WO2006026976A1

WO2006026976A1 - Calibration method for contrast agent-based perfusion imaging

Info

Publication number: WO2006026976A1
Application number: PCT/DE2005/001573
Authority: WO
Inventors: Peter Brunecker; Jens Steinbrink; Arno Villringer
Original assignee: Charite- Universitätsmedizin Berlin
Priority date: 2004-09-08
Filing date: 2005-09-08
Publication date: 2006-03-16
Also published as: DE102004043809A1; DE102004043809B4

Abstract

The invention relates to a calibration method for contrast agent-based perfusion imaging for a section of vascular blood circulation of a test person. According to said method, image representation of contrast agent measuring data of contrast agent-based perfusion imaging are coregistered for the section of the vascular blood circulation together with the image representation of angiographic measuring data of an angiography measurement of the section of the vascular blood circulation. For this purpose, an optimum overlap of two or more arteries in the section of the vascular blood circulation is produced between the image representation of the contrast agent measuring data and the image representation of the angiographic measuring data, thereby allowing to correlate image elements of the image representation of the contrast agent measuring data with image elements of the image representation of the angiographic measuring data. A calibration measure is then determined for the correlation between the signal intensity of voxels in the image representation of the contrast agent measuring data and a respective blood volume pertaining to one of the voxels by correlating the geometrical parameters for one or all arteries in the section of the vascular blood circulation, which parameters are derived from the angiographic measuring data, with the voxels derived from the contrast agent measuring data of the contrast agent-based perfusion imaging method.

Description

Method for calibrating contrast agent-assisted perfusion imaging

The invention relates to a method for calibrating a contrast medium-based Perfusi¬ onsbildgebung.

In medicine, the question of the perfusion state of tissue is increasingly asked, which is also referred to as perfusion. This can be an important piece of information, especially in the case of severe neurological diseases such as stroke, the cause of which is a partial hypofusion of the brain.

In this case, imaging processes play a special role, since they provide information on the size and the spatial distribution of the perfusion disorder (see Barbier et al .: Methodology of Brain Perfusion Imaging, J.Magn. Reson. Imaging 13 (2001) 496-520). ,

Contrast-enhanced perfusion imaging of the human brain with nuclear tomography is a standard clinical diagnosis and is thus in competition with perfusion images with positron emission tomography (PET) and computer tomography (CT). Of importance are these methods for the diagnosis of neurological diseases that are associated with a disturbed blood supply, for example, a stroke.

In the case of contrast agent-assisted perfusion formation, a paramagnetic contrast agent is administered venously to the patient in a short time interval. The stimulation and discharge behavior in the brain can be determined dynamically with fast T2 * -weighted sequences with a typical time resolution of one second in voxels of approximately 2 × 2 × 5 mm ³ . Numerous algorithms have been published which translate measured contrast agent profiles into parameters of cerebral perfusion (cf., for example, Helenius et al., Cerebral modmodynamics in a healthy population measured by dynamic susceptibility contrast MR image, Acta Radiologica, 44 (2003) 538-546). In particular, the physical parameter flux as the product of velocity and volume is of particular importance here. For the absolute quantification of the volume, however, a device-specific and patient-specific relationship between measured MR signal change and the corresponding blood volume in a voxel is required.

One of the known methods is the dynamic contrast measurement in Kernspintomo¬ graphene. Here, a paramagnetic contrast agent is administered intravenously as a bolus and its propagation spatially and temporally detected. The measuring method used here is usually an echo-planar imaging (EPI), which permits a high temporal resolution with a sufficiently large measuring volume. This method, which belongs to the group of gradient echo sequences, supplies an image weighted with the relaxation time T2 *, whose time-dependent signal intensity S (t) in each voxel is determined by the concentration of the contrast agent C (t)

C (t) ~ -log (S (t) / S (0)) (1)

is determined. However, this behavior only applies in the limiting case of low concentration or low capillary density. In large arteries, for example, saturation and partial volume effects as well as local field inhomogeneities destroy this connection.

