WO2011026923A1

WO2011026923A1 - Super-resolution magnetic resonance imaging

Info

Publication number: WO2011026923A1
Application number: PCT/EP2010/062918
Authority: WO
Inventors: Stefan Ropele; Gernot Reishofer
Original assignee: Medizinische Universität Graz
Priority date: 2009-09-03
Filing date: 2010-09-02
Publication date: 2011-03-10

Abstract

A method for obtaining a magnetic resonance image of an object with a given pulse sequence, wherein the method comprises applying a first modulation of the longitudinal magnetization for producing a periodic spatial modulation of the longitudinal magnetization with a wavelength of a pixel width, performing the pulse sequence with a given field of view to acquire a first image of the object, applying a second modulation of the longitudinal magnetization that is spatially shifted by a part of the wavelength, performing the pulse sequence with a field of view shifted by a part of a pixel width to acquire a second image of the object, merging the first image and the second image in the image domain in an interleaved fashion pixel by pixel, and deconvolving the merged image using the modulation function.

Description

Super-Resolution Magnetic Resonance Imaging

This invention generally relates to magnetic resonance imaging (MRI) and to an image acquisition and reconstruction method in order to increase the spatial image resolution.

More specifically, the invention relates to a method for obtaining a magnetic resonance image of an object with a given pulse sequence.

Further, the invention relates to a controller for a magnetic resonance imaging device.

Beyond this, the invention relates to a magnetic resonance imaging device.

Moreover, the invention relates to a program element.

Furthermore, the invention relates to a computer-readable medium.

Magnetic resonance imaging (MRI) is primarily a medical imaging technique most commonly used to visualize internal structures and functions of the body. In addition to its non-invasive nature, it provides much greater contrast between the different soft tissues of the body than other imaging methods do.

A high spatial resolution is required to accurately delineate the boundaries of different features and tissue types. For diagnostic purposes it is also necessary to depict small structures or lesions. Since MRI data is acquired in the Fourier domain (i.e. in the k-space or spatial frequency space, not directly in the image space), the maximum resolution is given by the maximum frequency (or wave number) measured k_max. According to the Nyquist's theorem, the spatial resolution Ax,y is a function of the maximum sampling time t_{X/y ma}x during the application of a magnetic gradient field G_x,y [T/m] at a given Lamor constant γ In view of equation ( 1 ) , to increase the spatial resolution at a given field of view (FOV), the gradient field G_x,y, the sampling time, or both, have to be increased . The sampling time, however, is often constrained by other parameters, particularly the echo time and the transverse relaxation time, so the achievable resolution is usually limited by the maximum gradient strength available or by the rate at which data has to be digitised. Additional limitations can arise in single shot pulse

sequences where all k-space data necessary to reconstruct a magnetic resonance image is acquired after just one single radio frequency (RF) excitation pulse. In such sequences the spatial resolution is significantly limited due to the fact that the echo signal from the transverse

magnetization decays rapidly during filling up the k-space with digitized data.

Many attempts have been made to improve the spatial resolution using only a single image. These post-processing techniques involve the estimation of frequencies beyond the maximum frequency measured, kmax, by using a priori knowledge. However, in most situations the amount of new information is relatively small and possibly not even measurable in the presence of noise.

Another approach to increase spatial resolution in MRI is multi-shot imaging, i.e. only a part of the k-space is sampled after each excitation. If a higher resolution is wished, a larger k-space has to be acquired which usually results in more excitations. The resolution of images acquired in this manner is limited by the imaging time available and by the ability of the imaged subject to keep still. However, some techniques like diffusion weighted imaging (DWI) are incompatible with such segmented k-space approaches because of the large phase variations resulting from even minimal physiological motion during the application of the diffusion weighting field gradients. Super-resolution algorithms have been proposed recently as another method to increase the resolution in M RI. In the context of this application, the term super-resolution may particularly refer to an image processing method that incorporates multiples images - related by known spatial transformations - into a single high resolution image.

Usually, the original images have a relatively low resolution and are linearly shifted in sub-pixel steps, but they can also be blurred, rotated, or scaled . Super-resolution techniques have been used, for example, for reconstructing a high-resolution image from several pictures taken by a moving satellite.

The first super-resolution M RI (SRM RI) approach was disclosed by Peled S and Yeshurun Y in "Superresolution in M RI : Application to Human White Matter Fiber Tract Visualization by Diffusion Tensor Imaging", Magnetic Resonance in Medicine, Vol . 45, pp. 29-35, 2001. This paper, however, was received with controversy. It was argued that, due to the properties of the Fourier transformation, no new information beyond the first image can be acquired when applying SRM RI in the frequency- encoding or phase-encoding directions as pointed out by Scheffler K in "Superresolution in M RI?", Magnetic Resonance in Medicine, Vol . 48, pp. 408, 2002. Attention then was turned to the application of SRM RI to the slice-encoding direction . Initial results were presented by Greenspan H and co-workers in "M RI inter-slice reconstruction using super-resolution", Magnetic Resonance Imaging, Vol . 20, pp. 437- 446, 2002, and seemed to be more promising than super-resolution in frequency- or phase- encoding direction . Recently, much effort was again put on SRM RI in the frequency-encoding direction . The new idea was to apply a phase ramp prior to the anti-aliasing filter for the acquisitions to yield independent information, and was proposed by Carmi E and co-workers in "Resolution enhancement in M RI", Magnetic Resonance Imaging, Vol . 24, pp. 133- 154, 2006. A similar concept was disclosed in US 5,054,489 (Synthetic Aperture MRI). However, in "Measuring information gain for frequency- encoded super-resolution MRI", Magnetic Resonance Imaging, Vol . 24, pp. 1058-1069, 2007, Mayer G - one of the co-inventors - concluded after a comprehensive mathematical analysis of this method, that only little progress may be possible by using this approach to perform resolution enhancement in the frequency-encoding direction.

