techniques, on the contrary, acquires signals laying on non-uniformly spaced and generally non parallel, trajectories. They have been developed to effectively acquire the information needed to reconstruct the image. In some cases (for example radial and PROPELLER trajectories), data are acquired with a variable sampling density, such that in some regions of the k-space the acquisition is redundant. The acquisition redundancy has been exploited in some reconstruction techniques to estimate motion states and correct for the artefact that they can cause, although residual motion artefacts or other structured artefacts may still be present.

One important example of structured artefact that hinder some applications of these techniques is the T2 blurring, that reduces the PD or Tl contrast in the images.

Residual errors may be present in the data representing the acquired portion of the k-space.

For example after rigid motion correction in PROPELLER MRI, errors due to non-rigid motion or through plane motion may still remain in the data. Structured errors are also caused by T2 blurring, hindering the application of some techniques on Tl or PD weighted imaging.

A common approach in this case consists in discarding or filtering the data according to some measure of its reliability. The weighting procedure is typically performed taking also into consideration the sampling density in the k-space, as usually done in non-Cartesian MRI reconstructions (sampling density compensation).

Techniques belonging to this class of methods are KWIC (US7439737B2), developed for radial MRI, or the “intra-blade bow-tie filter” for contrast weighting developed for PROPELLER MRI (US20140077813A1).

The energy distribution in the k-space of the above mentioned errors may be non-uniformly distributed. Filters used to suppress this error should take also this energy distribution into consideration. Erroneous assumptions on the energy distributions of such structured errors can cause the discarding of region not affected by artefacts. This in turn may lead to SNR losses in the final image.

In PROPELLER MRI approaches based on filters require the knowledge of all the profiles belonging to the blades. For this reason they are applied after the application of Parallel Imaging techniques used to synthesize missing profiles (if there are missing profiles). The application of these filters before this process may be non-trivial.

MRI image reconstruction using inverse problems has gained interest during the years, bringing to the development of advanced techniques (like Parallel Imaging and Compressed Sensing). In inverse problems, the image reconstruction is performed enforcing data consistency only on the acquired data in the k-space. Regularization terms are used to stabilize and improve the image quality. In this scenario the application of artefact correction filters is not possible, since the filter need to be applied to the whole blade, of which only few profiles are known. This hinders the application of methods, like contrast weighting, to MRI reconstruction based on the solution of inverse problems.

SUMMARY OF THE INVENTION It is, inter alia, an object of the invention to provide an improved magnetic resonance imaging. The invention is defined by the independent claims. Advantageous embodiments are defined in the dependent claims.

One aspect of the invention provides a magnetic resonance imaging involving a Propeller-type acquisition sequence, wherein a reconstruction based on an inverse problem approach is used, T2-blurring effect is included as an additive structural error term, and in the inverse problem reconstruction k-space profiles less subject to the error are preferentially weighted.

Embodiments of the invention thus provides a relaxation error suppression for redundant MRI acquisitions. The effect of T2 relaxation in PROPELLER MRI is suppressed by giving a preferential weight to data not corrupted by relaxation. This weighting procedure is realized by solving a weighted least squares problem.

These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter.

DESCRIPTION OF EMBODIMENTS A reconstruction based on an inverse problem approach is used. The T2blurring effect is included as an additive structural error term. In the inverse problem reconstruction k-space profiles less subject to the error are preferentially weighted. IN this way T2-blurring is mitigated in the reconstructed magnetic resonance image.

The effect of T2 relaxation in PROPELLER MRI is suppressed by giving a preferential weight to data not corrupted by relaxation. This weighting procedure is realized by solving a weighted least squares problem.

Image reconstruction based on inverse problems describes the MRI acquisition as a linear operation: y = Ex + η Here y represents the data acquired in the k-space, E is the non uniform Fourier operator and η is some additive acquisition noise. The inverse problem aims at estimating the image x starting from the data y: x = argminf(E,x,y) + X This is a general formulation, where f(E, x, y) is a data consistency term, (x) and are some regularization tenns and their corresponding weights. The regularization terms may also be not present.

Structured errors may be formalized as an additive tenn z in the signal equation: y = Ex + z + η This is a general formulation. Errors that are multiplicative by nature can also be reformulated in additive form. For example in the case of error due to T2 relaxation: At xe T2 + η = Ex 4- z + η In general the structured error is a function of the k-space location k and the time t: z = g(k, t) If z is known, or a model for z is provided, it is possible to exploit the redundancy of nonCartesian acquisitions, like PROPELLER, to give a preferential weight to the profiles that are less subject to the error. The data consistency term in this case can be: f(E,x,y) = IlEx-yllw In this case the problem is formulated as a weighted least squares (plus regularizations if present). The weighting matrix W = h(z) is such that a lower weight is applied, in the data consistency term, to some subset of data where \z\ is large.

An example could be: In this case SNR losses are avoided by properly modelling |z|, and the redundancy is exploited to select only the profiles less corrupted by errors. In the regions where there is low redundancy (i.e. the periphery of the k-space) all the profiles are automatically taken into consideration.

Given a model for the weighting matrix W it is possible to estimate the image as: x = argmin||Fx — y\\w = (EHWE) 1EHWy X The matrix E represents the non uniform Fourier transfonn. Since W is a positive diagonal matrix, the matrix P = EHWE is a Toeplitz matrix. Therefore P_1 is easily calculable after Fourier transform.

The procedure thus comprises: 1. Measure (or estimate) the blade k-space data y: a. Measured if parallel imaging is used b. Reconstructed applying parallel imaging 2. Correct for motion and trajectory inconsistencies according to PROPELLER motion correction algorithm 3. Estimate the weighting data W: a. Weighting for the T2 blurring b. Weighting based on correlation for motion suppression (PROPELLER correlation filter) 4. Apply inverse non uniform Fourier operator to the weighted data yw = Wy . Filter the intermediate image xw = EHyw applying the inverse filter P_1 The reconstruction process flow is for example acquired data => PROPELLER motion correction => Weighting matrix => inverse NUFFT => FilterP'1 => Output image, or acquired data => SENSE unfolding per blade => PROPELLER motion correction => Weighting matrix => inverse NUFFT => Filter P'1 => Output image It should be noted that the above-mentioned embodiments illustrate rather than limit the invention, and that those skilled in the art will be able to design many alternative embodiments without departing from the scope of the appended claims. The word comprising does not exclude the presence of elements or steps other than those listed in a claim. The word a or an preceding an element does not exclude the presence of a plurality of such elements. The invention may be implemented by means of hardware comprising several distinct elements, and/or by means of a suitably programmed processor.

Claims (2)

1. A method of magnetic resonance imaging involving a Propeller-type acquisition sequence, the method comprising: using a reconstruction based on an inverse problem approach, including a T2-blurring effect as an additive structural error term, and 5 preferentially weighing k-space profiles less subject to the error in the inverse problem reconstruction.

2. A device for magnetic resonance imaging involving a Propeller-type acquisition sequence, the device comprising a processor for 10 using a reconstruction based on an inverse problem approach, including a T2-blurring effect as an additive structural error term, and preferentially weighing k-space profiles less subject to the error in the inverse problem reconstruction.