DE102008055727A1 - Method for reducing partial scan artifact in image data of phase-correlated cardio-computed tomography, involves correcting raw data using extracted datasets, and reconstructing image from corrected raw data - Google Patents

Abstract

The method involves calculating a dataset for full rotation scan from raw data of real partial scans or a partial quantity of the partial scans, in an initially simulated manner. Datasets for virtual partial scans with different start angles are extracted from the calculated dataset. The raw data is corrected using the extracted datasets corresponding to initially produced partial scans with the same start angles or a start angle offset about 180 degrees. An image is reconstructed from the corrected raw data. An independent claim is also included for a device for reducing partial scan artifact in image data of phase-correlated computed tomography.

Description

The The present invention relates to a method and an apparatus to reduce partial scan artifacts in image data of the phase-correlated Computed tomography (CT), in particular cardio-CT, in which the Image data from raw data of several real partial scans at different Starting angles are generated, from a data set for a full scan information for the reduction of the Partialscanartefakte is won.

In cardio-CT, partial circulation scans, also known as partial scans, performed to the temporal resolution of the image captures to increase. At partial scans are used instead of the usual Full-rotation scans, the data over an angular range of 2π aquirieren, the data only over a partial circulation recorded and evaluated the recording system. A partial scan covers only a limited angular range <2π, For example, an angular range of π plus the fan angle of the inserted X-ray beam. Due to the limited angular range arise in the However, reconstructed image CT value fluctuations, from the starting angle depend on the particular partial scan. These CT value fluctuations are called partial scan artifacts. Partialscanartefakte have a particularly detrimental effect on perfusion measurements on the myocardium because there is already a slight inaccuracy of CT values wrong perfusion data supplies.

To reduce partial scan artifacts, see Primak et al., "A technical solution to avoid partial scan artifacts in MDCT", Med. Phys. 34 (2007), No. 12, pages 4726 to 4737 , proposed to use the same starting angle for each of the partial scans. This is achieved in this publication by synchronizing the heartbeat of the pig used as the test object with the rotation of the recording system of the computer tomograph. However, this technique is not suitable for examination on human patients because of the obvious risks.

In Meinel et al., "Reduction of half-scan shading artifact based on full-scan correction", Academic Radiology 13 (1), pp. 55-62, 2006 , a method is described in which the partial scan artifacts are reduced due to the additional information from a full-turn scan. In addition to the partial scans for the actual measurement, a full-rotation scan is performed in this method, from the measurement data set then several virtual partial scans are extracted at different starting angles. From the data records of the full-rotation scan and the virtual partial scans, image data sets are generated in each case by reconstruction. Subsequently, differential data sets are generated by forming the difference between the respective image data sets of the virtual partial scans and the image data set of the full-rotation scan. These difference data sets are then used to correct the images reconstructed from the real partial scans. For this purpose, the start angles of the virtual partial scans closest to the starting angle of the real partial scan are searched for each real partial scan. By appropriate weighting of the associated differential data records, the image data record of the real partial scan is finally corrected. This provides an image data set with reduced partial scan artifacts.

This However, procedure can be found in the phase-correlated cardio-CT unsatisfactory due to the required full scan apply, since the acquisition of a derarten 360 ° full rotation scan due to the movement of the beating heart to motion artifacts in the picture leads.

The The object of the present invention is a method and to provide a device that can reduce partial scan artifacts also in image data of the phase-correlated cardio-CT.

The Task is with the method and the device according to the Claims 1, 2, 6 and 7 solved. advantageous Embodiments of the method and the device are the subject the dependent claims or can be the following description and the embodiment remove.

In the proposed method, a data set for a virtual full scan scan is first formed from the raw data of all real partial scans recorded with the CT scanner at a constant z position or from a subset thereof. The number of real partial scans for the artificial formation of this data set is preferably chosen so large that each projection of the new data set is formed by averaging over the corresponding projections of several partial scans. Real scans are scans that were actually recorded with the computed tomography scanner. A virtual partial scan record or a full-scale virtual scan record is understood to mean a record that is merely artificially composed of one or more data rates of other real scans. From the artificially generated data set for the virtual full-rotation scan, also referred to below as a full-rotation virtual data record, data sets for virtual partial scans with different starting angles are then extracted. The raw data of the real partial scans are respectively corrected with data sets that belong to artificially generated partial scans with the same or π shifted starting angle. From the corrected raw data sets of the real partial scans, an image is subsequently reconstructed in a known manner, which can then be displayed on an image display device.

