WO2008135879A1

WO2008135879A1 - Accelerated t2-mapping in mri through undersampling and k-t reconstruction

Info

Publication number: WO2008135879A1
Application number: PCT/IB2008/051574
Authority: WO
Inventors: Julien Senegas; Hannes Dahnke; Christian Stehning; Stefanie Winkelmann
Original assignee: Philips Intellectual Property and Standards GmbH; Koninklijke Philips Electronics NV
Current assignee: Philips Intellectual Property and Standards GmbH; Koninklijke Philips NV
Priority date: 2007-05-03
Filing date: 2008-04-24
Publication date: 2008-11-13
Anticipated expiration: 2009-11-03

Abstract

The invention relates to a device for MR imaging of a body (7) placed in an examination volume. The device (1) is arranged to:a) generate a series of MR echo signals for multiple echo time values (TE1, TE2, TE3, TE4) by subjecting at least a portion of said body (7) to an MR imaging sequence of at least one RF pulse and switched magnetic field gradients, b) acquire MR signal data from a set of k- space positions with sub-sampling in the phase-encoding direction, c) estimate MR signal data for skipped k-space positions by combining MR signal data from acquired k-space positions, so as to complete the MR signal data set, d) reconstruct an MR image for each echo time value (TE1, TE2, TE3, TE4) from the completed MR signal data set, and e) compute a relaxation time map from the set of reconstructed MR images.

Description

ACCELERATED T2-MAPPING IN MRI THROUGH UNDERSAMPLING AND K-T RECONSTRUCTION

FIELD OF THE INVENTION

The invention relates to a device for magnetic resonance imaging (MRI) of a body placed in an examination volume.

Furthermore, the invention relates to a method for MRI and to a computer program for a magnetic resonance (MR) device.

BACKGROUND OF THE INVENTION

In MRI, pulse sequences consisting of RF pulses and switched magnetic field gradients are applied to an object (a patient) placed in a homogeneous magnetic field within an examination volume of an MR device. In this way, MR signals are generated, which are scanned by means of radio frequency (RF) receiving antennas in order to obtain information about the object and to reconstruct images thereof. Since its initial development, the number of clinically relevant fields of application of MRI has grown enormously. MRI can be applied to almost every part of the body, and it can be used to obtain important functional information about the human body. The pulse sequence, which is applied during an MRI scan, plays a significant role in the determination of the characteristics of the reconstructed image, such as location and orientation in the object, dimensions, resolution, signal-to-noise ratio, contrast, sensitivity for movements, et cetera. An operator of an MRI device has to choose the appropriate sequence and has to adjust and optimize its parameters for the respective application.

SUMMARY OF THE INVENTION

So-called molecular imaging and diagnostics (MID) is rapidly developing during the last years. MID is sometimes defined as the exploitation of specific molecules for image contrast and for diagnosis. This definition refers to the in- vivo measurement and characterization of cellular and molecular level processes in human subjects and to the analysis of bio molecules to screen, diagnose and monitor the human health status and to assess potential risks. An important prerequisite for molecular imaging is the ability to image specific molecular targets. At the moment, MRI is considered to be one of the most promising modalities in molecular imaging. Therefore, MRI is expected to play an essential role in the clinical use of MID for screening, targeted drug delivery and therapy evaluation. Highly sensitive contrast agents have recently been used to allow MRI of molecular targets and gene expression. As mentioned above, MRI can visualize the anatomy with good spatial resolution, is applicable to all body regions and will allow reproducible and quantitative imaging. It can also be used for intravascular and needle image-guided drug delivery. MR can also partly assess molecular information, for example through spectroscopy or relaxometry. One particularly important tool for MID is MR relaxometry. In MR relaxometry, a spatially resolved quantitative measurement of MR relaxation times is performed. Two examples of relaxation times often measured in MR relaxometry are the longitudinal relaxation time and the transverse relaxation time, usually referred to as Ti and T₂ respectively. The relaxation time values are determined by different physical and chemical parameters. This makes MR relaxometry useful for MID applications. Ti and T₂ sensitively depend on magnetic field strength, molecular structure, temperature, pH, and the presence of contrast agents. Multi-echo imaging sequences are typically used for MR relaxometry. The straightforward approach is to acquire full MR signal data sets for a set of different echo time values and to reconstruct an MR image for each echo time value separately. Finally, a relaxation time map comprising relaxation time values for the individual pixel positions is computed from the set of reconstructed MR images, usually by means of an appropriate exponential fitting algorithm.

