Method and magnetic resonance apparatus for correction of a B0 map for chemical shifts

Abstract

In a method for correction of a B0 field map measured with a magnetic resonance device, that describes deviations from a nominal field strength in the homogeneity area of the magnetic resonance device by deviations from a nominal frequency for protons bonded to water, the deviations being represented as Larmor frequency values for different picture elements shifted by chemical shifts, the B0 field map is recorded with spins of the fat and water protons not in phase. The B0 field map is segmented by evaluating the differences of the Larmor frequency values of adjacent picture elements of the B0 field map in at least two contiguous clusters. For each cluster, a decision is made on the basis of a smoothness criterion and a compactness criterion as to whether a cluster containing a majority of protons bonded into fat is involved. Clusters identified as containing a majority of protons bonded into fat are corrected by lowering the Larmor frequency values by the difference between the nominal frequency for protons bonded into water and the corresponding nominal frequency for protons bonded to fat.

Claims

1. A method for correction of a B0 field map measured with a magnetic resonance device in which an examination object is situated, describing deviations from a nominal field strength in a homogeneity area of the magnetic resonance device by deviations from a nominal frequency for protons bonded into water, the deviations being represented as Larmor frequency values for different picture elements resulting from chemical shifts of the Larmor frequencies, wherein the B0 field map was recorded at least in part with spins of fat and water protons in the examination object not in phase, said method comprising: segmenting the B0 field map according to a segmentation criterion that evaluates the differences of the Larmor frequency values of adjacent picture elements of the B0 field map in at least two contiguous clusters, each containing at least two picture elements, that are separated by a jump in the Larmor frequency values; for each cluster, determining, based on a smoothness criterion and a compactness criterion, whether the respective cluster contains a majority of protons bonded into fat; and correcting clusters identified as containing a majority of protons bonded into fat by lowering the Larmor frequency values thereof by a difference between the nominal frequency for protons bonded into water and the corresponding nominal frequency for protons bonded into fat.

2. The method as claimed in claim 1, comprising using a graph-based segmentation algorithm for the segmentation.

3. The method as claimed in claim 1, comprising identifying a water reference cluster before the application of the smoothness criterion and the compactness criterion.

4. The method as claimed in claim 3, comprising determining the water reference cluster as a cluster showing the smallest deviation from the nominal Larmor frequency.

5. The method as claimed in claim 4 comprising determining the water reference cluster as a cluster showing the smallest deviation from the nominal Larmor frequency among a set of the segmented clusters comprising between five and ten of the largest segmented clusters.

6. The method as claimed in claim 4, comprising forming the median of the Larmor frequency values of the respective cluster for determining the smallest deviation from the nominal Larmor frequency.

7. The method as claimed in claim 3, comprising using a smoothness of the B0 field map for clusters corrected in testing in relation to the chemical shift as the smoothness criterion for each cluster except for the reference cluster with the smoothness without correction and, when an improved smoothness is established for corrected clusters, accepting the correction and classifying the cluster as a cluster containing a majority of protons bonded into fat.

8. The method as claimed in claim 1, comprising, as the compactness criterion for each cluster not classified by the smoothness criterion as containing a majority of protons bonded into fat and, if a reference cluster has been defined, as not corresponding to the reference cluster, checking whether an average Larmor frequency value of that cluster is within or outside a permitted interval around an average Larmor frequency value determined for all clusters of the group lying within a region of interest around the cluster to be checked, which contains all clusters classified by the smoothness criterion as containing a majority of protons bonded into fat and, if a reference cluster has been defined, contains the reference cluster, wherein for an average Larmor frequency value lying outside the interval, the checked cluster is classified as containing a majority of protons bonded into fat and corrected.

9. The method as claimed in claim 7, comprising selecting a slice of the B0 field map containing the cluster to be checked as a region of interest.

10. The method as claimed in claim 1, comprising determining a fat mask from the clusters classified as containing a majority of protons bonded into fat and using said fat mask in at least one later recording and/or evaluation of magnetic resonance data of the same examination object.

11. The method as claimed in claim 10 comprising using said fat mask as a starting point for a Dixon technique for acquiring magnetic resonance data from said examination object.