Due to the physiological variability of the systemic circulation in the blood transporting vessels and the occurrence of dispersion effects, the contrast medium neither reaches the brain in a fixed time nor in the original concentration-time course predefined by the type and duration of the injection. In order to make the quantification independent of this strongly varying quantity, the concentration-time profile of each voxel (spatial generalization of the pixel) with the arteries supplying the brain is mathematically unfolded

C (t) = C ri _Arte _e (t) XR (t). (2)

In this case selected voxels serve as the source of the arterial signal curve, which in their signal characteristic (faster, early rise of the concentration to a high value) indicate the presence of arteries in the interior. The result of the mathematical unfolding then describes the perfusion, ie the perfusion state in each voxel by means of a residue function R (t), which is determined by the parameters cerebral blood volume (CBV), cerebral blood flow (CBF) and mean transit time ( MTT) is characterized. It is disadvantageous that the parameters CBV and CBF are directly dependent on the choice of arterial voxels and thus of C _A rterie (t) in this case. therefore, arterial voxels are required with reproducible signal behavior for the quantification of perfusion parameters.

Therefore, the evaluation on the basis of voxels, which are located completely within an artery, seems expedient. You can calibrate the cerebral blood volume (CBV) can be used by determining a global calibration coefficient so that the CBV of this voxel reaches 100% and all other voxels are scaled according to the measurement. However, the interference factors already mentioned above, such as saturation and partial volume effects as well as local field inhomogeneities, argue against the use of the large arteries necessary for this purpose. In practice, therefore, usually smaller, embedded in the voxel vessels of the bloodstream serve as a source of the measured arterial signal. These embedded vessels thus fill only a partial volume of the voxel under consideration. Here, however, there is no possibility of calibrating the CBV, since the volume fraction of these embedded vessels at the voxel as well as the contribution of the residual volume of the voxel which can likewise be reached by the contrast agent are not known.

In the case of studies involving groups of subjects or patients, this problem is addressed by setting a CBV group agent to physiologically sensible standards. These standard values are either taken from the literature or in a parallel comparative study with a gold standard, e.g. PET (Positron Emission Tomography). However, this approach does not allow the determination of parameters from which conclusions about the perfusion status of a single individual can be derived. The application of standard values to patients is also problematic, since their disease can potentially influence them.

In addition to determining the blood volume, the question of CBV calibration also applies directly to quantification of the CBF, since this variable is determined from the maximum of the residual function and therefore linearly scales with the CBV, the surface area of the residual function. For most applications in medicine, however, CBF is the decisive factor since, in contrast to CBV, it is able to assess the supply state of tissue.

The object of the invention is therefore to provide a method for calibrating a contrast agent-based perfusion imaging, which supports a more accurate determination of perfusion-relevant parameters.

This object is achieved erf durchndungsgemäß by a method according to the independent An¬ claim 1. The invention provides a method for calibrating contrast agent-based perfusion imaging for a section of a blood vessel circuit of a subject, in which an image representation of contrast agent measurement data of the contrast agent-based perfusion imaging for the section of the blood vessel circuit with an image representation of angiography measurement data of an angiography measurement the coregistration of the section of the blood vessel circulation is effected by forming an optimized overlap of one or more arteries in the section of the blood vessel circulation between the image representation of the contrast agent measurement data and the image representation of the angiography measurement data, whereby image elements of the image representation of the contrast medium measurement data and Picture elements of the image representation of the angiographic measurement data can be assigned to one another; and subsequently determining a calibration measure for the relationship between a signal intensity of voxels in the image representation of the contrast agent measurement data and a blood volume respectively associated with the voxels, by geometric parameters derived from the angiography measurement data for one or all arteries in the section of FIG Blutgefaßkreislaufes the voxels from the contrast agent measurement data of the contrast agent-based perfusion imaging are assigned.

