Method and device for selecting body model positions for SAR monitoring of a magnetic resonance transmit array

1. Field of the Invention

The present invention concerns methods and devices for monitoring an examination subject in a magnetic resonance apparatus.

2. Description of the Prior Art

Identification of the local SAR in an MR scanner has not previously been possible by means of measurement in clinical practice. The only known and used technique is to make use of a theoretical simulation (model) that accounts for both the patient and the structure of the transmission antenna as an electrical model. In the operation of MR scanners with a transmission array and with RF pulses that can exhibit an arbitrary pulse shape for each array element (variation of amplitude and phase), a multitude of overlay possibilities result. Comprehensive monitoring in the local SAR thus would involve a high degree of complexity, but monitoring of the local SAR value is absolutely necessary for the safety of the patient and is required by corresponding regulations. The overlaying of the electrical fields in array antennas is in particular critical because the E-vectors add linearly but the local power transfer is proportional to E 2 .

In a system with K elements, L phase steps and M amplitudes, (M*L) K combination possibilities result for every model point in the search for a maximum potential hotspot.

Given a typical exposed mass of 50 kg and a hotspot size of one gram, 50,000 model points result for which these combination possibilities must be considered. Given a higher channel count and appropriate resolution (phase and amplitude), the determination of all combinations for every model point is not applicable for clinical application. One possibility loophole is focusing on a reduced number of suitably selected model points that cover the possible amplitude and phase combinations as well as possible.

However, the monitoring for individually selected potential “hotspots” is generally not simple. For example, if a model point at which the E-fields of the individual antennas can theoretically superimpose at maximum (for example given antenna currents that are the same in terms of magnitude) is determined from among all model points in the patient model, this maximum is the case at this model point only for a specific distribution of the phases of these currents.

If the heat production determined by calculation at this model point is now monitored as being representative of all voxels, the actual maximum superimposition of the electrical fields can occur at other model points (that are not monitored) if the phase distribution of the currents deviates from the specific distribution (for example polarity reversal of a single antenna current). The SAR monitoring thus would be virtually blind to the actual occurring maximum.

An object of the present invention is to provide a method to systematically select model points that are best suited for an optimally comprehensive monitoring of the local SAR of the entire patient.

A systematic approach with regard to all possible field superpositions is not known. Conventional approaches normally use calculation methods for predetermined current distributions at the array elements, for example with the CP mode. Numerous basic tasks exist for electromagnetic calculation of the electrical fields and SAR values in patient models. This calculation of the fields of individual array elements with such methods is the basis for the combinatorial analysis of the fields that is the subject matter of this invention.

It is the intent of the present invention to improve the SAR monitoring of an examination subject in a magnetic resonance tomography apparatus.

The method is implemented in a computerized processor.

The above object is achieved in accordance with the present invention by a method for specific absorption rate (SAR) monitoring in a magnetic resonance (MR) system having multiple transmission coils, that are individually activated by being charged with respective currents that cause each coil to emit an electrical field, having a field variable pertaining thereto, with each current having a magnitude and a phase. In accordance with the present invention, a modeled examination subject is modeled in a computerized processor as a number of voxels, each respective voxel having a total electrical field therein that is produced by vector addition of respective contributions to the respective voxel by the respective electrical fields emitted by the respective transmit coils. This total electric field exhibits a total field variable. In the computerized processor, a primary model point voxel is automatically identified, from among the multiple voxels of the modeled examination subject, in which an absolute maximum of the total field variable exists within the modeled examination subject. Additionally in the computerized processor, at least one auxiliary model point voxel is identified, from among the multiple voxels of the modeled examination subject, in which a relative maximum of the total field variable exists within the modeled examination subject.

Further in accordance with the invention, in the computerized processor, a back-calculation is automatically executed that, from the total field variable for the primary model point voxel, calculates respective relative times of activation of the respective coils in the multiple transmit coils in order to produce the total field variable at the primary model point voxel. The same type of back-calculation is also undertaken for the at least one auxiliary model point voxel to determine respective relative times of activation of the respective coils in the multiple transmission coils to produce via total field variable at the auxiliary model point voxel.

