Diffusion mapping by MR fingerprinting

Abstract

The invention relates to a method for MR imaging of an object 10 placed in an examination volume of an MR system 1. The object of the invention is to enable quantitative diffusion MR imaging based on the generation of temporally incoherent MR signals by changing acquisition parameters (similar to MRF or MR STAT). The method comprises the steps of subjecting the object 10 to an imaging sequence comprising a train of sequence blocks, each sequence block having at least one RF pulse, at least one switched spoiling field gradient and at least one switched readout field gradient defining a k-space sampling pattern and having associated thereto a set of acquisition parameters including the zeroth moment of the spoiling field gradient, acquiring MR signals during the imaging sequence while varying at least the zeroth moment of the spoiling field gradient and any one or more other acquisition parameters, and reconstructing at least one MR image, in which at least one diffusion coefficient and any one or more other MR parameters are calculated for several image locations from the incoherent temporal evolution of the acquired MR signals caused by the change in the at least one acquisition parameter. Furthermore, the invention relates to an MR system 1 for carrying out this method, and to a computer program running on the MR system.

Description

[Technical field]

[0001] The present invention relates to the field of magnetic resonance (MR) imaging. The present invention relates to a method for MR imaging of an object. The present invention also relates to an MR system and to a computer program running on an MR system. [Background technology]

[0002] MR imaging methods, which utilize the interaction between a magnetic field and atomic nuclear spins to form two- or three-dimensional images, are widely used today, especially in the field of medical diagnostics, since they are superior in many ways to other imaging methods for imaging soft tissues, do not require ionizing radiation, and are usually non-invasive. Summary of the Invention

[0003] [Problem to be solved by the invention]

[0004] MR fingerprinting (MRF) uses the generation of temporally incoherent MR signal evolutions (forming "fingerprints") for different material (tissue) types through continuous variation of acquisition parameters such as flip angle, radio frequency (RF) phase, repetition time (TR) and k-space sampling pattern during the imaging sequence used (see Nature 495:187-192, 2013). For this purpose, an imaging sequence comprises a train of sequence blocks, each sequence block comprising at least one RF pulse and at least one switched readout magnetic field gradient that defines a k-space sampling pattern. An individual set of acquisition parameters is associated with each sequence block. A pattern matching algorithm matches the fingerprint against a predefined dictionary of expected MR signal evolution patterns. The MR parameter maps thus obtained, such as T1, T2, frequency shift and proton density maps, are likewise estimated from the best signal match. Due to the nature of the incoherent sampling and the matching procedure based on prior knowledge, MRFs have been shown to be relatively insensitive with respect to errors, even in combination with highly undersampled k-space acquisitions that give rise to severe aliasing artifacts.

[0005] In the so-called MR STAT technique (Magnetic Resonance Spin Tomography in Time-domain, see Magnetic Resonance Imaging 46:56-62, 2018), the ensemble of magnetic spins in the subject is treated as a large-scale nonlinear dynamical system and is probed by superimposing a train of imaging sequence blocks in which the acquisition parameters vary incoherently (as in MRF). Based on this, quantitative MR images are performed as a one-step process; signal localization and parameter quantification are obtained simultaneously by the solution of a large-scale nonlinear inverse problem. Quantitative parameter maps are reconstructed by employing a nonlinear optimization algorithm and a parallel computational infrastructure that does not rely on the Fourier transform. The advantage of MR STAT is that it allows relaxing constraints on the measurement process and also allows the use of acquisition plans that are time-efficient and widely available in clinical MR imaging scanners. The disadvantage is that the reconstruction procedure (which typically involves solving an inverse problem with about 105 unknowns) is computationally very complex and demanding.

[0006] Diffusion-weighted imaging (DWI) is an important MRI technique, e.g., for the detection of acute ischemic stroke or for the characterization and differentiation of brain tumors. In DWI, the apparent diffusion coefficient (ADC), which represents the magnitude of diffusion (of water molecules) in tissue, can be quantitatively evaluated by changing the amplitude and time of the magnetic field gradient, or TR.