Hence, conventional MRI systems may still suffer from a limited image resolution. It is an object of the invention to provide a magnetic resonance imaging system providing increased image resolution.

In order to achieve the object defined above, a method for obtaining a magnetic resonance image of an object, a controller for a magnetic resonance imaging device, a magnetic resonance imaging device, a program element, and a computer-readable medium according to the independent claims are provided.

According to an exemplary embodiment of the invention, a method for obtaining a magnetic resonance image of an object with a given pulse sequence is provided, wherein the method comprises applying a first modulation of the longitudinal magnetization for producing a periodic spatial modulation (which may also be denoted as a modulation function) of the longitudinal magnetization with a wavelength (according to the periodicity), performing said pulse sequence with a given field of view to acquire a first image of said object, applying a second modulation of the longitudinal magnetization that is spatially shifted along the modulation field by a part (or a fraction) of the wavelength, performing said pulse sequence with a field of view also shifted by the same part as the modulation function to acquire a second image of said object, merging said first image and said second image in the image domain in an interleaved fashion pixel by pixel (or in a multi-dimensional embodiment voxel by voxel), and deconvolving the merged image using the

modulation function.

According to another exemplary embodiment of the invention, a controller for (for instance controlling or controlling part of) a magnetic resonance imaging device for generating a magnetic resonance image of an object with a given pulse sequence is provided, the controller comprising a sequence control unit adapted for applying a first

modulation of longitudinal magnetization for producing a periodic spatial modulation of the longitudinal magnetization with a wavelength, and adapted for performing said pulse sequence with a given field of view, a receiver unit adapted for acquiring a first image of said object upon performing said pulse sequence with the given field of view, wherein the sequence control unit is adapted for applying a second modulation of the longitudinal magnetization for spatially shifting the modulation function by a part of the wavelength (which may be denoted as a sub-wavelength shift), and adapted for performing said pulse sequence with a field of view shifted by a part of a pixel width, wherein the receiver unit is adapted for acquiring a second image of said object upon performing said pulse sequence with the field of view shifted by the part of the pixel width (which may be denoted as a sub-pixel shift), and an arithmetic unit adapted for merging said first image and said second image in the image domain in an interleaved fashion pixel by pixel, wherein the arithmetic unit is adapted for deconvolving the merged image using the modulation function.

According to still another exemplary embodiment of the invention, a magnetic resonance imaging device for generating a magnetic resonance image of an object with a given pulse sequence is provided, the magnetic resonance imaging device comprising a controller having the above mentioned features. According to yet another exemplary embodiment of the invention, a computer-readable medium (for instance a CD, a DVD, a USB stick, a floppy disk or a harddisk) is provided, in which a computer program for obtaining a magnetic resonance image of an object with a given pulse sequence is stored which, when being executed by a processor, is adapted to control or carry out a method having the above mentioned features.

According to still another exemplary embodiment of the invention, a program element (for instance a software routine, in source code or in executable code) for obtaining a magnetic resonance image of an object with a given pulse sequence is provided, which program element, when being executed by a processor, is adapted to control or carry out a method having the above mentioned features.

Data acquisition and processing for MRI purposes which may be performed according to embodiments of the invention can be realized by a computer program, that is by software, or by using one or more special electronic optimization circuits, that is in hardware, or in hybrid form, that is by means of software components and hardware components.

The term "given pulse sequence" may particularly denote a sequence of gradient pulses, RF pulses, and data sampling in order to generate data for reconstructing an MRI image. The resulting image contrast and properties (e.g. sensitivity for motion induced artefacts or inhomogeneities of the main magnetic field) largely depend on the timing and amplitudes of the gradient and RF pulses. The "given pulse

sequence" can be a routinely used standard sequence with a

characteristic image contrast (like a spin echo sequence, a gradient echo sequence or an echo-planar imaging sequence), but can also denote a sequence for non-structural imaging (e.g. angiography, perfusion imaging, diffusion imaging, blood oxygenated level dependent contrast imaging). The term "modulation of longitudinal magnetization" may particularly denote a process of spatially varying the strength of the longitudinal magnetization, which is oriented parallel to the main magnetic field of an M RI apparatus. The longitudinal magnetization represents the macroscopic steady state magnetization (or net

magnetization) and is build up by the individual ¹H (hydrogen) spins which are oriented anti-parallel or parallel along the main magnetic field . Consequently, transversal magnetization may be a magnetization component in a plane perpendicular to such a direction of the net magnetization, wherein a dedicated pulse sequence may allow to tilt the magnetization out of the longitudinal direction into the transversal plane.

The term "field of view" may denote the size and position of the two or three dimensional spatial encoding area of an image, i .e. the part of an object that is depicted with an M RI instrument at a given moment.

The term "spatially shifting the modulation function" may particularly denote a process where the phase of the RF pulses for the modulation of the longitudinal magnetization is changed such that the modulation function is shifted in space, i .e. the minima and maxima of the modulated magnetization are shifted by a certain distance along a predefined direction .