With the help of the coregistration of the measurement results of different methods, which relate on the one hand to the course of the distribution of the contrast agent and on the other hand to angiography of a portion of the blood circulation, a measurement evaluation of the contrast agent-based perfusion imaging is made possible for the subject. In this way, sources of error are excluded which may occur during a calibration of the contrast agent-based perfusion imaging on the basis of mean values for groups of test persons. Overall, the evaluation accuracy of the determined measured values of the subject for the contrast agent-based perfusion imaging is increased.

The invention is explained below with reference to embodiments with reference to a drawing. Hereby show:

Fig. 1 is a schematic representation for explaining the measuring methods used; FIG. 2A shows an MR angiography of a head as an MIP representation; FIG.

FIG. 2B shows a co-registration of an image representation of the measurement data of the MR angiography according to FIG. 2A with an image representation of a slice from a T2 * -weighted imaging; FIG. FIG. 2C shows a representation limited to the volume of the T2 * -weighted layer, here represented by means of a superimposed, black-represented local MIP representation; FIG.

Figure 3A shows a rectangular selection from the T2 * weighted layer after marching cube reconstruction;

FIG. 3B shows the same rectangular selection as in FIG. 3A, but in isometric projection perpendicular to the layer; FIG.

Fig. 3C shows the result of a classification in which noise-related artifacts are removed; Fig. 3D shows the selection of a single arterial segment for further analysis; FIG. 4A shows the division of an arterial segment into voxel groups perpendicular to the layer guide; FIG.

Fig. 4B shows a T2 * weighted image consisting of much larger voxels;

FIG. 4C shows the superposition of the voxel-subdivided arterial segment with a T2 * -weighted image; FIG. 4D shows the result of an optimization of the position with the aim of a maximum correlation coefficient between the relative concentration, calculated from T2 * - Bildera, and the volume fraction of the arterial segment per T2 * -Voxel; and

5 shows a relationship between relative concentration and calculated volume proportions per T2 * voxel in the case of balanced contrast medium concentration within the blood vessels.

A method for calibrating a contrast medium-based perfusion imaging is explained below with reference to FIGS. 1 to 5. An essential element of the exemplary embodiment is that the cerebral blood volume (CBV) is calibrated with the aid of an MR imaging technique (MR magnetic resonance), MR angiography.

In the method, MR imaging techniques, known as such, are combined to calibrate the relationship between blood volume and signal intensity in the individual measurement situation and for individuals. For this purpose, large arteries of the brain are recorded with high-resolution MR angiography (resolution currently approx. 0.5 × 0.5 × 1 mm ³ ) and coregistered (superimposed) with rapid T2 * -weighted imaging. The spatial position of one or more arteries is changed on the contrast medium image until the greatest possible agreement between the two measured data representations consists. After the optimal match has been determined, a partial volume fraction of the artery (s) at each voxel can be determined for each voxel that passes through the artery (s) (volumetry). By looking at several voxels, there is a correlation between the signal intensity in T2 * -weighted imaging and the volume of blood in the voxel. An individual calibration is thus made possible.

The method automatically processes measured raw data. For this purpose, a conventional computer device is used, which has sufficient computing capacity to process the image data. For carrying out the method, an application program is installed on the used computer device, for example a personal computer, in order to process the acquired measurement data of the imaging techniques, as will be explained in detail below.

When measuring the raw data, an overlapping measurement at the level of a layer 10 according to FIG. 1 is carried out. Angiographic data are measured so that a plurality of branches of the middle cerebral artery are within the measurement volume. For evaluation, the calculation of an MEP ("maximum intensity projection") map can be performed.

When measuring the raw data, the following measuring methods are used, for example:

(i) Dynamic T2 * -weighted imaging (example: TE 54 ms, TR 0.8 s, TA 60 s, matrix 128 × 128, 8 layers, field-of-view 256 mm, layer thickness 6 mm) with paramagnetic contrast agent ( Example: 20 ml of 0.5 M Gd-DTPA and subsequently 20 ml physiological saline solution, delivery rate 4 ml / s); and

(ii) Time-of-flight MR angiography (Example: T _E 6.5 ms, T _R 39 ms, matrix 128 × 128, 8 layers, field-of-view 256 mm, slice thickness 0.875 mm, field-of-view 200 × 200 × 56 mm ).

