METHOD AND COMPUTER FOR AUTOMATIC CHARACTERIZATION OF LIVER TISSUE FROM MAGNETIC RESONANCE IMAGES

Abstract

In a computer and a magnetic resonance method and apparatus for automatic characterization (classification) of liver tissue in a region of interest of a liver, at least one value tuple of the region of interest of the liver is acquired, the value tuple including at least one T.sub.1 value determined from magnetic resonance images of the region of interest, or a reciprocal value thereof, and a T.sub.2 or T.sub.2* value or a reciprocal value thereof. The value tuple is transferred into a multidimensional parameter space and the characterization of the liver tissue is then performed on the basis of the position of the value tuple in the parameter space.

Claims

1. A method for automatic characterization of liver tissue in a region of interest of a liver, comprising: providing a processor with at least one value tuple of a region of interest of the liver of a subject, said value tuple comprising at least one T.sub.1 value deteiniined from a magnetic resonance image of the region of interest, or a reciprocal of said at least one T.sub.1 value, and a T.sub.2 value or a reciprocal thereof, and a T.sub.2* value or a reciprocal value thereof; in said computer, transferring the value tuple into a multi-dimensional parameter space; in said computer, generating a characterization of the liver tissue from a position of the value tuple in the parameter space; and making an electronic signal representing said characterization of the liver available as an output from the processor.

2. A method as claimed in claim 1 comprising, in said processor, characterizing the liver tissue based on the position of the value tuple in the parameter space with reference to a boundary hyperplane of the parameter space.

3. A method as claimed in claim 1 comprising characterizing the liver tissue on the basis of the position of the value tuple in the parameter space with reference to at least one cluster of reference value tuples in the parameter space.

4. A method as claimed in claim 1 comprising providing the processor with a plurality of value tuples of the region of interest of the liver, and transferring the plurality of value tuples into the multi-dimensional parameter space as a value-tuple group, and characterizing the liver tissue based on a collective position of the value-tuple group in the parameter space.

5. A method as claimed in claim 1 wherein said region of interest includes an entirety of the liver.

6. A method as claimed in claim 1 wherein said value tuple comprises at least one further value.

7. A method as claimed in claim 6 wherein said at least one further value is a fat value.

8. A method as claimed in claim 6 wherein said at least one further value is a liver texture value.

9. A method as claimed in claim 1 comprising providing the processor with said at least one value tuple by: acquiring magnetic resonance data from said region of interest of the liver; reconstructing an image of the liver from said magnetic resonance data, said image comprising voxels in said region of interest of the liver; and determining said T.sub.1 value or the reciprocal value thereof, said T.sub.2 value or the reciprocal value thereof, or said T.sub.2* value or said reciprocal value thereof, from said voxels in said region of interest of the liver in said image.

10. A characterization computer for automatic characterization of liver tissue in a region of interest of a liver of a subject, said computer comprising: an input interface that receives a magnetic resonance image of the liver tissue in said region of interest; a processor configured to determine at least one value tuple from said region of interest, said value tuple comprising at least one Ti value or a reciprocal of said at least one T.sub.1 value, and a T.sub.2 value or a reciprocal thereof, and a T.sub.2* value or a reciprocal value thereof; said processor being configured to transfer the value tuple into a multi-dimensional parameter space; said processor being configured to generate a characterization of the liver tissue from a position of the value tuple in the parameter space; and said processor being configured to make an electronic signal representing said characterization of the liver available as an output from the processor.

11. A magnetic resonance (MR) apparatus comprising: an MR data acquisition scanner; a control computer configured to operate the MR data acquisition scanner to acquire MR data from a region of interest of a liver of a subject; a reconstruction computer configured to reconstruct an image of the region of interest of the liver from said MR data; a processor configured to acquire at least one value tuple of the region of interest of the liver that includes at least one T.sub.1 value determined from said image, or a reciprocal of said at least one T.sub.1 value, and one of a T.sub.2 value determined from the image, or a reciprocal of said T.sub.2 value, or a T.sub.2* value determined from said image, or a reciprocal of said T.sub.2* value; said processor being configured to transfer the value tuple into a multi-dimensional parameter space; said processor being configured to characterize the liver tissue in said region of interest based on a position of the value tuple in said parameter space; and said processor being configured to generate an electronic signal representing said characterization of the liver tissue in the region of interest and to emit said electrical signal from said processor.