In an alternative embodiment of the method, the correction takes place not on the basis of the raw data, but on the basis of the image data of the virtual full-scale scans and virtual partial scans. To For this purpose, image data is taken from the corresponding raw data reconstructed. The correction then also takes place accordingly in the picture space.

Under the raw data or raw data sets are in the present Patent application understood the data or records that from the X-ray detectors after digitization and optionally further conversion or normalization. Such data For example, they are also based on a sinogram. From these Raw data are by suitable image reconstruction, eg. By filtered back projection, which generates image data that eventually the user as an image, for example, as a sectional image shown become.

The proposed method differs from the method of Meinel et al. especially in that the raw data set of the full-scale scan is not obtained by a real full-rotation scan, but that a virtual full-rotation data set is calculated from the partial scans actually performed, on the basis of which the further correction takes place. Since the real partial scans were recorded phase-correlated, the virtual full-circle data set obtained thereby is not disturbed by the movement of the heart. Furthermore, there are no disturbances due to the influence of the contrast agent on the partial scan artefacts, since averaging over a larger number of partial scans, which were recorded at different times of the contrast medium flow, also averages over this influence.

at the implementation of the proposed method according to The first alternative based solely on raw data will be added reduces the computational effort required for the otherwise required Reconstructions from the virtual partial scan records is required. Since in the proposed method no full-rotation scan must be performed, but an artificial Full-rotation data set from the measured partial scan records The reduction of partial scan artifacts can also be generated be achieved in phased-array cardiac CT. Of course however, the method can also be phase-correlated with other ones Use CT applications.

The Starting angles of the virtual partial scans are chosen that they match the start angles of the real recorded partial scans or shifted by an angle of π. This way you can exactly one virtual dataset for each partial scan Partialscan be obtained, with which the raw data or image data corrected for the real partial scan. Of course are the virtual partial scans in the present process like this chosen to be in its orbit, i. H. the angular interval, over the data are acquired, coincide with the wrap angle of the real partial scans or add 2π to it.

The proposed device, for example, as part of a computed tomography may be formed, comprises at least one data memory for the processed and all during processing accumulating data and a computing unit with which the corresponding Calculations are performed. The device comprises Furthermore, an image reconstruction module, which consists of raw data image data reconstructed, as well as a correction module, which generates the virtual full-round and partial-scan records as well correcting the data records of the real partial scans.

The proposed methods and associated apparatus be in the following with reference to an embodiment in Connection with the drawings explained in more detail. Hereby show:

1 an example of the dependence of the CT values on the starting angle when performing partial scans;

2 an example of the procedure of the present method in a highly schematic representation;

3 a representation for illustrating the artificial formation of the virtual full-rotation data set from the partial scan records;

4 an example of the effectiveness of the proposed method for the reduction of partial scan artifacts; and

5 a highly schematic representation of the proposed device.

Performing perfusion measurements on the myocardium using dynamic, phase-correlated cardiac CT only uses measurement data recorded in the same phase of the heart. For the perfusion measurements, a contrast agent is injected, whose propagation over time is detected by the partial scans. Although the required temporal resolution is achieved by these partial scans, CT value fluctuations occur in the reconstructed CT images, which depend on the starting angle of the respective partial scan. This starting angle can not be freely selected due to the phase control. 1 shows for illustration the dependence of the CT values on the starting angle. Here, N = 12 partial scans of a water phantom were simulated at 12 different starting angles α _n . The mean values of an ROI (region of interest) are plotted on the y-axis and the starting angles on the x-axis. The CT scans of the partial scans (solid curve with crosses) show significant variations of up to ± 4 HU, leading to partial scan artifacts in the reconstructed images. The CT values of a full-scale scan, which are plotted as a dashed curve with circles in the graph for comparison, are constant and thus independent of the starting angle.