The term "merging said first image and said second image in the image domain in an interleaved fashion pixel by pixel" may particularly denote mathematical combining image data related to the images in accordance with a super-resolution scheme. This may allow to obtain a higher resolution image from a plurality of lower resolution images by making use of complementary information of the various images.

The term "deconvolving" may particularly denote the numerical inversion of a convolution of the high resolution image with the

modulation function . A convolution, which is inherent in most super- resolution techniques, results in a smoothing of the super-resolution image and therefore has to be eliminated by a deconvolution .

The term "pixel" may particularly denote a picture element such as a two-dimensional image element. A "voxel" can be considered as a three-dimensional image element. Correspondingly, the three- dimensional analogue of a pixel is a voxel . In one embodiment, an improvement of the resolution can be achieved with a pixel shift. In another embodiment, a more-dimensional improvement of the resolution can be obtained with a voxel shift.

The term "k-space" may also be denoted as Fourier domain, spatial frequency space, wave number space, or even momentum space. The term "k-space" may denote the data space in which MRI raw data is collected or acquired. In contrast to this, the Fourier-transformed counterpart of the k-space may be denoted as image space or image domain, or even position space.

According to an exemplary embodiment of the invention, a magnetic resonance image acquisition and reconstruction system is provided that is capable of synthesizing a high-resolution magnetic resonance image from a series of low-resolution images. This may result in enhancing the spatial resolution of a magnetic resonance image of an object. A gist of an embodiment of the invention is to shift a field of view and modulate the transverse magnetization which allows to acquire a series of sub-pixel or sub-voxel shifted images with new and independent k-space data.

According to an exemplary embodiment of the invention, a super- resolution MRI system is provided which is based on a spatial modulation of the magnetization in a longitudinal direction. In other words, the magnetization may be forced to become spatially dependent with a certain periodicity within an object under analysis (such as tissue of a human being). With a corresponding modulation characteristic, a first image may be captured in a first field of view. For a subsequent second image, both may be shifted simultaneously, the modulation distribution as well as the field of view. By taking this measure, two (or more) relatively poorly resolved individual images may be captured which slightly differ as a result of the shift of the spatial modulation and the shift of the field of view. This double shift may allow to provide complementary information by individual low resolution images so that a combination of them may allow to reconstruct a high resolution image in accordance with a super-resolution principle. By reading out one slice after the other in a multi-slice architecture, a two-dimensional information can be obtained. Also a three-dimensional image can be obtained in another embodiment.

Thus, an exemplary embodiment of the invention provides an improvement of the resolution of an MRI scan. In contrast to

conventional MRI scans having a resolution which depends primarily on data sampling time and available gradient strength, an exemplary embodiment of the invention makes efficient use of the super-resolution concept, i.e. to produce an image with a high resolution from several slightly displaced images having a smaller resolution. An embodiment of the invention implements an additional microscopic or macroscopic modulation of the magnetization allowing to make the super-resolution principle also applicable to MRI systems.

In the following, further exemplary embodiments of the method will be explained . However, these embodiments also apply to the controller, to the magnetic resonance imaging device, to the program element and to the computer-readable medium.

In an embodiment, the first modulation of the longitudinal magnetization may be applied or adjusted with two subsequent radio frequency pulses (which may also be denoted as high-frequency pulses) and a gradient pulse between, thereby producing a periodic spatial modulation of the longitudinal magnetization for instance with a

wavelength of one pixel (other wavelengths may be possible as well, for instance a wavelength of two pixels, three pixels, etc.). Accordingly, the second modulation of the longitudinal magnetization may be applied or adjusted with two subsequent radio frequency pulses and a gradient pulse between, thereby spatially shifting the modulation function for instance by half of the wavelength (other wavelength shifts may be possible as well, for instance a wavelength shift of one and a half pixels, two and a half pixels, etc.). Such a sequence of two radio frequency pulses sandwiching a gradient pulse may be particularly appropriate for adjusting the modulation in the desired way. By adjusting the

characteristic of these three pulses, the properties and parameters of the modulation function may be set as well.

The pulse sequence may be performed with the field of view shifted by for instance half of a pixel width to acquire the second image of the object (other amounts of shifting the field of view may be possible as well, for instance a shift of one and a half pixels, two and a half pixels, etc.).

However, particularly the combination of a shift of the modulation function by an amount of half of a wavelength with the shift of the field of view by half of a pixel width may allow to obtain important

complementary information from the different low resolution images.

The method may further comprise repeating applying the first modulation and applying the second modulation with different gradient directions to allow for a two- or three-dimensional modulation of the longitudinal magnetization. Thus, the system may also allow to modulate the magnetizations in a multiple number of dimensions to further improve the resolution.

In an embodiment, the method may comprise N acquisitions of said images after modulation of the longitudinal magnetization, thereby shifting the field of view and the modulation function by the factor wavelength/N in one or more directions. N may be an integer number larger than two or equal to two (N = 2,3,...). The larger N, the more can the method be refined to further improve the resolution of the super- resolved image.

The method may further comprise repeating said radio frequency pulses and the gradient pulse for the modulation of the longitudinal magnetization several times to modify the modulation function. Also this feature allows to refine the modulation scheme which may have a positive impact on the resolution of the final image.

The method may further comprise generating said modulation of the longitudinal magnetization with more than two radio frequency pulses. Therefore, other pulse sequences as compared to two radio frequency pulses separated by a gradient pulse may be applied as well . However, even with such alternative pulse sequences, the above- discussed manipulation of the magnetization vector can be achievable.