12. A non-transitory, computer-readable data storage medium encoded with programming instructions, said storage medium being loaded into a computer of a magnetic resonance (MR) apparatus, and said programming instructions causing said computer to: determine at least one value tuple of a region of interest of the liver of a subject, said value tuple comprising at least one T.sub.1 value determined from a magnetic resonance image of the region of interest, or a reciprocal of said at least one T.sub.1 value, and a T.sub.2 value or a reciprocal thereof, and a T.sub.2* value or a reciprocal value thereof; transfer the value tuple into a multi-dimensional parameter space; generate a characterization of the liver tissue from a position of the value tuple in the parameter space; and make an electronic signal representing said characterization of the liver available as an output from the computer.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

[0035] FIG. 1 schematically illustrates a magnetic resonance imaging system according to one exemplary embodiment of the invention.

[0036] FIG. 2 schematically shows a liver for determining the region of interest.

[0037] FIG. 3 shows an example of a possible classification in a two-dimensional parameter space.

[0038] FIG. 4 shows an example of a possible classification in a three-dimensional parameter space.

[0039] FIG. 5 shows a further example of a possible classification in a three-dimensional parameter space.

DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0040] FIG. 1 is a schematic illustration of a magnetic resonance imaging apparatus 1. It includes the actual magnetic resonance scanner 2 with an examination chamber 3 or patient tunnel in which a patient or test subject is positioned on a bed 8. The actual examination object O, the liver, is located in the body of the patient or test subject.

[0041] The magnetic resonance scanner 2 is typically equipped with a basic field magnet 4, a gradient coil arrangement 6 and an RF transmission antenna 5 and a RF reception antenna 7. In the exemplary embodiment depicted, the RF transmission antenna system 5 is a whole-body coil permanently installed in the magnetic resonance scanner 2, the RF reception antenna 7 is formed by local coils to be arranged on the patient or test subject (in FIG. 1 only symbolized by an individual local coil). In principle, however, it is also possible for the whole-body coil to be used as an RF reception antenna and the local coils as an RF transmission antenna as long as these coils can in each case be switched to different operating modes. The basic field magnet 4 is typically designed so as to generate a basic magnetic field in the longitudinal direction of the patient, i.e. along the longitudinal axis of the magnetic resonance scanner 2 extending in the z direction. The gradient coil arrangement 6 typically includes individually controllable gradient coils in order to be able to activate gradients in the in x, y or z directions independently of one another. The magnetic resonance scanner 2 also contains shim coils (not shown), which can be embodied conventionally.

[0042] The MR scanner 2 shown in FIG. 1 is a whole-body system with a patient tunnel into which a patient can be completely introduced, In principle, however, the invention can also be used in other MR systems, for example with a laterally open C-shaped housing. The only essential factor is that it is possible to prepare appropriate images of the liver O.

[0043] The MR apparatus 1 furthermore comprises a central control computer 13 that is used to control the MR apparatus 1. This central control computer 13 includes a sequence controller 14 that controls the sequence of radio-frequency pulses (RF pulses) and gradient pulses in dependence upon a selected pulse sequence PS or a sequence of multiple pulse sequences to acquire a number of slices in a volume region of interest of the examination object during a scanning session. Such a pulse sequence PS can be specified and parameterized in a scan or control protocol P. Different control protocols P for different scans or scanning sessions are typically stored in a memory 19 and can be selected by an operator (and optionally changed if necessary) and then used to carry out the scan. In the present case, the control computer 13 contains pulse sequences for the acquisition of the raw data for the determination of the T.sub.1 values or reciprocal RI values thereof and the T.sub.2 or T.sub.2* values or the reciprocal R.sub.2 or R.sub.2* values thereof required according to the invention (as is typically already the case with standard scans) and optionally further values, such as fat values, texture, etc.

[0044] To emit the individual RF pulses of a pulse sequence PS, the central control computer 13 has a radio-frequency transmitter 15 that generates and amplifies the RF pulses and feeds them into the RF transmission antenna 5 via a suitable interface (not shown in detail). To control the gradient coils of the gradient coil arrangement 6 in order to activate the specified pulse sequence PS suitably, the control computer 13 has a gradient interface 16. It could also be possible for the shim coils to be activated via this gradient interface 16 since the gradient coils are used by setting the DC offset currents for shimming the Bo-Felds (namely for compensating the linear portion of the field distortions). The sequence controller 14 communicates with the radio-frequency transmitter 15 and the gradient system interface 16 in a suitable way, for example by emitting sequence control data SD to execute the pulse sequences PS.

[0045] The control computer 13 also has a radio-frequency receiver 17 (which likewise communicates with the sequence controller 14 in a suitable manner) in order to receive magnetic resonance signals within the readout window specified by the pulse sequence PS and thus to acquire the raw data.