[0007] However, known quantitative techniques such as MRF and MR STAT are not suitable to handle diffusion as an important MR contrast parameter, especially because the Extended Phase Graphs formalism (EPG, e.g. Journal of Magnetic Resonance Imaging 41:266-295, 2015 or MT Vlaardingerbroek and JA den Boer: Magnetic Resonance Imaging, Theory and Practice, 3. ed, Springer, 2002), used for the calculation of the necessary dictionary of predicted MR signal evolution patterns, presents several problems in the treatment of diffusion. The main drawback of the EPG approach is that the TR and the zero-order gradient moment k0 of the spoiled field gradient included in the imaging sequence must use fixed values ​​throughout the pulse train in order to have synchronous phase evolution of the switched and static field gradients (field inhomogeneity, susceptibility or T2* weighting), which ensures that the MR echo signals generated by both the switched and static field gradients are refocused simultaneously (hereafter referred to as pure coherence). Asynchronous phase evolution refocuses the MR echo signals generated by the switched and static field gradients at different times, which results in artifacts in the MR signal data due to superposition of different echo times (hereafter referred to as impure coherence). On the other hand, using a fixed k0 throughout the sequence makes it difficult to distinguish between diffusion and T2 decay in the evolution of incoherent MR signals, since both lead to decay of transverse magnetization. Diffusion mapping requires large variations in k0 in order to properly encode diffusion in the data evolution of acquired MR signals.

[0008] Against this background, it is readily appreciated that there is a need for improved quantitative MR imaging techniques. It is an object of the present invention to enable quantitative diffusion MR imaging based on the generation of temporally incoherent MR signal evolution by varying acquisition parameters (as in MRF or MR STAT).

[0009] [Means for solving the problem]

[0010] According to the invention, a method for MR imaging of an object placed in an examination volume of an MR system is disclosed, comprising the steps of: subjecting the object to an imaging sequence including a train of sequence blocks, each sequence block having at least one RF pulse, at least one switched spoiling magnetic field gradient, at least one switched readout magnetic field gradient defining a k-space sampling pattern, and having a set of acquisition parameters associated therewith including the zeroth moment of the spoiling magnetic field gradient, acquiring MR signals while varying at least the zeroth moment of the spoiling magnetic field gradient and any one or more other acquisition parameters during the imaging sequence, and reconstructing at least one MR image, in which at least one diffusion coefficient and any one or more other MR parameters are calculated for several image locations from an incoherent time evolution of the acquired MR signals caused by the variation of the at least one acquisition parameter.

[0011] The inventive technique generally corresponds to the above-mentioned MR imaging techniques (MRF or MR STAT). MR signals are generated by a train of sequence blocks and varying acquisition parameters to generate characteristic incoherent MR signal evolutions (fingerprints) from which MR parameters are quantitatively derived for each image location. Instead of fixed spoiled field gradients as commonly applied in spoiled gradient echo imaging sequences, the present invention proposes a gradient echo sequence scheme with variable spoiled field gradients and any other variable acquisition parameters. The variation of the zeroth moment (k0) of the spoiled field gradient provides diffusion encoding in the generated incoherent time evolution of the acquired MR signal, and the inventive method allows diffusion mapping, since such diffusion is well distinguishable from other MR parameters (such as transverse relaxation time).

[0012] The insight is that it is possible to suppress the above-mentioned impure coherence tomography (by appropriately selecting the variation pattern of k0) so that the MR echo signals caused by the inhomogeneity of the static magnetic field are not completely refocused. Thus, the present invention allows to avoid undesirable artifacts in the acquired MR signal data due to the superposition of different echo times.

[0013] In a possible embodiment, the imaging sequence has two or more types of sequence blocks, for example type a and type b, and each type has a fixed value of k0 (k 0a and k 0b Each type is assigned a fixed repetition time TR a and TR b can be further allocated. The two (or more) types of sequence blocks can be arbitrarily interleaved.