The method may comprise performing said merging of said first image and said second image in the Fourier domain. The merging may be performed in the Fourier (wavelength number) space, whereas the final image may then be reconstructed into the image space, i.e. in a domain perceivable by a human user.

The wavelength of the modulation function may be larger than the width of one pixel (or voxel). Therefore, the above example of a wavelength of one pixel is only exemplary, and other wavelengths are possible as well.

Next, further exemplary embodiments of the controller will be explained . However, these embodiments also apply to the method, to the magnetic resonance imaging device, to the computer-readable medium and to the program element. In an embodiment, the sequence control unit may be programmed to acquire N>2 acquisitions of said images after modulation of the longitudinal magnetization, thereby shifting the field of view and the modulation function by the factor wavelength/N in one or more

directions. The sequence control unit may be a processor or a part thereof allowing to control the sequences applied by the system.

Particularly, the sequence control unit may be programmed to apply the first modulation of the longitudinal magnetization with two subsequent radio frequency pulses and a gradient pulse between thereby producing the periodic spatial modulation of the longitudinal

magnetization with a wavelength of one pixel, and to apply the second modulation of the longitudinal magnetization with two subsequent radio frequency pulses and a gradient pulse between, thereby spatially shifting the modulation function by half of the wavelength. However, other sequences of radio frequency pulses and/or gradient pulses are possible as well, as well as different spatial modulations and different wavelengths differing from one pixel. Accordingly, also the shift by half of the wavelength is only exemplary and can be substituted by other

appropriate values allowing to sufficiently improve the resolution of a final image of the object.

The sequence control unit may be programmed to subsequently repeat said radio frequency pulses and gradient pulses several times to modify the modulation function. The sequence control unit may further be programmed to modulate the longitudinal magnetization with more than two radio frequency pulses. Hence, any appropriate user-defined sequence can be applied.

The arithmetic unit may be adapted for firstly transferring the individual images in the image space and to perform the merging of the data in the image space. However, it is also possible to perform said image fusion before performing a Fourier transformation transforming the acquired and processed data from the k-space into the image space. In other words, the fusion of the images may be performed in the k- space before transferring the k-space image into the inverse Fourier space, i.e. the image space.

The sequence control unit may be programmed to generate a modulation with a wavelength being longer than the width of one pixel . Therefore, one pixel is only an example for a suitable wavelength and other values may be used as well. In the following, further exemplary embodiments of the magnetic resonance imaging device will be explained . However, these

embodiments also apply to method, to the controller, to the program element and to the computer-readable medium.

In an embodiment, the magnetic resonance imaging device comprises a magnet adapted for generating a constant magnetic field about said object. Such a magnetic field may be the main magnetic field defining the longitudinal direction along which a net magnetization of spins of material of the object is aligned.

Furthermore, the magnetic resonance imaging device may comprise one or more gradient coils adapted for applying one or more gradient pulses to the object. One or more radio frequency coils may be provided for applying one or more radio frequency pulses to the object. Additionally, driving circuitry may be provided which may be adapted for driving said gradient coil(s) and radio frequency coil(s) in accordance with the desired driving scheme. The receiver unit may be adapted for receiving a signal from said object in said magnetic field upon application of said gradient pulses and radio frequency pulses. With these features, it is possible to perform a complete magnetic resonance imaging analysis.

The magnetic resonance imaging device may furthermore comprise a display device for displaying said received and processed signals, particularly for displaying the reconstructed images. Such a display device may form or may form part of an input/output unit allowing a user to unidirectionally or bidirectionally communicate with the magnetic resonance imaging device. Via such an input/output unit, a user may input control commands to the magnetic resonance imaging device and may also observe the result image.

The sequence control unit may be adapted for controlling the radio frequency coil(s) and gradient coil(s) to generate a periodic modulation of the longitudinal magnetization and may be adapted for performing the pulse sequence to acquire the first image and the second image of the object with the receiver unit. Thus, the sequence of RF pulses, gradient pulses and receiving data may be controlled by the sequence control unit.

The sequence control unit may be further programmed to repeat the modulation steps with different gradient coils to allow for a two- dimensional modulation of the longitudinal magnetization or even a three-dimensional modulation of the longitudinal magnetization.

The aspects defined above and further aspects of the invention are apparent from the examples of embodiment to be described hereinafter and are explained with reference to these examples of embodiment.

BRIEF DESCRIPTION OF THE DRAWINGS

The invention will be described in more detail hereinafter with reference to examples of embodiment but to which the invention is not limited .

Fig. 1 shows a pulse sequence timing for achieving a spatial modulation of the longitudinal magnetization (SPAMM) for a given sequence according to an exemplary embodiment of the invention.

Fig . 2 shows the spatial variation of the longitudinal magnetization M_z as a consequence of the SPAMM preparation of Fig . 1 according to an exemplary embodiment of the invention. Fig . 3 demonstrates the effect of spatial modulation according to an exemplary embodiment of the invention, wherein compared to the native sequence without SPAMM preparation (left) the SPAMM prepared sequence results in stripes which are orientated orthogonal to the direction of the field gradient (right).

Fig. 4a illustrates that in conventional MRI all spins within a voxel (cube) contribute equally to the magnetic resonance signal because there is no spatial modulation of the magnetization.

Fig . 4b and Fig . 4c show the result of a ID and 2D modulation, respectively, of the longitudinal magnetization according to an exemplary embodiment of the invention, wherein areas with low magnetization will only marginally contribute to the magnetic resonance image.

Fig. 5 illustrates that repeating the SPAMM preparation for several times can lead to a sharper point spread function with a smaller width at half maximum according to an exemplary embodiment of the invention.