12. A method as claimed in claim 10 comprising using said fat mask in an acquisition of magnetic resonance spectroscopic data from said examination object.

13. A magnetic resonance apparatus comprising: a magnetic resonance data acquisition scanner in which an examination subject is situated; a control computer configured to operate said magnetic resonance data acquisition scanner to acquire a B0 field map describing deviations from a nominal field strength in a homogeneity area of the magnetic resonance data acquisition scanner, by deviations from a nominal frequency for protons bonded into water, the deviations being represented as Larmor frequency values for different picture elements resulting from chemical shifts of the Larmor frequencies, with said B0 field map being acquired at least in part with spins of fat and water protons in said examination subject not being in phase; a processing computer provided with said B0 field map, said processing computer being configured to segment the B0 field map according to a segmentation criterion by evaluating differences of the Larmor frequency values of adjacent picture elements of the B0 field map in at least two contiguous clusters, each containing at least two picture elements, that are separated by a jump in the Larmor frequency values; in said processing computer, for each cluster, determining, based on a smoothness criterion and a compactness criterion, whether the respective cluster contains a majority of protons bonded into fat; in said processing computer, correcting clusters identified as containing a majority of protons bonded into fat by lowering the Larmor frequency thereof by a difference between the nominal frequency for protons bonded into water and the corresponding nominal frequency for protons bonded into fat; and generating a corrected B0 field map from all corrected clusters, and making the corrected B0 field map available in electronic form as a data file at an output of said processing computer.

14. A non-transitory, computer-readable data storage medium encoded with programming instructions, said storage medium being loaded into a computer of a magnetic resonance apparatus and said programming instructions causing said computer to: receive a B0 field map, for correction of said B0 field map, acquired by operation of said magnetic resonance apparatus with an examination object situated therein, said B0 field map describing deviations from a nominal field strength of a homogeneity area of the magnetic resonance apparatus by deviations from a nominal frequency for protons bonded into water, the deviations being represented as Larmor frequency values for different picture elements resulting from chemical shifts of the Larmor frequencies, said B0 field map having been acquired at least in part with spins of fat and water protons in the examination object not being in phase; segment the B0 field map according to a segmentation criterion that evaluates the differences of the Larmor frequency values of adjacent picture elements of the B0 field map in at least two contiguous clusters, each containing at least two picture elements, that are separated by a jump in the Larmor frequency values; for each cluster, determine, based on a smoothness criterion and a compactness criterion, whether the respective cluster contains a majority of protons bonded into fat; and correct clusters identified as containing a majority of protons bonded into fat by lowering the Larmor frequency values thereof by a difference between the nominal frequency for protons bonded into water and the corresponding nominal frequency for protons bonded into fat.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) FIG. 1 shows a flowchart of the inventive method.

(2) FIG. 2 illustrates correction by use of the smoothness criterion in accordance with the invention.

(3) FIG. 3 shows the option of correction by use of the compactness criterion in accordance with the invention.

(4) FIG. 4 shows the effects of correction in accordance with the invention.

(5) FIG. 5 schematically illustrates a magnetic resonance device in accordance with the invention.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

(6) The starting point for the inventive method, of which a flowchart is shown in FIG. 1, is a B0 field map 1, which contains artifacts resulting from chemical shifts between protons bonded into fat and protons bonded into water. The B0 field map 1 has been recorded in the present example in a multi-echo method, thus a method which uses different dephasing times and is thus not capable of insuring that fat signals and water signals are in phase. Therefore artifacts are produced as a result of chemical shift. These are to be identified and corrected on the basis of the correction method presented here. In this case the method is based on the knowledge that B0 field inhomogeneities, which the B0 field map 1 is designed to map, vary slowly spatially, thus exhibit a high degree of smoothness. By contrast artifacts of chemical shift are characterized by sudden detached areas.

(7) It should be noted in advance that areas without a signal in which thus only noise components are present can of course easily be recognized in a B0 field map and can be excluded from consideration in the present method.