One problem of MR relaxometry is that the measurements are characterized by long acquisition times. These long acquisition times are due to the large number of echo time values required. It is possible in principle to decrease the number of echo time values acquired and therefore the repetition time, in order to speed up the MR data acquisition, but this approach has two limitations: First, a reduced repetition time is associated with a loss of SNR, and would therefore yield less precise estimates of the relaxation time. Moreover, the range of relaxation time values that needs to be covered is large (from < 10 ms to > 1000 ms), and is usually not known in advance. Hence, a large number of echo times with a relative short echo spacing is best suited for acquisition. A further limiting factor is that conventional MR relaxometry techniques based on multi-spin-echo sequences (e.g. for T₂ mapping) are characterized by a high specific absorption rate (SAR). Therefore, it is readily appreciated that there is a need for an improved MR relaxometry technique. It is consequently an object of the invention to provide an MR device that enables MR relaxometry with a reduction of acquistion time and a reduction of SAR. In accordance with the present invention, an MR device for MRI of a body placed in an examination volume is disclosed. The device comprises means for establishing a substantially homogeneous main magnetic field in the examination volume, means for generating switched magnetic field gradients superimposed upon the main magnetic field, means for radiating RF pulses towards the body, control means for controlling the generation of the magnetic field gradients and the RF pulses, means for receiving and sampling MR signals, and reconstruction means for forming MR images from the signal samples. The invention proposes that the device is arranged to a) generate a series of MR echo signals for multiple echo time values by subjecting at least a portion of the body to an MR imaging sequence of at least one RF pulse and switched magnetic field gradients, b) acquire MR signal data from a set of k- space positions with sub- sampling in the phase-encoding direction, c) estimate MR signal data for skipped k-space positions by combining MR signal data from acquired k-space positions, d) reconstruct an MR image for each echo time value from the completed MR signal data set, and e) compute a relaxation time map from the set of reconstructed MR images.

The invention is based on the recognition of the fact that the MR signal data acquired in an MR relaxometry measurement for different echo time values exhibit a high degree of temporal correlation. Therefore, there is a certain amount of redundancy within the data. The acquisition time is reduced in accordance with the invention by applying sub- sampling in the phase-encoding direction for each echo time value. The reduction of the acquisition time is directly proportional to the number of skipped phase-encoding steps. This number can also be referred to as acceleration factor. The MR signal data missing at the skipped k-space positions is estimated according to the invention by combining MR signal data from acquired k-space positions. This enables the reconstruction of MR images from a completed MR signal data set, such that the reconstructed images are free from aliasing artifacts. From these artifact-free images the relaxation time values can be determined by means of a standard exponential fitting algorithm. The MR relaxometry approach of the invention relies on the reduction of MR signal data acquisition steps and on the exploitation of correlations in k-space, in time, and/or both in k-space and time. The invention reduces the number of phase-encoding steps, while preserving both spatial and temporal resolution. The reduction in phase-encoding steps directly translates into a corresponding reduction of acquisition time and SAR.

Preferably, the MR device of the invention is arranged to acquire the MR signal data from a different set of k-space positions for each echo time value. This can easily be performed by shifting the k-space positions, from which the MR signals are acquired, in k- space from one echo time value to the next. In this way, a complete set of k-space positions is sampled after a corresponding number of echo time increments, and a uniform coverage of k- space is achieved. This approach can be applied generally to both Cartesian and Non- Cartesian sampling schemes (e.g. radial, spiral). The quality of the reconstructed MR images and the resulting accuracy of the relaxation parameters is optimized in this way.