[0014] In one embodiment, the fixed values ​​of k0 for different sequence block types are unit spoiled gradient values ​​multiplied by mutually prime integers. For example, k 0aand k 0b can be selected depending on the relationship: k 0a =n a k0, k 0b =n b k0

[0015] In the above formula, n a and n b are mutually prime positive integers (i.e., they share no divisors except 1), and k0 is a unit spoiled gradient moment large enough to dephase the MR signal (i.e., k0 should be large enough to induce a phase shift of the transverse magnetization of 2π or more). The latter condition ensures that there is no leakage of adjacent magnetization states into the acquired MR signal. 0a and k 0b By selecting at least n sequence blocks of type a, b repetitions, and n sequence blocks of type b a It can be seen that n iterations are needed before impure coherence refocusing occurs. The actual number also depends on the interleaving scheme of the different sequence block types. Therefore, for mutually prime numbers n a and n b By choosing a sufficiently large value for at least one of the two relatively prime numbers, the ratio of the two relatively prime numbers is preferably at least 5, more preferably at least 10, and most preferably at least 20. The suppression of the impure coherence is thus performed by a long train of sequence blocks, which this coherence undergoes before it has a chance to be refocused. A train of sequence blocks containing many RF pulses of different flip angles reduces the absolute strength of this coherence from RF pulse to RF pulse. The impure coherence tomography is therefore diluted and attenuated by relaxation and is therefore too weak to significantly degrade the acquired MR signal data. The relatively prime numbers n a and n bBoth the selection of TR and the selection of the interleaving pattern of the various sequence block types can be used to optimize the suppression of impurity coherence tomography. Note that the TR of two or more sequence block types can be selected arbitrarily. It is conceivable to use three or more different sequence block types.

[0016] In possible embodiments, other acquisition parameters that are changed during the imaging sequence are one or more of the following: repetition time, echo time, flip angle, RF pulse phase and / or frequency, k-space sampling pattern, readout field gradient and / or longitudinal magnetization preparation. As with conventional MRF techniques, an MR signal acquisition scheme is applied that makes the MR signals from different materials spatially and temporally incoherent by continuously varying the acquisition parameters throughout the data acquisition process. The aforementioned acquisition parameters can be used for this purpose in addition to the variation of the zeroth moment of the spoiling field gradient as described above. The acquisition parameters can be changed in a random, pseudorandom or other manner that makes the MR signals from different materials spatially incoherent, temporally incoherent, or both.

[0017] As mentioned above, the change of acquisition parameters achieves spatial incoherence, time incoherence, or both, by varying the acquisition parameters (including k0) from one sequence block to the next. This produces a time series of MR images with different contrasts. The MRF reconstruction process is designed to map any of a wide variety of MR parameters, such as the diffusion coefficient, and optionally the longitudinal relaxation time T1, the transverse relaxation time T2, the main or static magnetic field B0, the radio frequency magnetic field B1, and the spin (proton) density, as mentioned above. To achieve this, the time evolution of the acquired MR signal data is compared to a dictionary of MR signal evolution patterns previously generated for different acquisition parameters, based on a physical simulation based on the Bloch equations, or preferably on an MR signal model such as the extended phase graph (EPG) format (see above), which is computationally much simpler and therefore more efficient. This comparison allows for an estimation of the MR parameters of interest. In general, the MR parameters at a given image location are estimated to be those that provide the best match between the acquired MR signal evolution and the predicted MR signal evolution pattern. For this purpose, conventional pattern matching algorithms can be used.

[0018] In one embodiment of the present invention, at least some RF pulses of a sequence block are chemical shift selective, which can be used, for example, to achieve fat suppression.

[0019] The method of the invention as described above can be performed by an MR system having at least one main magnet coil for generating a main magnetic field in an examination volume, a number of gradient coils for generating switched magnetic field gradients in different spatial directions in the examination volume, at least one RF coil for generating RF pulses in the examination volume and / or receiving MR signals from a subject positioned in the examination volume, a control computer for controlling the time sequence of the RF pulses and the switched magnetic field gradients based on an examination protocol, and a reconstruction unit for reconstructing MR images from the received MR signals.

[0020] The method of the invention can for example be implemented by corresponding programs in a reconstruction unit and / or in a control unit of an MR system.

[0021] The method of the invention can be advantageously implemented in most MR systems currently in clinical use. For this purpose, it is only necessary to utilize a computer program which controls the MR system to execute the above-mentioned method steps of the invention. The computer program may be present on a data carrier or in a data network, to be downloaded for installation in the control unit of the MR system.