Various Factors, such as B. object areas with high attenuation coefficients or asymmetrical geometries, contribute to the one from the projection angle dependent stray radiation profile is generated. In one Full-rotation scan will scan the object from all projection angles irradiated, resulting in a uniform distribution of Scattered radiation over all angles possible. In In this case, the result is invariant under change of Start angle. In a partial scan, however, the data becomes one Angular range of π limited, what this invariance picks up and thus each Partialscandatensatz depends on Start angle assigns a specific scattered beam profile. These inconsistencies are the main cause of the unwanted partial scan artifacts.

at the proposed method becomes a virtual full rotation data set generated from the measured partial scan records. Thereby may be the time-varying CT values of the contrast agent be taken into account, as shown below. First, however, is based on the generation of real partial scans received.

In a perfusion study, the temporally varying CT values of an in-flowing contrast agent are examined at a constant z-position. In the phase-correlated cardio-CT, from a continuously recorded scan p ⁰ (α, ξ, γ), retrospective N partial scans p ^P _n (α, ξ, γ) are generated (n = 1,... N), which are selected in this way in that they are temporally in the nth diastole of the cardiac cycle. The p ^P _n (α, ξ, γ) correspond to sinograms with the projection angle α, the detector coordinate ξ and the cone angle γ. For reasons of clarity, the variables ξ and γ are no longer explicitly stated below. It is further assumed that a suitable fan-beam to parallel beam geometry rebinning has taken place. In this notation α is continuously counted, ie α ε [0; 2πk], where k is the number of rotations of the gantry performed during a perfusion study. Decisive for the determination of the reconstruction window is the starting angle α _n , from which over an angular range of π data are acquired. The α _n are chosen so that the reconstruction window comes to lie temporally in the n th diastole of the heart cycle. For the sake of simplicity, it is assumed here that the heart rate is low enough to obtain a complete π partial scan record within one diastole. Since the p ^P _{n are} successive, n increases in proportion to the scanned time t. The p ^P _n is obtained by projection multiplication of p ⁰ with a weighting function w (α - α _n ): p P n (α) = w (α - α n ) · P 0 (α) (1)

Here α denotes the projection angle in parallel beam geometry. The weighting function is chosen as follows:

Of the Factor 2 ensures suitable standardization.

This is a known method for obtaining the phase-corrected partial scan records. Of course, it is also possible by suitable triggering of the computer tomograph only then data record when the correct cardiac phase is reached, to the To obtain partial scan records.

For the subsequent calculations, a new projection angle θ ε [0; 2π] is introduced, which calculates over α = θ + 2πk. The variable k = 0, ... K counts the rotations of the gantry. Accordingly, for the starting angles α _n = θ _n + 2πk _n , where k _{n stands} for the k th rotation in which the starting angle α _n comes to lie. With this formulation, equation (1) yields: p P n (θ + 2πk) = w k (θ, θ n ) p 0 (θ + 2πk) (2)

Here we have defined for short: w k (θ, θ n ) = w (θ - θ n + 2π (Θ (θ n - θ - π) + k - k n )).

Θ (x) denotes the Heaviside step function, which allows a correct continuation of the weighting function for angles θ _n > π, which corresponds to the modulo operation, where Θ (x) = 0 for x <0 and Θ (x) = 1 for x ≥ 0.

An essential step of the present method is the creation of a full-scale artificial scan p ^F _n (θ). The p ^F _n (θ) are calculated by projection from the p ^P _n (θ) of the partial scans via averaging. For this purpose, the measured partial scan data sets along the interval [0; 2π], the starting angles θ _n corresponding to the respective angular positions in [0; 2π] are assigned.