In a practical embodiment of the invention, the MR signal data for skipped k- space positions are estimated by combining MR signal data acquired from k-space positions adjacent in k-space to the skipped k-space positions and/or by combining MR signal data acquired for echo time values temporally adjacent to the echo time values of the missing MR signal data. In MR relaxometry, correlations of the MR signal data both in k-space and in time can be exploited advantageously, in order to achieve a maximum acceleration of the signal acquisition without significantly reducing the image quality.

According to a further preferred embodiment of the invention, the MR device comprises two or more separate receiving antennas, via which the MR signal data are acquired. In this case, the correlation of the MR signal data acquired via the different antennas is additionally exploited to reconstruct aliasing- free images for each echo time value. Hence, MR signal data acquired via separate receiving antennas are combined in the way described above to estimate the MR signal data for the skipped k-space positions. MR images can be reconstructed for each receiving antenna individually, wherein the reconstructed MR images are combined afterwards using known methods (such as sum-of- squares or Roemer's optimal combination). Alternatively, combined MR images can be reconstructed directly from the MR signal data acquired via the separate antennae.

According to a still further preferred embodiment of the invention, calibration MR signal data are acquired from a central portion of k-space without sub-sampling. By acquiring a central portion of k-space completely, low-resolution MR images are acquired for each echo time value. These low-resolution data are used in accordance with the invention for calibration purposes. For completing the MR signal data set at the skipped k-space positions, linear combinations of MR signal data from acquired k-space positions may be computed according to the invention. The weighting factors for this linear combination can be derived from the calibration MR signal data. For example, a least-squares fit can be performed to compute optimal weighting factors. Translation invariance in k-t-space of the weighting factors can be exploited advantageously during the calibration step. Since the calibration data is acquired without sub-sampling from a central portion of k-space only, the total acquisition time is not significantly increased.

The invention not only relates to a device but also to a method for MRI of at least a portion of a body placed in an examination volume of an MR device, wherein the method comprising the following steps: a) generating a series of MR echo signals for multiple echo time values by subjecting at least a portion of said body to an MR imaging sequence of at least one RF pulse and switched magnetic field gradients, b) acquiring MR signal data from a different set of k-space positions for each echo time value with sub-sampling in the phase-encoding direction, c) estimating MR signal data for skipped k-space positions by combining MR signal data from k-space positions acquired in step b), d) reconstructing an MR image for each echo time value from the MR signal data set completed in step c), and e) computing a relaxation time map from the set of MR images reconstructed in step d).

A computer program adapted for carrying out the imaging procedure of the invention can advantageously be implemented on any common computer hardware, which is presently in clinical use for the control of magnetic resonance scanners. The computer program can be provided on suitable data carriers, such as CD-ROM or diskette. Alternatively, it can also be downloaded by a user from an Internet server.

BRIEF DESCRIPTION OF THE DRAWINGS The enclosed drawings disclose preferred embodiments of the present invention. It should be understood, however, that the drawings are designed for the purpose of illustration only and not as a definition of the limits of the invention. In the drawings

Fig. 1 shows an MR scanner according to the invention;

Fig. 2 illustrates the k-t space sampling scheme proposed by the invention; Fig. 3 illustrates the estimation of skipped k-space data in accordance with the invention.