The primary model point voxel and the least one auxiliary model point voxel are stored. When MR data are subsequently acquired from an actual examination subject in the MR system, the stored primary model point voxel and the stored auxiliary model point voxel are retrieved and the SAR of the actual examination subject is monitored at voxels in the actual examination subject respectively corresponding to the primary model point voxel and the auxiliary model point voxel.

The first through penultimate step of the method are advantageously implemented for multiple different possible examination subjects before a present examination subject is examined in the magnetic resonance apparatus, wherein—if a present examination subject is examined in the magnetic resonance apparatus—it is determined which of the most similar of the possible examination subjects corresponds best to the present examination subject (the examination subject to be examined), in particular with regard to weight and shape; whereupon a respective variable pertaining to the electrical field is determined for the most similar of the examination subjects for specific voxels in the present examination subject (the examination subject to be examined) and is used for the SAR monitoring. This enables a fast and efficient monitoring of patients of different statures.

The invention also encompasses a computerized device that implements the above method with a processor configured (programmed) to implement the aforementioned method steps.

FIG. 1 schematically illustrates, a whole-body MRT coil and a local coil.

FIG. 2 shows an eight-channel system of a whole-body MRT coil, wherein every two antenna rods are combined into one transmission coil.

FIG. 3 illustrates the overlaying of both E-vectors of two antennas.

FIG. 4 illustrates the overlaying of the E-vectors of two antennas.

FIG. 5 illustrates the overlaying of the E-vectors of more than two antennas.

FIG. 6 shows the search for the phase for which a maximum total variable results.

FIG. 7 shows model points at which the maximizing phase vector deviates only minimally from that of the first model point.

FIG. 8 shows prior tests with hypothetical data.

FIG. 9 shows histograms of the different overlay types in comparison.

FIG. 10 shows the decrease of a weighted E-field “E 2 -weighted” (“Sine-Square”) over the number of concurrently considered phase vectors (or, respectively, model points connected with these), obtained within the scope of a test of the method on hypothetical data.

FIG. 11 shows, as a test, the coverage of the E 2 maxima at the 1,000 model points by the 50 selected virtual test points (10,000 random phase and amplitude combinations of the antenna currents).

FIG. 12 shows, as a test, which virtual measurement point detects, and how many hotspot maxima.

FIG. 1 shows a magnetic resonance apparatus 1 with a whole-body coil 2 and a tube-shaped space 3 into which a patient bed 4 with, for example, a patient 5 and a local coil 6 can be driven in order to generate exposures of the patient 5 . Here a local coil with which good exposures are enabled in a local region is placed on the patient. The maximum SAR in the patient 5 should be monitored; for this multiple points 8 , 9 , 10 (also called model points 8 , 9 , 10 in the following) in the patient 5 should be determined at which absolute or relative maximum power density is realized (thus a greater power density than at other points in the patient). The point at which the highest power density occurs is called the primary model point voxel 8 in the following; points at which lower power density occurs than at the primary model point voxel 8 (but more than is expected elsewhere in the patient) are called auxiliary model point voxels 9 , 10 (voxels are small volumes, for example small cuboids in the patient or examination subject).

As an example, FIG. 2 shows the patient 5 in an eight-channel MRT system 2 , wherein every two antenna rods are respectively combined (including a 22.5° relative phase rotation). Electrical fields are simulated with an exemplary model of a TX array, together with a HUGO patient model, 4 mm resolution, FDTD with Microwave Studio, wherein a central disc in the direction of the body axis lies in the resonator center, with 27 incorporated slices.