In order to generate a full-scale synthetic data set, a number A = 2a + 1 of the obtained partial scans are added up and normalized according to their starting angle. The partial scans are additionally multiplied by a time weight g _{n '} , which will be discussed in more detail later. The nth full-scale artificial scan p F n (Θ) is calculated as follows:

stellt sicher, dass jede Projektion korrekt gemittelt wurde. Die Variable A muss so gewählt werden, dass W ≠ 0∀ϑ und außerdem muss 0 ≤ a ≤ N erfüllt sein. Je größer a wird, desto weniger ähnelt die Perfusionsinformation im n'-ten Partialscan der interessierenden Zeit n. Demzufolge werden die Partialscans mit dem zeitlichen Gewicht g_n' multipliziert, der diesem Sachverhalt Rechnung trägt: je größer |n – a|, desto kleiner g_n'. Die g_n' können z. B. als Gaußverteilung implementiert werden. Im einfachsten Fall setzt man einfach A = N und g_n' = 1, d. h. alle vorhandenen Partialscans tragen in gleichem Maße zur Mittelung bei.The normalization factor

Ensures that every projection has been averaged correctly. The variable A must be chosen such that W ≠ 0∀θ and also 0 ≤ a ≤ N must be satisfied. The larger a becomes, the less the perfusion information in the n'th partial scan is similar to the time of interest n. As a result, the partial scans are multiplied by the time weight g _{n '} which takes account of this fact: the larger | n - a |, the smaller g _{n '} . The g _{n '} can z. B. be implemented as a Gaussian distribution. In the simplest case one simply sets A = N and g _{n '} = 1, ie all existing partial scans contribute to the same extent to the averaging.

The creation of a full-scale artificial scan corresponds to averaging over time (n α t), which ensures that the projection value of the in-flow contrast agent has the same average size in all angular positions. The morphological information is not changed by the averaging, however, since the operations are performed at a constant z-position (constant position on the system axis of the computer tomograph) and the partial scan datasets used originate from the same temporal reconstruction window, ie diastole. By way of illustration, it can be said that the start angles θ _n , and thus also the partial scan rates p P n (Θ) along the interval [0; 2π] are arranged.

This is based on the 3 3 illustrates, by way of example, three partial scans having different starting angles θ ₁ , θ ₂ and θ ₃ . The graph illustrates the process leading to a full-scale synthetic data set p F n (Θ) (black circle) leads. The three partial scan records p P 1 (Θ) . p P 2 (Θ) and p P 3 (Θ) (Semicircles) each cover a π-range whose position from the position of the respective starting angle θ _n in the interval [0; 2π]. The generation of an artificial projection in p ^F _n (θ) is shown for an arbitrary angle θ and the case A = 3. The three projections (black dots) from the three partial scans are averaged into an artificial projection in p ^F _n (θ), shown as a rectangle. In the case shown, a fourth data record would still be needed to find the left, still open area in p F n (Θ) cover. In this way, an artificial full-rotation data set is obtained, which is indicated in the representation as a closed circle.

It was explained above that the partial scan artifacts are due to the missing data areas in the partial scans. The proposed method eliminates the inconsistencies in p P n (θ + 2πk) by filling in the missing angle ranges.

For this purpose, the missing areas are filled with data from the nth full-scale synthetic data record. The latter is weighted with a weighting function shifted by the range π. There p F n (Θ) to an angular range of only [0; 2π], k = 0 is selected for the shifted weighting function. p C n (θ + 2πk) below denotes the nth corrected partial scan data set:

The second term on the right can also be used as a virtual partial scan p V n (Θ) since it also only covers an angular range of π: p V n (θ) = p F n (θ) (2 - w 0 (θ, θ n )).

The factor ½ in Eq. (3) ensures that

A schematic overview of the method is in 2 shown.

For the reconstruction of the corrected data, the transformation between α and θ is reversed. The desired, corrected CT image f C n (R) , which can be used for perfusion examinations, z. B. calculated via a filtered rear projection:

Here R denotes the operator of the radon transformation in 3-D. A distinction is made between a backprojection R -1 / α with a definition range α ε [0; ∞] and R -1 / θ with a domain of definition θ Ε [0; 2π].