Fig. 2A shows an MR angiography of a head in an MEP representation. The direction of projection here is perpendicular to the T2 * -weighted layers over the full volume of MR angiography. 2B, a co-registration of an image representation of the measurement data of the MR angiography (cf., FIG. 2A) with an image representation of the slice 10 from the T2 * -weighted imaging is carried out, which means that the Image representations of the determined using the two measurement methods raw data are stacked. The position information contained in the measured raw data from the nuclear spin tomography measurement define an affine transformation, which converts the data space (voxel space) into a Cartesian coordinate system.

In the next step, a local MIP map (MIP map) is determined, which is limited to the selected T2 * -weighted slice 10 in order to select suitable arterial segments (see FIG. 2C) for a subsequent 3D reconstruction. This is done by selecting a rectangular area which further includes segments to be used. FIG. 2C shows a representation limited to the volume of the T2 * -weighted layer 10, represented here by means of a superimposed, black-represented local MIP map. This representation later serves to select regions for 3D segmentation.

This is followed in the illustrated exemplary embodiment within the selected rectangular area by a three-dimensional reconstruction of the surface of the arteries from the volume data obtained with MR angiography using a Marchmg Cube algorithm (compare FIGS. 3A, 3B). Figure 3 A shows a rectangular selection from the T2 * weighted layer 10 after marching cube reconstruction. The shortest side length here is the layer thickness. FIG. 3B shows the same rectangular selection as in FIG. 3A, but in an isometric projection perpendicular to the layer 10. A threshold value for the signal intensity of the volume data is determined under visual inspection of the reconstruction in real time. The threshold is preferably set so that noise-related reconstruction artifacts in the vicinity of the vessels are just beginning to emerge. The entire three-dimensional reconstruction takes place in a congruent to the T2 * data space, which was higher in the slice direction by an integer factor (Example: factor 5, horizontal resolution of T2 * imaging: 2mm, resolution of the reconstruction space: 0.4mm) and to which the angiographic data set was linearly interpolated.

The 3D-reconstructed data is divided into clusters according to identity and surface content. The goal is to distinguish between real objects and noise-related ones

Artifacts. To do this, the triangulation of the surface returned by the marching cube algorithm is used to generate a unique vertex list. Starting with The accessibility of other vertices via the triangular network can be checked for an arbitrary vertex until the selection of vertices not yet reached is exhausted and a cluster is thus generated. The method is applied to the vertices which are not yet clusters until all vertices are assigned to a cluster. The clusters are sorted by surface area and the largest representatives select the arterial segment 30. FIG. 3C shows the result of a classification according to the size of the segments where noise-induced artifacts are removed. FIG. 3D shows a representation of a selected arterial segment 30, with which the further analysis is carried out.

The determination of the volume fraction per voxel of the reconstruction space is then carried out for the segment 30 on the basis of the positional relationship to the cluster (inside / outside) and then its combination in voxel groups perpendicular to the layer plane. FIG. 4A shows the subdivision of an arterial segment in voxel groups perpendicular to the layer guide. For each of these groups, the volume fraction of the vessel is determined on the total volume. The volume fraction is thus mapped in two dimensions with high resolution.

Subsequently, this volume distribution, which has a high resolution in the slice direction, is superimposed with a representation of T2 * measurement data at an early point in time at which the applied contrast agent has just reached the arteries and the capillary bed as a rule does not yet contain any contrast agent. 4B shows a T2 * weighted image consisting of much larger voxels. The common feature at this time with the MR angiography of the exclusive representation of the arteries is used for the position correction of the angiographic reconstruction with the T2 * -weighted image data.