(8) In a first step S1 the portions (containing a signal) of the B0 field map 1 are segmented into N disjoint, contiguous clusters, which are to be labeled C.sub.n below, wherein n runs from 1 to N. Expressed schematically in formulas, wherein B0 symbolizes the B0 field map 1, this means that
B0=Σ.sub.n=1.sup.NC.sub.n

(9) For segmentation a graph-based segmentation algorithm reacting extremely sensitively to jumps is employed, which assesses neighboring relationships between the individual picture elements of the B0 field map 1 as part of a segmentation criterion. In this way it can be established whether picture elements or regions belong to a common cluster or not. The clusters are thus separated from one another by jumps in the deviation from the nominal Larmor frequency for protons bonded into water for Larmor frequency values describing the magnetic resonance device, as is usual for segmentation methods, wherein it is entirely conceivable for a slow, smooth change of the Larmor frequency value to also occur within a cluster. In general, however, a cluster herein is a group of picture elements that, at least in adjacent areas, exhibit a comparable Larmor frequency value.

(10) The clusters ultimately represent candidates for fat clusters, i.e. those which primarily contain signals from protons bonded into fat. Whether a cluster is a fat cluster is checked in the following steps on the basis of two different criteria, namely a smoothness criterion, which is more sensitive than the segmentation carried out in step S1, and a compactness criterion. Clusters identified as fat clusters are corrected by subtraction of the chemical shift between water and fat.

(11) Initially in step S2, a reference cluster is defined that is innately a water cluster, thus reproduces signals of a region of the examination volume in which there are more protons bonded into water than protons bonded into fat present. To do this, the clusters found in step S1 are arranged in ascending order according to their size and in this description only the seven largest clusters will be considered further. Since the Larmor frequency values are selected here so that a Larmor frequency value of 0 describes no deviation from the specified Larmor frequency of the magnetic resonance device for protons bonded to water, that cluster among the largest clusters that exhibits the smallest average Larmor frequency value is selected. For this purpose, the median is selected as the median value of the Larmor frequency values, since this is especially robust with respect to noise effects. Thus
C.sup.ref=arg.sub.i min(median(C.sub.i−0))

(12) The reason for this is that, in the examination volume recorded in the B0 field map 1, as a rule a contiguous water region is always to be found among the four largest clusters. The smoothness is now optimized in the following step S3 in relation to the reference cluster C.sub.ref.

(13) In this step each cluster C.sub.n, except for the reference cluster C.sub.ref, is now corrected individually for a possible chemical shift. If the overall smoothness of the B0 field map 1 is increased by this step, it is to be assumed that the cluster is a fat cluster, so that it is marked accordingly and the correction is retained. If the smoothness is not increased the step continues with the next cluster. This method is explained in greater detail schematically by the diagram in FIG. 2. This diagram shows an example of a slice 2 of the B0 field map 1 in which for purposes of presentation only a few clusters 3, 4, 5 and 6 are shown here. Areas without clusters in slice 2 indicate no signal was present here, especially since air or the like existed there.

(14) In this example, cluster 5 represents the reference cluster, from which the other clusters 3, 4 and 6 have deviating average Larmor frequency values, which is represented by the different shadings in the part image on the left. The arrow 7 symbolizes the correction of the cluster 6 for a possible chemical shift. In the part image 2 on the right it can be seen, as is represented by the removal of the shading 6, that the Larmor frequency value of the cluster 6 now essentially corresponds to the Larmor frequency value of the cluster 5, a smooth transition is produced. The cluster 6 is thus a fat cluster, the correction is retained. The improved smoothness is clearly to be seen in the right-hand part image of FIG. 2.

(15) This method of operation means:
cost.sub.old=Σabs(∇B0)

(16) For each C.sub.n, which does not correspond to the reference cluster C.sub.ref, do:
cost.sub.new=Σabs(∇(B0|C.sub.n+(C.sub.n−Δω.sub.FW))),
wherein, when cost.sub.new≦cost.sub.old:
cost.sub.old=cost.sub.new
B=B0|C.sub.n+(C.sub.n−Δω.sub.FW)

(17) In this case, cost symbolizes the smoothness able to be expressed as a cost function, B0 the B0 field map 1, Δω.sub.FW the chemical shift between protons bonded into fat and into water at the nominal Larmor frequency and B0|C.sub.n symbolically B0 without the current cluster C.sub.n to be checked (in FIG. 2 the cluster 6).