DETAILED DESCRIPTION OF THE EMBODIMENTS In Fig.l an MR imaging device 1 in accordance with the present invention is shown as a block diagram. The apparatus 1 comprises a set of main magnetic coils 2 for generating a stationary and homogeneous main magnetic field and three sets of gradient coils 3, 4 and 5 for superimposing additional magnetic fields with controllable strength and having a gradient in a selected direction. Conventionally, the direction of the main magnetic field is labelled the z-direction, the two directions perpendicular thereto the x- and y- directions. The gradient coils 3, 4 and 5 are energized via a power supply 11. The imaging device 1 further comprises an RF transmit antenna 6 for emitting RF pulses to a body 7. The antenna 6 is coupled to a modulator 9 for generating and modulating the RF pulses. Also provided is a receiver for receiving the MR signals, the receiver can be identical to the transmit antenna 6 or be separate. If the transmit antenna 6 and receiver are physically the same antenna as shown in Fig. 1, a send-receive switch 8 is arranged to separate the received signals from the pulses to be emitted. The antenna 6 can either be a single RF coil or an array of two or more separate RF coils, via which MR signals can be received in parallel. The received MR signals are input to a demodulator 10. The send-receive switch 8, the modulator 9, and the power supply 11 for the gradient coils 3, 4 and 5 are controlled by a control system 12. Control system 12 controls the phases and amplitudes of the RF signals fed to the antenna 6. The control system 12 is usually a microcomputer with a memory and a program control. The demodulator 10 is coupled to reconstruction means 14, for example a computer, for transformation of the received signals into images that can be made visible, for example, on a visual display unit 15. For the practical implementation of the invention, the control system 12 comprises a programming for generating a series of MR echo signals for multiple echo time values by subjecting the body 7 to an MR imaging sequence of at least one RF pulse and switched magnetic field gradients. MR signal data are acquired via antenna 6 from a different set of k-space positions for each echo time value by applying sub-sampling in the phase-encoding direction. The reconstruction means 14 is programmed in accordance with the invention to estimate MR signal data for skipped k-space positions by combining MR signal data from acquired k-space positions, to reconstruct an MR image for each echo time value from the completed MR signal data set, and to compute a relaxation time map from the set of reconstructed MR images. This relaxation time map is finally displayed via the visual display unit 15.

Fig. 2 illustrates the k-space sampling scheme of the invention. The thick black lines represent acquired k-space lines at phase-encoding positions k_y for different echo time values TEi, TE₂, TE₃, and TE₄. The frequency-encoding direction is k_x. The thin dashed lines represent skipped k-space lines. As indicated by the skipped k-space positions in Fig. 2, the MR signal data are acquired with sub-sampling in the phase-encoding direction in order to reduce the total acquisition time. The acceleration factor of the depicted embodiment equals 3. The thick dashed lines in the diagram represent additional k-space lines acquired from the central portion of k-space for calibration purposes. The k-space positions of the acquired k-space lines are shifted from one echo time value TE_n to the next echo time value TE_n+I. This k-space shift can be achieved by the introduction of an additional blip gradient into the imaging pulse sequence, as it is done in echo planar imaging, for example. In this way, a uniform coverage of k-space is achieved after 3 echo time increments. The k-space acquisition scheme is then repeated cyclically. The size of the complete MR signal data set in the phase-encoding direction equals 14 in the depicted embodiment.

In Fig. 3, the k-t-space reconstruction according to the invention is illustrated. In this embodiment, the total number of phase-encoding steps is again 14, and the acceleration factor equals 2. MR signal data for skipped k-space positions is estimated by combining MR signal data from acquired k-space positions in accordance with the invention. The circle in the diagram of Fig. 3 exemplarily indicates a skipped k-space position, for which the missing MR signal data is to be estimated. The acquired MR signal data at the positions marked by a cross contribute to the estimation of the missing MR signal data. MR signal data acquired from k-space positions spatially and temporally adjacent to the skipped k-t position are combined by computing linear combinations of the acquired MR signal data. In the depicted embodiment, the size of the k-t neighborhood used for the computation of the linear combination is 3x3x3 in the k_y-, k_x-, and echo time directions respectively.