Fig.6 shows an SRMRI method for a simple ID case according to an exemplary embodiment of the invention, wherein two SPAMM prepared acquisitions are shifted such that the spatial independent information is maximized, wherein a high resolution image can then be obtained by merging these two images voxel by voxel in an interleaved way.

Fig . 7 is a simplified flow diagram of an embodiment of an SRMRI process according to the invention to simultaneously increase the resolution in two dimensions.

Fig . 8 shows a block diagram of a magnetic resonance imaging system that may be programmed for SRMRI according to an exemplary embodiment of the invention.

Fig. 9 illustrates an application of a method according to an embodiment of the invention to increase the resolution of a single shot echo planar imaging sequence. Fig . 10 illustrates a principle of a method according to an

embodiment of the invention for a 1-dimensional scenario.

DESCRIPTION OF EXEMPLARY EMBODIMENTS

The illustration in the drawing is schematically.

A concept for increasing the resolution of magnetic resonance imaging according to an embodiment of the invention is based on SRMRI. The SRMRI process combines information from a plurality of low- resolution images with a field of view (FOV) shifted by a distance less than a pixel width to create a synthesized image with a higher resolution.

A shift of the field of view is usually done as a post-processing step, either by shifting the reconstructed image or by manipulating k- space data (in two dimensions: k_x, k_y). The inverse Fourier space of the k-space is the image space (in two dimensions: x,y). As given by the Fourier-shift theorem, a spatial shift in image space is equivalent to a linear phase modulation in k-space along the shift direction. Therefore, the image shift can be accomplished by multiplying the k-space data (for instance in phase encoding direction) according to

S'(k_x , k_y ) = S(k_x , k_y )e (2) where S' is the shifted k-space data and S is the original k-space data with :

k = G t (4)

In equation (3), M_xy is the transverse magnetization, i.e. a magnetization component perpendicularly to the longitudinal z-direction. Equations (3) and (4) show that a magnetic resonance image can be obtained by a Fast Fourier Transformation (FFT) of k-space data S(k_x,k_y), i.e. the domain of data acquisition.

The sampling positions (k_x,k_y) are uniquely defined by the given resolution and by the field of view. Therefore, no new k-space

information is acquired just by changing the phase of the k-space data. It is presently believed that this is the reason why so far all super- resolution approaches have failed to significantly improve the in-plane resolution in MRI.

An exemplary embodiment of the current invention provides a method to circumvent the limitation of the Fourier theorem by

modulating the transverse magnetization M_xy before a field of view shift. If M_xy in equation (3) is modulated differently for each of n field of view shifts, n different and independent k-space data can be acquired. Then, a high resolution image can be composed from a series of such shifted and modulated images because the modulation function and the spatial relationship of the field of view shifts are known .

In an exemplary embodiment of the invention, the modulation of the transverse magnetization M_xy may be achieved by a spatial

modulation of the longitudinal magnetization (SPAMM) M_z prior to the excitation RF pulse. The excitation pulse then flips the longitudinal modulation into the transverse plane thus changing M_xy.

The concept of SPAMM as such has been disclosed in US 5,054,489 (Magnetic Resonance Imaging using spatial modulation of magnetization) and is explicitly incorporated in the disclosure of this patent application by reference. The concept of SPAMM has been used predominantly to investigate motion patterns of myocardial tissue over the cardiac cycle or motion of other tissues.

As can be taken from Fig. 1 showing the time (t) axis reflecting a sequence of pulses applied during an MRI analysis, SPAMM employs two RF pulses and a magnetic field gradient between them. The first RF pulse ai flips the longitudinal magnetization, or a part of it, into the transverse plan. The field gradient then causes a linear change of the resonance frequency along the direction of the gradient. When the gradient is turned off, spins exhibit a different phase according to their position. The second RF pulse a₂ then flips the magnetization back into the longitudinal direction, where the sign and amplitude of the resulting magnetization may depend on the phase of the individual spins.

As can be taken from Fig. 2, these steps result in a sinusoidal modulation of the longitudinal magnetization. Flip angles ai and a ₂ determine the extremes of the periodic modulation function, whereas the wavelength is controlled by the timing of the gradient field.

The sinusoidal modulation of the longitudinal magnetization according to Fig . 2 can be appreciated as stripes with sinusoidal variation of the signal intensity in the corresponding image shown in Fig. 3.

Thereby, the spatial wavelength R is given by the amplitude and duration of the gradient field G [T/m] :

When repeating the SPAMM preparation with a different orientation of the gradient field, also a two-dimensional (2D) or even 3-dimensional (3D) modulation of the magnetization is possible.

In an embodiment of the invention, SPAMM is employed such that the longitudinal magnetization is maintained only in the central part of a voxel. Then, only central spins will contribute to the actual image.

Consequently, the measured magnetization is no longer related to the actual voxel dimensions but provides sub-voxel information only.

In an embodiment of the invention, one-dimensional (I D), 2D, or 3D SPAM M can be performed immediately prior to any desired pulsed sequence, whereas the timing of the field gradient G, i .e. the amplitude and duration, is controlled such that the spatial wavelength R

corresponds exactly to the width of one voxel . Under such a condition the SPAM M profile is invisible on the M R image ("microscopic SPAM M") . The phase of the RF pulses should be controlled such that the maximum of the sinusoidal function is located exactly in the centre of each voxel . In addition, the flip angle of the RF pulses ai and a₂ should preferably be equal or less than 45° so that the resulting longitudinal magnetization has only positive values. Otherwise, negative components could contribute to intra-voxel annihilation of the magnetization .