For this purpose, the position of the 3D reconstructed vessel is varied with the goal of maximum linear correlation between concentration and volume fraction in the slice direction. Fig. 4C shows the T2 * weighted image of Fig. 4B superimposed with the volume distribution of the arterial segment. Fig. 4D shows the result of optimizing the position. The aim here is a maximum correlation coefficient between the relative concentration calculated from T2 * images and the volume fraction of the arterial segment per T2 * voxel. The determined Lage¬ optimum is used in the following for the calibration of the blood volume. For this purpose, concentration and volume are evaluated at a late point in time in which the concentration of the contrast agent is completely balanced. Assuming that the contrast agent concentration at this time has balanced itself throughout the bloodstream, the measured concentration is proportional to the blood volume within a voxel. Outside the reconstructed segment, it is in the form of small vessels up to the capillary bed. This portion is not accessible to measurement by MR angiography and provides a random contribution to each voxel. This compensates itself with the observable blood volume located in the reconstructed vessel segment to form a total blood volume per voxel. The T2 * voxels penetrated by the selected arterial segment with different volume fractions serve in the final step of the analysis to calibrate the blood volume.

For this purpose, the calculated volume fractions per T2 * voxel in the vicinity of the arterial segment 30 are plotted against the relative concentration per T2 * voxel. FIG. 5 shows a relationship between the relative concentration and the calculated volume fractions per T2 * voxel in the case of balanced contrast agent concentration within the blood vessels. Each circle of the graph marks a voxel. The measured concentration is composed of indeterminate portions outside the arterial segment modulated with the volume fractions of the segment varying between the voxels. The base of the line is represented by the T2 * voxels in which no part of the segment lies. The increase in the straight line represents the conversion factor between the measured, relative concentration of the contrast agent at a late time of the measurement and the blood volume (without consideration of hematocrit corrections or the like) and can subsequently be applied globally to the entire T2 * -weighted data set for calibration become.

The method has been described in detail with reference to a T2 * weighted imaging method and measurement data from an MR angiography measurement. In principle, the method for calibration can be used for any contrast agent-assisted perfusion imaging, it also being possible to use image representations of measured data from other measuring methods for determining the geometric parameters of the arteries if these allow coregistration.

The features of the invention disclosed in the foregoing description, in the claims and in the drawing may be of importance both individually and in any combination for the realization of the invention in its various embodiments.

Claims

claims

A method of calibrating a contrast agent-based perfusion imaging for a portion of a subject's blood vessel circuit, comprising:

an image representation of contrast agent measurement data of the contrast agent-based perfusion imaging for the section of the blood vessel circuit is coregistered with an image representation of angiography measurement data of an angiography measurement of the section of the blood vessel circulation, by switching between the image representation of the contrast measurement data and the image representation the angiographic measurement data is formed an optimized overlap of one or more arteries in the portion of the blood vessel circuit, whereby picture elements of the image representation of the contrast agent measurement data and image elements of the image representation of the angiography measurement data are mutually assignable; and

a calibration measure for the relationship between a signal intensity of voxels in the image representation of the contrast-agent measurement data and a blood volume associated with the voxels is then determined by selecting from the angiography

Measured data derived, geometric parameters for one or all arteries in the Ab¬ section of the blood vessel circuit are assigned to the voxels from the contrast agent measurement data of the contrast agent-based perfusion imaging.

2. The method of claim 1, characterized g ek enn e z ei chn et that angiography measurement data of an MR angiography measurement for coregistering the angiographic measurement data and the contrast agent measurement data are used.

3. The method of claim 1 or 2, characterized gek ennz eic hn et that contrast agent measurement data of a T ₂ * -weighted imaging for coregistering the angiographic measurement data and the contrast agent measurement data are used.

4. Method according to one of the preceding claims, characterized in that the geometric parameters are derived from the angiographic measurement data by means of a 3D reconstruction of one or all of the arteries in the section of the blood vessel circuit.

5. The method according to claim 4, characterized in that a marching cube algorithm is used in the SD reconstruction.