A preferred embodiment of the invention is described in detail in the following in the context of a T₂ mapping measurement. If it is assumed that a constant echo time increment AT is used during the acquisition. Hence, the magnetization values m_t (x, y) and m_t__x (x, y) for two consecutive echo time values are linked in image space by the relation m_t(x,y) = m_t__l(x,y)e-^AT/T^^y) , wherein T₂(x,y) represents the relaxation time map. Consequently, the corresponding k- space signals S_t (Jc_x , k_y ) and S_t__γ (Jc_x , Jc _y ) are linked by a convolution product. By computing the Fourier transform of the above equation, the following relation is obtained:

The kernel / of this product is the Fourier transform of the exponential function in the first equation, which does not depend on the echo time index t .

As mentioned above, the technique of the invention includes the acquisition of calibration MR signal data from the central k-space region. These are represented in Fig. 2 by thick dashed k-space lines. If N denotes the total size of the MR signal data set in the phase- encoding direction, R the acceleration factor mentioned above, and N_b the size of the k- space block used for calibration, the effective acceleration factor reads:

R ^N e^ff N/R + N_b x(R - \)

Due to the sub-sampling applied according to the invention in the phase- encoding direction, the MR images obtained for the different echo time values contain aliasing artifacts, if reconstructed by means of a standard Fourier transform algorithm. According to the invention, the missing MR signal data are completed prior to image reconstruction in order to avoid such artifacts. This is done directly in k-t-space as shown schematically in Fig. 3. For the embodiment shown in Fig. 3 (R = 2 ), the missing MR signal data at k-space position (Ji_x ,2k + 1) can be computed for echo time index 2t by the following linear combination:

S_2t(k_x,2k + \) = £ £ ^w_τ^S_2^ (k_x + i,2k + 2j + d_τ ) , τ =— Yi₁ ι=—n_x J=—n_y wherein w_{τ t }} denotes the (yet unknown) weighting coefficients for the linear combination, and d_τ equals 1 if τ is odd, and 0 otherwise. The computation of the estimated MR signal data for the odd echoes with index 2^ + 1 can be performed in a similar manner. Here, for simplicity, neighboring samples have been chosen to lie symmetrically around the skipped k- space position. More sophisticated neighborhood selections are of course possible. It is important to note that the weighting factors w_{τ t }} depend only on the local sub-sampling scheme and on the neighborhood selected for estimating the missing MR signal data, and not on the absolute position in k-space or in time. If an array of n_c receiving antennas is used for signal acquisition, the estimation of the missing data for coil γ reads:

S_jat(k_x,2k + \) = ∑ ∑ £ ∑w^^S^ i^ + iak + lj + dJ , γ= 1 τ =— Yi₁ ι=—n_x J=—n_y with W_{1 τ t j} representing the weighting coefficients for the linear combination. The yet unknown weighting coefficients w_{τ t }} are determined on the basis of the calibration MR signal data. For example, the weighting coefficients w_{τ t }} can be computed by means of a least-squares fit. The sum of the squared differences between the acquired calibration MR signal data and their respective estimates according to the above equation are minimized with respect to the coefficients w_{τ t }} , according to:

w_τM = argmin _tlJS_2tψτ (k_x + i,2k + 2j + d_τ ) j=—n_γ

The solution of this minimization problem is computed via standard algorithms.

Once the missing k-space data are estimated, artifact-free MR images are reconstructed for each echo time value using standard Fourier transform. The relaxation time map T₂ (x, y) is finally computed by pixel- wise fitting an exponential function to the time series of MR images.