Fig . 4a to Fig . 4c show the longitudinal magnetization for a 6x2 voxel array in different scenarios. Fig. 4a relates to a scenario without or before 2D SPAM M preparation, Fig. 4b relates to a scenario after I D SPAMM preparation, and Fig. 4c relates to a scenario after 2D SPAM M preparation .

Basically, the preparation produces a sinusoidal shape of the longitudinal magnetization . Other shapes can be achieved with binomial RF pulses of higher order, i.e. with more than two RF pulses for the SPAM M preparation or by repeating the SPAM M preparation scheme several times.

Fig. 5 illustrates the effect of repeating the SPAM M preparation several times before running the image acquisition part to achieve a sharper point spread function .

For the sake of simplicity, the super-resolution method according to an exemplary embodiment of the invention is now explained using a I D scenario with two SPAM M prepared acquisitions. As shown in Fig. 6, these two acquisitions can have a low resolution but should be shifted by half the voxel width along the direction of the desired resolution improvement. Although there is a 50 % overlap of the corresponding voxels in both acquisitions, due to the SPAMM preparation there is only very few overlap in available longitudinal magnetization, i.e. the two images represent different regional information of the imaged object. Portions of the shifted images are then merged into a single image by interleaving their voxels. Finally, as SPAMM does not provide a

rectangular modulation of the longitudinal magnetization, the merged image can be deconvolved by the modulation function, which however is known.

In an exemplary embodiment of the invention, more than two shifted images may be acquired thereby obtaining even higher spatial resolution of the reconstructed image. Also, field of view shifting may be done in different directions and SPAMM may be done in 2D or 3D to increase the spatial resolution in more than one dimension.

Referring to Fig. 7, a simplified flow diagram of an exemplary embodiment of a 2D SRMRI process according to the invention is shown .

In box 1 the actual position of the field of view and the position of maxima of the 2D SPAMM grid are calculated . Shifting is done in small incremental steps which are multiples of voxel width/maximum number of shifts in one dimension . The variables x and y count the actual shifts in both dimensions. With the updated spatial information a 2D SPAMM preparation is performed as indicated in box 2, followed by the

performance of the respective sequence as indicated in box 3. The loops over the two dimensions are controlled by box 4 and box 5, respectively. The maximum number of shifts in one dimension is identical to the resolution improvement factor. The total number of shifts then is given by the maximum number of shifts in one direction multiplied by the maximum number of shifts in the other direction. The steps indicated by box 6 indicate the image fusion portion of the SRMRI method . The high resolution image matrix is filled with the shifted images by interleaving their voxels in both dimensions. Referring to box 7, the high resolution image then is deconvolved with the 2D modulation function.

A system for acquiring shifted and modulated images and for reconstructing high resolution images according to an exemplary embodiment of the invention is shown in Fig. 8. This system can be a magnetic resonance microscope, a 1.5T or 3T whole-body scanner, or any other suitably equipped MRI system that may be programmed for SRMRI.

As illustrated in Fig. 8, a magnet 10 generates a static,

fundamental magnetic field along a z-axis 12, in which an object or the body of a patient 14 to be examined is situated . The system additionally comprises gradient amplifiers 16, gradient coils 18, transmitter 20, RF power amplifier 22, RF coils 24 for generating pulse sequences for application to selected slices of the patient's anatomy or the sample. The control of the pulse sequence is done by a sequence control unit 26, which can be programmed through a scan control interface 28. Since software techniques for generating pulse sequences with the

characteristics defined below are believed to be well-known to those skilled in the art, such pulse generating techniques will not be described in further details herein. The signal generated by the pulse sequence is received by signal receiver 30 and digitized at digitizer 32 for application to an arithmetic unit 34 for processing in accordance with the technique of an embodiment of the invention. The processed signal is then displayed on a display unit 36. Data storage 38 and filming with a camera device 40 may be provided additionally.

Magnetic resonance imaging device of Fig. 8 comprises magnet 10 which generates a magnetic field about object 14, gradient coils 18 which apply gradient pulses to said object 14, RF coils 24 which apply RF pulses to said object 14, driving circuitry 16, 22 which drives said gradient coils 18 and RF coils 24, receiving circuitry 30 which receives a signal from said object 14 in said magnetic field upon application of said gradient and RF pulses, an arithmetic unit 34, a display device 36 for displaying said received and processed signals, and a sequence control device 26 which controls said RF coils 24 and gradient coils 18 to generate a periodic modulation of the longitudinal magnetization and which performs said pulse sequence to acquire a first image and a second image of said object 14 with said receiving circuitry 30 and digitizer 32, where the field of view and the modulation function of the second image is shifted by half of the wavelength, and which arithmetic unit 34 merges said first and second image in an interleaved fashion pixel by pixel and deconvolves the resolution high resolution image. Said sequence control device 26 is programmed to repeat the modulation steps with different gradient coils 18 to allow two- or three- dimensional modulation of the longitudinal magnetization. Said sequence control device 26 is programmed to acquire N acquisitions of said images after modulation of the longitudinal magnetization, thereby shifting the field of view and the modulation function by the factor wavelength/N in one, two, or three directions. Said sequence control device 26 is programmed to subsequently repeat said RF pulses and gradient pulses several times to modify the modulation function. Said sequence control device 26 is programmed to modulate the longitudinal magnetization with more than two RF pulses. Said image fusion is done with said arithmetic unit 34 after Fourier transformation. Said sequence control device 26 is programmed to generate a modulation with a wavelength that is longer than the width of one pixel .