[0022] The accompanying drawings disclose preferred embodiments of the present invention, however, it is to be understood that the drawings are designed for illustrative purposes only and not as a definition of the limits of the invention. [Brief description of the drawings]

[0023] [Figure 1] FIG. 1 shows an MR system for implementing the method of the present invention. [Diagram 2] FIG. 2 shows a schematic diagram of an example of an imaging sequence for use in accordance with the present invention; [Diagram 3] FIG. 1 shows phase graphs of different coherence tomography pathways of the MR signal for an imaging sequence with alternating spoiled magnetic field gradients. [Figure 4] FIG. 1 shows a schematic of an MRF sequence design with alternating spoiled magnetic field gradients and a corresponding dictionary for distinguishing fat and water. [Diagram 5] FIG. 1 shows MRF T1, T2 and ADC maps obtained from water and oil phantoms using the method of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0024] With reference to Figure 1, there is shown an MR system 1. The system comprises superconducting or resistive main magnet coils 2 such that a substantially uniform and temporally constant main magnetic field is generated along the z-axis through an examination volume.

[0025] The magnetic resonance generation and manipulation system applies a series of RF pulses and switching magnetic field gradients to invert or excite nuclear magnetic spins, induce magnetic resonance, refocus magnetic resonance, manipulate magnetic resonance, spatially and otherwise encode magnetic resonance, saturate spins, and the like to perform MR imaging.

[0026] More specifically, gradient pulse amplifiers 3 apply current pulses to selected ones of whole-body gradient coils 4, 5 and 6 along the x, y and z axes of the examination volume. A digital RF frequency transmitter 7 transmits RF pulses or pulse packets to the whole-body volume RF coil 9 via a transmit / receive switch 8 to transmit RF pulses into the examination volume. A typical MR imaging sequence is composed of packets of short duration RF pulse segments which, taken together with each other, together with the applied magnetic field gradients, achieve selected manipulation of nuclear magnetic resonance. The RF pulses are used to saturate resonance, excite resonance, invert magnetization, refocus resonance or manipulate resonance to select portions of the subject (patient's body) 10 located within the examination volume.

[0027] For the production of MR images of limited regions of the body 10 by parallel imaging, a set of local RF coils 11, 12, 13 are positioned adjacent to the region selected for imaging. The resulting MR signals are picked up by the RF coils 11, 12, 13 and demodulated by a receiver 14, preferably having a preamplifier (not shown). The receiver 14 is connected to the RF coils 9, 11, 12 and 13 via a transmit / receive switch 8.

[0028] A host computer 15 controls the gradient pulse amplifiers 3 and the transmitter 7 to generate any of a number of MR image sequences, such as echo planar images, echo volume images, gradient and spin echo images, fast spin echo images, etc. For the selected sequence, the receiver 14 receives single or multiple MR data lines in rapid succession following each RF excitation pulse. A data acquisition system 16 performs analog-to-digital conversion of the received signals, converting each MR data line into a digital form suitable for further processing. In modern MR systems, the data acquisition system 16 is a separate computer, or in some cases a remote server "in the cloud" dedicated to collecting the raw image data. Finally, the digital raw image data is reconstructed into an image representation by a reconstruction processor 17 that applies a Fourier transform or other suitable reconstruction algorithm. The MR image may represent a planar slice across the patient, an array of parallel planar slices, a three-dimensional volume, etc. The image is then stored in an image memory, which can be accessed to convert slices, projections, or other portions of the image representation into a suitable format for visualization, for example, via a video monitor 18 that provides a human-readable display of the resulting MR image.

[0029] With continuing reference to FIG. 1, and further with reference to FIGS. 2-5, an embodiment of the imaging approach of the present invention will now be described.

[0030] As mentioned above, MR signals are generated by a train of sequence blocks of a spoiled gradient echo imaging sequence and varying acquisition parameters to generate a characteristic incoherent MR signal evolution (fingerprint) from which maps of diffusion coefficients and further MR parameters can be derived. The present invention uses a gradient echo sequence scheme with variable spoiled field gradients to enable diffusion mapping, whereby contributions from impure coherence tomography are suppressed and MR echo signals caused by static magnetic field inhomogeneities are not fully refocused.

[0031] In one embodiment of the present invention, an imaging sequence is composed of two types of sequence blocks, type a and type b, and each type is assigned a fixed value of the zeroth moment of the spoiling magnetic field gradient applied to each type of sequence block. The two fixed values ​​are k 0a and k 0b It is called.