[0046] A reconstruction processor 18 accepts the acquired raw data and reconstructs therefrom magnetic resonance image data, i.e. in particular the T.sub.1-weighted images (or T.sub.1 maps) or reciprocal R.sub.1 maps thereof and the T.sub.2 or T.sub.2*-weighted (or T.sub.2 or T.sub.2* maps) or the reciprocal R.sub.2 or R.sub.2* maps thereof. This reconstruction is performed on the basis of parameters, which can be specified in the respective scan or control protocol P. This image data can then be stored in a memory 19.

[0047] The details of how suitable raw data are acquired by the irradiation of RF pulses and the activation of gradient pulses, and MR images or parameter maps reconstructed therefrom, are known to those skilled in the art and thus need not be explained in further detail here.

[0048] Here, the compilation of the images or parameter maps required in the further method and optionally also an evaluation of the raw data to identify the higher-ranking parameters, such as fat values, texture values of the region of interest etc., are performed in a parameter map compilation processor 20 of the reconstruction processor 18.

[0049] All these values can then be sent to an evaluation computer 21, which here also includes the characterization processor 22, here specifically a classification processor, for the method according to the invention. In principle, the parameter map compilation processor 20, which is depicted in FIG. 1 as part of the reconstruction processor 18, can also be part of the evaluation computer 21 or classification processor 22. In the example shown, it is assumed without restricting the generality that the characterization of the liver tissue is performed in the form of a classification.

[0050] Here, a (software) module of this classification processor 22 is a data conversion unit 23, which analyzes the parameter maps and forms the value tuples therefrom and then transmits these into the multidimensional parameter space. For example, here a mean T.sub.1 value, a mean R.sub.2*-value, a fat value FF for the fat content etc. within the region of interest ROI of the liver O are formed and the value tuple for this region of interest ROI formed therefrom and then transferred into the parameter space. In this context, the determination of the mean values can be performed such that, first, the values for the individual image points in the region of interest are determined and then a mean value is formed on the basis of these individual values.

[0051] Alternatively, as described it is also possible for separate value tuples to be formed for individual image points and/or regions and transferred into the parameter space.

[0052] Subsequently, a classifier 24 (as a further software module) then performs the actual classification, as will be explained later in more detail with reference to FIGS. 3 to 5. The infotmation required for this on the location of the boundary hyperplanes and/or the different clusters of reference tuples can, for example, be stored in a memory 25 to which the classifier 24 has access.

[0053] The central control computer 13 can be operated via a terminal 11 with an input unit 10 and a display unit 9 via which the entire MR apparatus 1 can hence be operated by an operator. It is also possible for MR images to be displayed on the display unit 9, and the input unit 10, optionally in combination with the display unit 9, can be used to plan and start measurements and in particular select and optionally modify control protocols P.

[0054] Here, it is also possible for results of the automatic classification to be presented, for example as specific suggested classifications, which can be accepted or rejected by the person tasked with performing the examination. To this end, it is also possible to display the parameters spaces used by the display 9, particularly when a two-dimensional parameter space or three-dimensional parameter space is involved, as is the case in FIGS. 3 to 5.

[0055] The MR apparatus 1 according to the invention and its control computer 13 can also have further components, not shown in detail here, but typically present in systems of this kind, such as a network interface in order to connect the entire system with a network and to enable the exchange of raw data and/or image data or parameter maps but also further data, such as patient-relevant data or control protocols.

[0056] The image points taken into account during the evaluation—i.e. for the spatial region of the liver tissue for which such a suggested classification is to be compiled—depend upon which region of interest is defined. As depicted in FIG. 2 a region of interest ROI can be defined in the form of a box or the like if, for example, a classification is only to be performed for a specific region of the liver. This can take place, for example, by the liver O or a region of the body comprising the liver O being displayed in the form of two or three-dimensional image data on the display unit 9 and the examiner then enters coordinates in order to set a corresponding box or draws this box using a mouse or similar graphical tools. However, instead of a box, it is also possible to enter any other shape. If classification of the entire liver tissue of the liver is desired, ultimately the entire liver O is the region of interest ROI′. To identify the tissue associated therewith, it is possible to perform manual segmentation, i.e. with the use of the user interface, by the boundary lines being drawn therein, for example in the image data, or automatic segmentation.

[0057] When the region of interest ROI, ROI′ has been defined, the image points can be evaluated as described above in order to form the desired value tuples, which are then entered in the parameter space for the further classification. This is shown in FIG. 3 in a very simple example. Here, the parameter space PR is only two-dimensional, wherein, the R.sub.2* value is shown in s.sup.−1 along one coordinate axis, here the abscissa, and the T.sub.1-relaxation time in ms along the other coordinate, here the ordinate. Three straight lines are drawn in this two-dimensional parameter space PR as boundary hyperplanes H1, H2, H3, wherein these straight lines overall divide four regions from one another within the parameter space PR. Each one of these regions is assigned its own class which is assigned to a specific fibrotic condition of the liver tissue from which the value tuple originates. The boundary hyperplanes H1, H2, H3 or the individual regions for the individual classes are defined with the aid of reference value tuples RT originating from liver tissue from livers for which the degree of fibrosis is known. Then, for the current examination, it is only necessary to define the position of the value tuple T belonging to the liver to be examined and it is hence very easy to define the correct class.