Claims

CLAIMS:

1. Device for MR imaging of a body (7) placed in an examination volume, the device (1) comprising means (2) for establishing a substantially homogeneous main magnetic field in the examination volume, means (3, 4, 5) for generating switched magnetic field gradients superimposed upon the main magnetic field, means (6) for radiating RF pulses towards the body (7), control means (12) for controlling the generation of the magnetic field gradients and the RF pulses, means (10) for receiving and sampling MR signals, and reconstruction means (14) for forming MR images from the signal samples, the device (1) being arranged to a) generate a series of MR echo signals for multiple echo time values (TEi, TE₂, TE₃, TE₄) by subjecting at least a portion of said body (7) to an MR imaging sequence of at least one RF pulse and switched magnetic field gradients, b) acquire MR signal data from a set of k- space positions with sub- sampling in the phase-encoding direction (k_y), c) estimate MR signal data for skipped k-space positions by combining MR signal data from acquired k-space positions, d) reconstruct an MR image for each echo time value (TEi, TE₂, TE₃,

TE₄) from the completed MR signal data set, and e) compute a relaxation time map from the set of reconstructed MR images.

2. Device of claim 1, wherein the device (1) is further arranged to acquire the

MR signal data from a different set of k-space positions for each echo time value (TEi, TE₂, TE₃, TE₄).

3. Device of claim 1 or 2, wherein the device (1) is further arranged to estimate

MR signal data for skipped k-space positions by combining MR signal data acquired from k- space positions spatially adjacent to the skipped k-space positions.

4. Device of any one of claims 1-3, wherein the device (1) is further arranged to estimate MR signal data for a given echo time value (TEi, TE₂, TE₃, TE₄) by combining MR signal data acquired for temporally adjacent echo time values (TEi, TE₂, TE₃, TE₄).

5. Device of any one of claims 1-4, comprising two or more separate receiving antennas, via which the MR signal data are acquired, wherein the device is arranged to estimate MR signal data for skipped k-space positions by combining MR signal data acquired via different receiving antennas.

6. Device of any one of claims 1-5, wherein the device (1) is further arranged to acquire calibration MR signal data from a central portion of k-space without sub-sampling.

7. Device of claim 6, wherein the device (1) is arranged to derive weighting factors from the calibration MR signal data, which weighting factors are used for computing linear combination of acquried MR signal data.

8. Method for MR imaging of at least a portion of a body placed in an examination volume of an MR device, the method comprising the following steps: a) generating a series of MR echo signals for multiple echo time values by subjecting at least a portion of said body to an MR imaging sequence of at least one RF pulse and switched magnetic field gradients, b) acquiring MR signal data from a different set of k-space positions for each echo time value with sub-sampling in the phase-encoding direction, c) estimating MR signal data for skipped k-space positions by combining MR signal data from k-space positions acquired in step b), d) reconstructing an MR image for each echo time value from the MR signal data set completed in step c), and e) computing a relaxation time map from the set of MR images reconstructed in step d).

9. Method according to claim 8, wherein MR signal data for a given echo time value are estimated in step c) by combining MR signal data acquired from k-space positions spatially adjacent to the skipped k-space positions and MR signal data acquired for temporally adjacent echo time values.

10. Method of claim 8 or 9, wherein calibration MR signal data are acquired in step b) from a central portion of k-space without sub-sampling, and wherein weighting factors are derived from the calibration MR signal data, which weighting factors are used for combining the MR signal data in step c).

11. Computer program for an MR device, the program comprising instructions for

a) generating an MR imaging sequence of at least one RF pulse and switched magnetic field gradients, b) acquiring MR signal data for each echo time value from a set of k- space positions with sub-sampling in the phase-encoding direction, c) estimating MR signal data for skipped k-space positions by combining MR signal data from k-space positions acquired in step b), d) reconstructing an MR image for each echo time value from the MR signal data set completed in step c), and e) computing a relaxation time map from the set of MR images reconstructed in step d).

12. Computer program according to claim 11, comprising further instructions for estimating MR signal data for a given echo time value in step c) by combining MR signal data acquired from k-space positions spatially adjacent to the skipped k-space positions and MR signal data acquired for temporally adjacent echo time values.

13. Computer program according to claim 11 or 12, comprising further instructions for acquiring calibration MR signal data in step b) from a central portion of k- space without sub-sampling, and for deriving weighting factors from the calibration MR signal data, which weighting factors are used for combining the MR signal data in step c).