[0032] In the embodiment shown in FIG. 2, the first four sequence blocks of the gradient echo sequence used are shown, where k a = k0 and k b =k0+Δk, where TR a = TR1 and TR b =TR2.

[0033] In the embodiment shown in FIG. 3, two types of sequence blocks are applied alternately. 0a and k 0b is the unit spoiled gradient k0 multiplied by a relatively prime integer: k 0a =n a k0, k 0b =n b k0 In the above formula, n a and n b are relatively prime positive integers, and k0 is a unit spoiled gradient moment large enough to dephase the magnetic resonance signal. a=2 and n b The combination of =11 is used. Figure 3 shows the alternating gradient moment (n a / n b Figure 1 shows simulated phase diagrams (φ, as a function of time) for visualizing the various coherence tomography pathways for a gradient echo sequence with TR = 2 / 11, but for simplicity, using a fixed TR for the two sequence block types. Coherence tomography pathways resulting in impure coherence tomography (curves a, b) and pure coherence tomography (curves c, d, e, f) are illustrated. The first impure coherence tomography is refocused (φ = 0) up to a number of 21 sequence blocks. In contrast, pure coherence tomography occurs already after three sequence blocks (stimulated echo, curves c and e) and four sequence blocks (spin echo, curves d and f). Since 21 sequence blocks are required to generate the first impure coherence tomography, but only three sequence blocks are required to generate the pure coherence tomography, the undesired impure coherence tomography is diluted and attenuated by relaxation and is therefore too weak to degrade the acquired MR signal data. 0a and k 0b and an integer n a and n b can each be influenced by the sequence programmer depending on the purpose, but the unknown main field inhomogeneities cannot contribute to the accumulated phase per TR either. It is therefore important to note that in the case of pure coherence tomography, the net contribution of all these main field inhomogeneity induced phases cancels out in the echo, but in the case of impure coherence tomography, these additional phases do not disappear and result in a non-zero phase at the time of gradient-induced echo formation. This effect results in a further dephasing of the impure signal contributions from the impure coherence tomography, which are already diluted in the long pathways.

[0034] Experiments were performed on a 1.5T MRI scanner (Philips Ingenia) with phantom bottles of water (phantom liquid 13) and oil (Spectrasyn 4). For the MRF according to the invention, a spoiled interleaved spiral imaging sequence was used to map T1, T2 and ADC (1 mm resolution, 8 ms acquisition window, 3 shots with 800 sequence blocks per shot, repetition time TR = 24 ms, undersampling factor 50 / 3). The shots were preceded by an inversion RF pulse and had an optimized flip angle variation. The top diagram of Figure 4 shows the evolution of the flip angle (FA) during the imaging sequence to improve the sensitivity of the quantitative mapping. In the (approximate) first half of the train of sequence blocks, a small spoiled magnetic field gradient (b value, approximately 0.2 s / mm 2 ) was used, while in the second half of the MRF sequence, strong spoiled field gradients (b-values, approximately 90 s / mm 2 ) were used, corresponding to relatively prime factors of 3 and 59, respectively. 2 ) was used. In the experiment, the first impure coherence tomography is refocused by the first pulse of the second half of the train of the sequence block to a spin echo in the fourth pulse of the second half. Apart from the dilution of the coherence pathway (among many pathways that do not generate impure coherence, only one pathway generates impure coherence) and relaxation decay, the attenuation of T2* due to the long echo time of more than 1 second also contributes to the efficient suppression of this coherence. For the adopted sequence, dictionaries were calculated for a range of T1, T2, and ADC values ​​using the EPG format. The corresponding fingerprints of water and oil reflect the different diffusion constants of water and oil. The lower diagram in Figure 4 shows two fingerprints of water and oil reflecting the different diffusion coefficients of water and oil.

[0035] From the dictionary and undersampled spiral images, T1, T2, and ADC maps were fitted using a custom MRF simulation environment. The resulting maps are shown in Figure 5. These maps are in good agreement with the expected values. In particular, we were able to successfully reveal low and high ADC values ​​for oil and water, respectively. The results confirm that the variable spoil field gradient allows for the differentiation of T2 and diffusion.