Methods and devices are described herein that enable the SAR calculation to be implemented for arbitrarily combined RF pulse shapes at relatively few monitoring points (as noted above, the location of the first monitoring point is also called a primary model point voxel and the locations of the additional monitoring points are also called auxiliary model point voxels), and therefore the entire (modeled) body is covered as well as possible with regard to the relevant local heating (known as “hotspots”). In particular, the method in accordance with the invention is directed to finding these model points (primary model point voxel and auxiliary model point voxels) after the calculations of the individual fields of the array elements that cover the entire exposed volume have been implemented once with a sufficiently large number of model points for a reference condition. In addition to the reduction of the computing effort for monitoring the local SAR, this method allows the compression of the very considerable model data for the evaluation of sensitivity considerations with regard to model changes. Beyond this, through the compression it is also possible to implement “worst case” overlays from different patient models, and thus to expand the scope of validity of an employed basis for local SAR calculation. The idea of the model point search method can also be considered for overlaying multiple patient models.

The invention utilizes the fact that the electrical field that every array element produces at every model point depends linearly on the excitation current of this array element, and the fact that the individual fields of multiple array elements combine linearly.

The (electrical) field at the model point i that is caused by the array elements indexed with j is thus represented as the following sum:

E i =ΣS ij I j .

All involved variables are generally complex in nature, the values of E three-dimensional complex vectors.

However, in principle this method can also be expanded for a representation with a conductivity sensor.

The values of the sensitivity matrix must have been determined by a preceding calculation with an electromagnetic model (for example the FDTD method).

The locally produced RF power density (for example in a voxel) is then the product Re (E*conj(E)*σ), wherein cy is the locally complex conductivity of the body. Here the additional considerations for a modified “effective” field E′ and the sensitivity matrix S′ that is connected with this (in which a has already been included, thus Re (E′*conj(E′):=Re (E*conj(E)*σ) and E i ′=: ΣS ij ′I j ) are appropriately implemented.

At the point of a scalar value σ, in general a diagonal tensor can also occur with different conductivities in the three primary axis directions (σ x , σ y , σ z ). In this case the powers that result due to the electrical field components E x , E y , E z are calculated and added separately per model point. All additional considerations remain valid without limitation.

The invention in particular concentrates on the scenario in which the fields of the TX array elements maximally superimpose, i.e. exhibit relatively high current strengths. The cases in which a single array element dominates do not need to be additionally considered.

The model point (primary model point in the primary model voxel) is initially determined at which the fields of antenna currents that are identical in terms of magnitude can theoretically maximally overlap, as are the phases of the antenna currents that produce this maximum overlap.

The phase distribution of the currents at which the largest SAR value locally results is also determined for each model point.

Due to the linearity of the above equations for the entire volume, SAR values that are now calculated for this model point (primary model point) cover the local SAR value calculation, but only for current distributions that correspond except for a complex factor of the found current distribution.

Current distributions that are linearly independent relative to the found current distributions (i.e. have other amplitudes or phase ratios) are not monitored in the monitoring of SAR at this hotspot, but can likewise cause relevant local heating at other model points.

One method to find an additional model point (thus an auxiliary model point voxel) at which a maximally high SAR can occur (as a relative maximum) to which the monitoring of the first model point is blind can be achieved by the following steps:

Of the local maximum current distributions that are found in the first search, those that produce a minimal local heating at the first model point (primary model point) are additionally considered. These are now evaluated with regard to their heating at other model points.

This method can be applied iteratively until the resultant heating changes only slightly due to the remaining possible current distributions.

This result is now theoretically still limited to the framework condition that all antenna elements are operated with the same current magnitude. If the antenna currents have different amplitudes in the actual excitation, this can be approximated by a corresponding adjustment of the relative phases in the overlaying. However, the model points are selected so as to be optimally sensitive to this.

Separate consideration of the case that only individual antenna elements with different currents are actually operated can be done due to the absent overlaying effect, which cannot be utilized here. For example, this can ensue in that the potential hotspots of the individual excitations are to be added to those for the combinations.

How the E-vectors of different TX contributions can be superimposed is apparent from FIG. 3-6 .