Magnetic resonance imaging device of Fig . 8 is capable of executing a method for enhancing the spatial resolution of a magnetic resonance image of an object with a given pulse sequence. Such a method may comprise the steps of:

1. applying a first modulation of the longitudinal magnetization with two subsequent RF pulses and a gradient pulse between thereby producing a periodic spatial modulation of the longitudinal magnetization with a wavelength of one pixel

2. performing said pulse sequence with a given field of view to acquire a first image of said object

3. applying a second modulation of the longitudinal magnetization with two subsequent RF pulses and a gradient pulse between thereby spatially shifting the modulation function by half of the wavelength

4. performing said pulse sequence with the field of view shifted by half one pixel width to acquire a second image of said object

5. merging said first and second image in the image domain in an interleaved fashion pixel by pixel

6. deconvolving the merged image using the modulation function. A 2D super-resolution method according to an embodiment of the invention was implemented on a 3.0 Tesla whole-body scanner and tested in a phantom that contains geometric objects with different shapes and sizes. The method was implemented with a single shot echo planar imaging sequence which is commonly known to suffer from a limited spatial resolution. The in-plane resolution provided by such a sequence typically lies between 2 and 3 mm. The sequence parameters were :

repetition time = 3 s, echo time = 70 ms, image matrix = 128x128, in- plane resolution = 2x2 mm, slice thickness = 4 mm, parallel imaging factor = 2. Prior to the RF excitation pulse a 2D SPAMM preparation was performed . Signal reception was done with a 12-element array coil. A total of 16 shifted and SPAMM prepared images were acquired as a combination of 4 shifts in each direction. This resulted in a resolution improvement factor of 4, i.e. the high resolution image had an in-plane resolution of 0.5 x 0.5 mm. The resolution improvement achieved with the method according to an embodiment of the invention can be clearly appreciated in Fig. 9.

Fig. 10 illustrates a principle of a method according to an embodiment of the invention for a 1-dimensional scenario similar to Fig .6 but providing 4 shifts instead of two.

Fig. 10 illustrates the ideal case that rectangular SPAMM excites only the central part of the voxel. A high resolution image can be composed from a series of images, where the field of view and the modulation field are shifted with small subvoxel steps (if not rectangular there may be a deconvolution with the SPAMM profile). Although one voxel covers a larger area, signal comes from a central subvoxel area only. Composing the 4 shifted images into a high resolution image results in a resolution enhancement factor of 4.

Exemplary embodiments of the invention may particularly provide the following advantages:

- The system and method can be combined with any given pulse sequence and with any image encoding technique (2D/3D-Fourier, spiral, and radial)

- The final image resolution is independent of the sampling bandwidth, therefore echo time becomes largely independent from image resolution

- High resolution imaging with short echo time becomes feasible

- Image fusion may be done in the image domain (no reference scan, no navigator, and no additional k-lines are necessary)

- Image fusion is simple, fast, and robust (no fitting required) Exemplary embodiments of the invention may particularly allow the following applications:

- High resolution single shot techniques (boosting resolution for many application fields) such as diffusion tensor imaging (DTI), perfusion weighted imaging (PWI), or blood oxygenated level dependent contrast (BOLD) imaging

- Correction of image warping or motion using "macroscopic" SPAMM . If the wavelength of the modulation function is larger than one pixel it becomes visible in the image and can be used to un-warp or register the image.

- MR-microscopy with less constraints and faster reconstruction

Embodiments of the invention are based on a combined shift of the field of view of an image and the modulation of the longitudinal magnetization. Above, a detailed embodiment of the invention has been described where the wavelength of the modulation function corresponds exactly to the width of one voxel . It should be noted, that the method of the invention can also be implemented with a wavelength that is substantial longer than one voxel or also with a nonlinear modulation function. Those skilled in the art will appreciate that many additional modifications are possible in the exemplary embodiment without materially departing from the novel teachings and advantages of the invention as defined in the appended claims.

It should be noted that the term "comprising" does not exclude other elements or features and the "a" or "an" does not exclude a plurality. Also elements described in association with different

embodiments may be combined.

It should also be noted that reference signs in the claims shall not be construed as limiting the scope of the claims.

Claims

C l a i m s

1. A method for obtaining a magnetic resonance image of an object (14) with a given pulse sequence, the method comprising :

applying a first modulation of longitudinal magnetization for producing a periodic spatial modulation of the longitudinal magnetization with a wavelength;

performing the pulse sequence with a given field of view to acquire a first image of the object (14);

applying a second modulation of the longitudinal magnetization for spatially shifting the modulation function by a part of the wavelength; performing the pulse sequence with a field of view shifted by a part of a pixel width to acquire a second image of the object (14);

merging the first image and the second image in the image domain in an interleaved fashion pixel by pixel;

deconvolving the merged image using the modulation function.

2. The method according to claim 1, wherein the first modulation of the longitudinal magnetization is applied with two subsequent radio

frequency pulses and a gradient pulse between, thereby producing the periodic spatial modulation of the longitudinal magnetization with a wavelength of one pixel.

3. The method according to claim 1 or 2, wherein the second modulation of the longitudinal magnetization is applied with two subsequent radio frequency pulses and a gradient pulse between, thereby spatially shifting the modulation function by half of the wavelength.

4. The method according to claims 1 to 3, wherein the pulse sequence is performed with the field of view shifted by half of a pixel width to acquire the second image of the object (14).