(18) While it is already possible with this step to identify and to correct the greater part of the fat clusters, fat clusters remain that are spatially isolated or have a high noise level, since these can only be identified with difficulty by the smoothness criterion in step S3. This is to be illustrated in greater detail by the basic diagram of FIG. 3, which shows a further slice 2′ of the B0 field map 1, in which once again, for the sake of simplicity, only a few clusters 8, 9, 10, 11 are shown. Evidently the cluster 11 is isolated from the clusters 8, 9 and 10, since a region without a signal is located between them, so that in a correction of the cluster 11 by means of testing for no influence or only an extremely low influence on the smoothness of the B0 field map 1 occurs. If however for example the slice 2′, in which the cluster 11 lies, is viewed as a region of interest, in which a compactness of the distribution of the Larmor frequency values can be used as a starting point, and if the clusters 9 and 10 have also already been identified in step S3 as fat clusters, it can be assumed that the corrected version of the clusters 9 and 10 are water clusters, thus contain usual Larmor frequency values in protons bonded into water for the region of interest. A compactness of the distribution of the Larmor frequency values would now mean that the average Larmor frequency value of the cluster 11 might not deviate too greatly from the average Larmor frequency value of the cluster 9 and 10, so that the cluster 11 can also be identified as a water cluster.

(19) Thus in step S4, the average Larmor frequency values are initially determined for the cluster 11, but also for the corrected fat clusters 9 and 10, which are water clusters in their corrected form, wherein it should be pointed out that of course the reference cluster is basically viewed and considered as a water cluster if it is contained in the region of interest. Only when the average Larmor frequency value of the cluster 11 to be examined within an interval lies around the average Larmor frequency value of the water cluster in the region of interest, is it to be assumed that the cluster 11 is a water cluster, otherwise it is classified as a fat cluster 11 and subjected to a corresponding correction. Again expressed schematically in formulas this means

(20) $|\stackrel{\_}{C_n}{|}_{\mathrm{ROI}}<f*|\stackrel{\_}{\underset{i\in \mathrm{Water}}{.Math.}{C}_i}{|}_{\mathrm{ROI}},$
wherein the horizontal line over the clusters indicates the averaging of the Larmor frequency values and f represents the factor which defines the interval. This is to be selected so that the interval, although it extends into the area of the chemical shift between water and fat, naturally does not reach this area, for example up to 60-80% of the value for the chemical shift between protons bonded into water and protons bonded into fat.

(21) The result of the inventive method is then the corrected B0 field map 1″ and also, since the identified fat clusters have been marked, a fat mask 12 which can be employed to advantage in further applications.

(22) FIG. 4 again explains schematically the effects of the inventive method on the basis of a slice 2″ of the B0 field map 1, which forms the left-hand part image of FIG. 4. As is shown by the shading, markedly different, clearly separated regions 13, 14 of Larmor frequency values exist in this diagram, as is shown by the dotted/hatched areas. The execution of the inventive method, arrow 15, leads to a correction of this chemical shift and to a contiguous smooth region 16 of Larmor frequency values, which reproduce the B0 inhomogeneities, as is shown in the slice 2″ of the corrected B0 field map 1′, i.e. the right-hand part image.

(23) Finally FIG. 5 shows a block diagram of an inventive magnetic resonance device 17, which, as is basically known, has a basic field magnet unit 18 with magnets generating the basic magnetic field B0. The basic field magnet unit 18 defines a patient receptacle 19, which is surrounded (not shown in greater detail) by the gradient coil arrangement and the radio frequency coil arrangement, as is basically known.

(24) The components of the magnetic resonance device 17 are controlled by a control device 20, which is also designed to implement the inventive method. In particular, a computer program represented by code on a storage medium can run in the control device 20, when the storage medium is loaded into the control device 20.

(25) Although modifications and changes may be suggested by those skilled in the art, it is the intention of the inventors to embody within the patent warranted hereon all changes and modifications as reasonably and properly come within the scope of their contribution to the art.