Claims

1. 1. A method of MR imaging of an object disposed within an examination volume of an MR system, comprising: subjecting a subject to an imaging sequence including a train of sequence blocks, each sequence block having at least one RF pulse, at least one switched spoiling magnetic field gradient, and at least one switched readout magnetic field gradient defining a k-space sampling pattern, and having associated therewith a set of acquisition parameters having a zeroth moment of the spoiling magnetic field gradient, the imaging sequence having two or more types of sequence blocks, each type being assigned a fixed value of the zeroth moment of the spoiling magnetic field gradient, the fixed value being a unit spoiling gradient value multiplied by a pair of relatively prime integers; acquiring MR signals while varying at least the zeroth moment of the spoiling field gradient and any one or more other acquisition parameters during the imaging sequence; reconstructing at least one MR image, wherein at least one diffusion coefficient and optionally one or more other MR parameters are calculated for a plurality of image locations from an incoherent time evolution of the acquired MR signal caused by variations in the at least one other acquisition parameter; A method having the following.

2. The method of claim 1 , wherein the ratio of two of the fixed values ​​is at least 5, at least 10, or at least 20.

3. 3. The method of claim 1, wherein the other acquisition parameters that are varied during the imaging sequence are one or more of a repetition time, an echo time, a flip angle, an RF pulse phase and / or frequency, a k-space sampling pattern, a readout field gradient, and a longitudinal magnetization preparation.

4. The one or more other MR parameters calculated from the time evolution of the acquired MR signals are T １ , T ２ , spin density, main magnetic field B ０ , RF magnetic field B １ 3. The method of claim 1 or 2, wherein the method is one or more of:

5. 3. The method of claim 1, wherein at least some RF pulses in the sequence block are chemical shift selective.

6. 3. The method of claim 1, wherein the diffusion coefficient and the one or more other MR parameters are calculated by matching the time evolution of the acquired MR signal at a given image location against a set of predicted MR signal evolution patterns.

7. The method of claim 6 , wherein the predicted MR signal evolution pattern is calculated using an extended phase graph.

8. 1. An MR system comprising: at least one main magnet coil for generating a main magnetic field in an examination volume; gradient coils for generating switching magnetic field gradients in different spatial directions in the examination volume; at least one RF coil for generating RF pulses in the examination volume and / or receiving MR signals from a subject positioned in the examination volume; a control computer for controlling the time sequence of RF pulses and switching magnetic field gradients based on an examination protocol; and a reconstruction unit for reconstructing MR images from the received MR signals, wherein the MR system comprises: subjecting the subject to an imaging sequence including a train of sequence blocks, each sequence block having at least one RF pulse, at least one switched spoiling magnetic field gradient, and at least one switched readout magnetic field gradient defining a k-space sampling pattern, and having a set of acquisition parameters associated with the sequence block including a zeroth moment of the spoiling magnetic field gradient, the imaging sequence having two or more types of sequence blocks, each type assigned a fixed value of the zeroth moment of the spoiling magnetic field gradient, the fixed value being a unit spoiling gradient value multiplied by a pair of relatively prime integers; acquiring MR signals while varying at least the zeroth moment of the spoiling field gradient and any one or more other acquisition parameters during the imaging sequence; reconstructing at least one MR image, wherein at least one diffusion coefficient and optionally one or more other MR parameters are calculated for a plurality of image locations from an incoherent time evolution of the acquired MR signals caused by variations in the at least one acquisition parameter; An MR system that performs

9. instructions for executing an imaging sequence including a train of sequence blocks, each sequence block having at least one RF pulse, at least one switched spoiling magnetic field gradient, and at least one switched readout magnetic field gradient defining a k-space sampling pattern, and having a set of acquisition parameters associated with the sequence block including a zeroth moment of the spoiling magnetic field gradient, the imaging sequence having two or more types of sequence blocks, each type assigned a fixed value of the zeroth moment of the spoiling magnetic field gradient, the fixed value being a unit spoiling gradient value multiplied by a pair of relatively prime integers; instructions to acquire MR signals during the imaging sequence while varying at least the zeroth moment of the spoiling magnetic field gradient and optionally one or more other acquisition parameters; A computer program having the following:

10. 10. The computer program of claim 9, further comprising instructions for reconstructing at least one MR image, wherein at least one diffusion coefficient and optionally one or more other MR parameters are calculated for several image locations from incoherent time evolution of the acquired MR signals caused by the variation of the at least one acquisition parameter.

Images