5. The method according to claims 1 to 4, wherein the method comprises repeating applying the first modulation and applying the second modulation with different gradient directions to allow two-dimensional modulation of the longitudinal magnetization or three-dimensional modulation of the longitudinal magnetization.

6. The method according to claims 1 to 5, comprising N acquisitions of the images after modulation of the longitudinal magnetization, thereby shifting the field of view and the modulation function by the factor wavelength/N in one direction, two directions, or three directions, wherein N is an integer number larger than or equal to two.

7. The method according to claims 2 to 6, wherein the method comprises repeating the radio frequency pulses and the gradient pulse between for the modulation of the longitudinal magnetization multiple times to modify the modulation function.

8. The method according to claims 2 to 7, wherein the method comprises generating the modulation of the longitudinal magnetization with more than two radio frequency pulses.

9. The method according to claims 1 to 8, wherein the method comprises doing the merging of the first image and the second image in a Fourier domain.

10. The method according to claims 1 to 9, wherein the wavelength of the modulation function is larger than the width of one pixel.

11. A controller (50) for a magnetic resonance imaging device for generating a magnetic resonance image of an object (14) with a given pulse sequence, the controller (50) comprising :

a sequence control unit (26) adapted for applying a first

modulation of longitudinal magnetization for producing a periodic spatial modulation of the longitudinal magnetization with a wavelength, and adapted for performing the pulse sequence with a given field of view; a receiver unit (30) adapted for acquiring a first image of the object (14) upon performing the pulse sequence with the given field of view;

wherein the sequence control unit (26) is adapted for applying a second modulation of the longitudinal magnetization for spatially shifting the modulation function by a part of the wavelength, and adapted for performing the pulse sequence with a field of view shifted by a part of a pixel width;

wherein the receiver unit (30) is adapted for acquiring a second image of the object (14) upon performing the pulse sequence with the field of view shifted by the part of the pixel width;

an arithmetic unit (34) adapted for merging the first image and the second image in the image domain in an interleaved fashion pixel by pixel;

wherein the arithmetic unit (34) is adapted for deconvolving the merged image using the modulation function.

12. The controller (50) according to claim 11, wherein the sequence control unit (26) is programmed to acquire N acquisitions of the images after modulation of the longitudinal magnetization, thereby shifting the field of view and the modulation function by the factor wavelength/N in one direction, two directions, or three directions, wherein N is an integer number larger than or equal to two.

13. The controller (50) according to claim 11 or 12, wherein the sequence control unit (26) is programmed to

apply the first modulation of the longitudinal magnetization with two subsequent radio frequency pulses and a gradient pulse between, thereby producing the periodic spatial modulation of the longitudinal magnetization with a wavelength of one pixel; and

apply the second modulation of the longitudinal magnetization with two subsequent radio frequency pulses and a gradient pulse between, thereby spatially shifting the modulation function by half of the

wavelength.

14. The controller (50) according to claim 13, wherein the sequence control unit (26) is programmed to subsequently repeat the radio frequency pulses and gradient pulses multiple times to modify the modulation function.

15. The controller (50) according to claim 13 or 14, wherein the sequence control unit (26) is programmed to modulate the longitudinal magnetization with more than two radio frequency pulses.

16. The controller (50) according to claims 11 to 15, wherein the arithmetic unit (34) is adapted for doing the image fusion before Fourier transformation.

17. The controller (50) according to claims 11 to 16, wherein the sequence control unit (26) is programmed to generate a modulation with a wavelength that is longer than the width of one pixel.

18. A magnetic resonance imaging device for generating a magnetic resonance image of an object (14) with a given pulse sequence, the magnetic resonance imaging device comprising a controller (50) according to claim 11 to 17.

19. The magnetic resonance imaging device according to claim 18, comprising a magnet (10) adapted for generating a static magnetic field about the object (14).

20. The magnetic resonance imaging device according to claim 19, comprising

one or more gradient coils (18) adapted for applying gradient pulses to the object (14);

one or more radio frequency coils (24) adapted for applying radio frequency pulses to the object (14);

driving circuitry (16, 22) adapted for driving the one or more gradient coils (18) and the one or more radio frequency coils (24);

wherein the receiver unit (30) is adapted for receiving a signal from the object (14) in the magnetic field upon application of the gradient pulses and the radio frequency pulses.

21. The magnetic resonance imaging device according to claim 20, comprising a display device (36) for displaying the received signal after processing .

22. The magnetic resonance imaging device according to claim 20 or 21, wherein the sequence control unit (26) is adapted for controlling the one or more radio frequency coils (24) and the one or more gradient coils (18) to generate the periodic modulation of the longitudinal

magnetization and which is adapted for performing the pulse sequence to acquire the first image and the second image of the object (14) with the receiver unit (30).

23. The magnetic resonance imaging device according to claims 20 to 22, wherein the sequence control unit (26) is programmed to repeat the modulation steps with different gradient coils (18) to allow a two- dimensional modulation of the longitudinal magnetization or a three- dimensional modulation of the longitudinal magnetization.

24. A computer-readable medium, in which a computer program for obtaining a magnetic resonance image of an object (14) with a given pulse sequence is stored, which computer program, when being executed by a processor (26, 30, 34), is adapted to carry out or control a method according to claims 1 to 10.

25. A program element for obtaining a magnetic resonance image of an object (14) with a given pulse sequence, which program element, when being executed by a processor (26, 30, 34), is adapted to carry out or control a method according to claims 1 to 10.