[0058] It should be noted that the depiction in FIG. 3 is only a schematic representation intended to demonstrate the principle and neither the values nor the boundary hyperplanes H1, H2, H3 are based on medical data. In particular, a significantly higher number of reference value tuples is used in order to define the boundary hyperplanes H1, H2, H3. The boundary hyperplanes can also consist of curved lines or surfaces.

[0059] As mentioned, preferably further values are also added to the value tuple. An advantageous further attribute in the value tuple is a fat value or a fat portion FF, which can be defined in %.

[0060] FIG. 4 is a schematic representation of such a three-dimensional parameter space PR′ with which the T.sub.1 relaxation time is again shown in ms over the fat value FF in % and the reciprocal T.sub.2* relaxation time, i.e. the R.sub.2*-value, is shown in s*

[0061] Here, once again reference value tuples RT were used to select a suitable hyperplane H4 in order to divide the parameter space PR′ into different regions assigned to specific classes. In the simplest case, as depicted here, there is only one hyperplane H4, which divides the parameter space PR′ into two regions, namely one signaling, for example, that there is no fibrosis as yet, and a further region indicative of fibrosis. This representation also only entails values and hyperplanes H4 used to elucidate the system used, but not medically substantiated values.

[0062] The same applies to the depiction in FIG. 5, which also shows a possibility of classification within a three-dimensional parameter space as shown in FIG. 4. This depicts how the reference value tuples RT in the space form clusters C1, C2, C3. Each of these clusters C1, C2, C3 can be assigned to a specific class representing a specific condition of the liver. The reference value tuple RT can also be determined here in that corresponding value tuples are compiled on the basis of liver tissue from patients with livers in a wide variety of pathological conditions. It is then possible for a corresponding data base to be extended within the framework of further examinations in that the value tuples for livers that have been examined and classified with the aid of the method according to the invention can be added.

[0063] These clusters could be used, for example, to form boundary hyperplanes forming the boundaries between the individual clusters. This has the advantage of enabling rapid classification with reference to the location between the boundary hyperplanes or the reference to the boundary hyperplanes and it is not necessary to retain all the reference tuples as data.

[0064] There are various possibilities for carrying out an evaluation using the point clouds of the reference value tuples RT in the clusters C1, C2, C3. One possibility consists in determining a collective value for each cluster C1, C2, C3, for example the focal point S1, S2, S3 (once again only marked symbolically in FIG. 5) of the respective point cloud, representing the respective cluster C1, C2, C3 in the parameter space. It would then be simple to determine the distance dl, d2, d3 to the respective focal point S1, S2, S3 for a value tuple T originating from the liver to be examined, wherein it is, for example, possible to use the Euclidean distance. The assignment of the value tuple T to one of the clusters C1, C2, C3 is then performed by simply accepting the cluster C1, C2, C3 with a focal point 51, S2, S3 closest to the value tuple T.

[0065] Alternatively, it possible in each case to use the distance to the closest reference tuple RT in the individual clusters C1, C2, C3. Similarly, the main axes of the clusters C1, C2, C3 could be determined and then the point of intersection of the main axes in each case form a representative point for the respective cluster C1, C2, C3. Here, a wide variety of determination possibilities is conceivable.

[0066] FIG. 5 also shows that it is in principle possible for a number of value tuples to be transmitted as a value-tuple group TG into the parameter space PR″ for a liver that is currently to be examined and for the classification then to be performed on the basis of this value-tuple group TG. For example, here once again the focal point of the value-tuple group TG could be determined and a distance between the focal point of value-tuple group TG and the focal points S1, S2, S3 of cluster C1, C2, C3 determined in each case (in FIG. 5, by way of example, only the distance dG to the focal point S2 of the middle cluster C2 is depicted) in order then to perform the classification in a similar way to that described above.

[0067] The way in which the characterization, in particular classification, is ultimately performed is in particular also dependent upon which further parameters are added if, instead of a three-dimensional parameter space, a four- or five-dimensional space is used, how the position of the individual value tuples in this parameter space is depicted and how effectively it is possible to demarcate the individual regions from one another.

[0068] Although modifications and changes may be suggested by those skilled in the art, it is the intention of the Applicant to embody within the patent warranted hereon all changes and modifications as reasonably and properly come within the scope of the Applicant's contribution to the art.