FIG. 3 shows (Part 1-Step 1) for two antennas the spatial worst case of an overlaying of two E-vectors E 1 and E 2 that are both produced at a model point via one antenna current per antenna.

A real case—meaning + and − signs are allowed—is considered.

FIG. 4 shows (Part 1-Step 2) a complex case; here a real field (per definition) E 1 of a first coil is shown, the phase φ 2 (relative to the first field E 1 ) of the field E 2 of an additional coil is selected so that the real part of E 2 maximally overlays three-dimensionally with the real part of E 1 . (In general an exact parallelism of E 1 and E 2 cannot be achieved in 3D space.)

FIG. 5 shows (Part 2-Step 1) more than two (thus three here) E-field vectors. These are successively combined: it begins here with the largest E-vector; the E-vector that maximally superimposed with this is added (here only the real portion is shown). For efficiently superimposing portions, in the worst case a projection of the maximum sum vector is implemented.

FIG. 6 shows (Part 2-Step 2) more than two (here eight) vectors of E-fields that are respectively generated via a current (I 1 through I 8 ) in a respective coil of a coil array. Here the phase (model point transmission phases) of a current at which a maximum results in the combination of the E-fields is respectively sought in the complex region.

Here an 8-dimensional, complex vector (thus a vector with magnitude and phase of the E-field) results for the currents (I 1 through I 8 ) at which the E-fields (generated by the respective currents I 1 through I 8 ) would maximally superimpose, as well as the magnitude of the maximal E-field overlay (I 1 =1+0i per definition).

An identification of “virtual measurement points” (respectively in particular at the primary model point voxel or one of the auxiliary model point voxels) follows for the local SAR monitoring (if a patient is examined), which “virtual measurement points” take into account the (most relevant) overlaying possibilities of the E-vectors.

For the monitoring a first “virtual measurement point” (at a primary model point voxel) is initially chosen so that the E-fields can theoretically maximally superimpose at its position (for example such that a maximum Re(σE 2 ) results).

This model point also covers all of the model points at which the maximizing current phase vector (the currents I 1 through I 8 ) is the same except for one factor.

Additional model points (in auxiliary model point voxels) with E-vectors deviating from these for additional difficult cases (worst cases) of one of the multiple (8 if there are 8 coils) complex E-vectors are not yet covered. The greater the deviation, the greater the risk.

Model points in which the maximizing phase vector deviates only slightly from that of the first model point will deliver lower SAR than the first model point if the σE 2 maximum value of the first model point has a sufficient separation from that of the second model point:

I new =I max,pos1 +ΔI orthog with I max,pos1 *conj (Δ I orthog )=0

Δ E=S pos1 *ΔI

Δ E=S pos2 *ΔI

E ( I max,pos1, pos2 )= S pos2 *I max, pos1

E ( I max,pos1, pos1 )= S pos1 *I max, pos1

Additional model points are thus sought that optimally expand, i.e.

that likewise have a high potential maximum SAR value whose SAR-maximizing current phase distribution deviates maximally from that of the already found model point.

Heuristic approach:

,,Sine-Square”=1−,,Cos-Square=1−Abs(I 1 *con(I 2 )) 2 is used as a measure of the difference of the 8th phase vectors I 1 and I 2 . The potential SAR values of all points are multiplied with the difference measurement of the phase vectors relative to the first point. A model point with newly weighted maximum value is sought. The potential SAR values of all points are multiplied with the difference measurement of the phase vectors relative to the second point. . . . (continued iteration)

Auxiliary model point voxels are thus identified at which the phases (auxiliary model point transmission phases) of the currents in the coils are markedly different than in a primary model point voxel (for example displaced by 180° relative to the phase in a primary model point voxel).

This can ensue for multiple models of different weight, different size, etc.

When a patient is examined, the model that best corresponds to the patient (examination subject) can be identified with a (if necessary brief) scan, and the currents in the coils are limited so that the model points identified for the selected model (examination subject) do not exhibit any electrical fields greater than predetermined electrical fields.

FIGS. 8-12 show results of tests.

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