In an alternative embodiment variant, the virtual partial scans p V n (Θ) also be defined so that they have the same angular range p P n (θ + 2πk) cover, ie p V n (θ) = p F n (Θ) w 0 (θ, θ n ) , In this case, the virtual partial scan records become p V n (Θ) only from the artificial full circulation data set p F n (Θ) subtracted and the difference to the p P n (θ + 2πk) added: p C n (θ + 2πk) = p P n (θ + 2πk) + p F (θ) - p V n (Θ).

It is - as in the first case - assumed that the virtual as well as the original Partialscandatensät ze with the Partialscanartefakt are afflicted. If the summands are reconstructed as in Eq. (4), the result is f C n (r) = f P n (r) + f F (r) - f V n (r) ⇒ f C n (r) = f P n (r) - (f V n (r) -f F (R)).

Here describes f C n (R) the corrected partial scan image, f P n (R) the originally measured partial scan image, f V n (R) the virtual partial scan picture and f F n (R) the reconstructed full scale synthetic data set.

In other words, we obtain the pure nth partial scan artefact by f F n (R) from f V n (R) is deducted. This is possible because the morphology information disappears and the perfusion information in p ^F (θ) is the same for all angular positions. If you pull from f P n (R) the partial scan artifact, so f V n (r) - f F (R) , you get the corrected partial scan picture f C n (R) ,

The presented method was verified by means of a simulated dynamic head phantom. The graphic of the 4 shows the mean values of the ROI as a function of the time and thus the starting angle, which was used for the calculation of the partial scan. The starting angles result from a simulated heart rate of 70 bpm. The solid curve (+) shows the evaluation of the full-circle images, which serve as a reference and represent a setpoint curve. The dashed curve (x) represents the average values of the reconstructed partial scans. It can be clearly seen that the CT values of the reconstructed partial scans are subject to a large fluctuation. The mean deviation from the reconstructed full-scale scans is 2.5 HU. The dotted curve (o) describes the course of the CT values corrected by the method described above. The CT value fluctuations are significantly reduced. The mean deviation from the reconstructed full-rotation scans is only 0.4 HU. This results in an improvement of the mean deviation of 84%.

5 finally shows in a highly schematic representation of an exemplary apparatus for performing the method. This device can be, for example, an image processor with a memory unit 1 , a data store 2 and an image reconstruction module 3 be in which a module 4 is implemented to reduce the partial scan artifacts. This module 4 is designed to perform the above-described process steps for reducing partial scan artifacts. The corrected and reconstructed images of the partial scans can then be displayed on an image display device 5 represent.

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited non-patent literature

• Primak et al., "A technical solution to avoid partial scan artifacts in MDCT", Med. Phys. 34 (2007), No. 12, pages 4726 to 4737 [0003]
• Meinel et al., "Reduction of half-scan shading artifacts based on full-scan correction", Academic Radiology 13 (1), pp. 55-62, 2006 [0004]
• Meinel et al. [0011]

Claims (10)

Method for reducing partial scan artifacts in image data of the phase-correlated computed tomography, the Raw data of several real partial scans at different starting angles be generated, in which - from the raw data of all real ones Partial scans or one of a subset thereof first artificially a record for a full rotation scan is calculated, - from the artificially formed Record for the full scan data records for extracted virtual partial scans with different starting angles become, - the raw data of the real partial scans respectively be corrected with records that are too artificial generated partial scans with the same or shifted by π Start angles include, and - from the corrected Raw data of the real partial scans reconstructed one image at a time becomes. Method for reducing partial scan artifacts in image data of the phase-correlated computed tomography, the Raw data of several real partial scans at different starting angles be generated, in which - from the raw data of all real ones Partial scans or one of a subset thereof first artificially a record for a full rotation scan is calculated, - from the artificially formed Record for the full scan data records for extracted virtual partial scans with different starting angles become, - from the artificially formed data set for the full scan and the records for Virtual Partialscans each reconstructed image data sets become, - for each image dataset of a virtual Partialscans a difference data set between this image data set and the image data set of the full-rotation scan is calculated - out The raw data of the real partial scans is reconstructed one image at a time will and - the images of the real partial scans respectively be corrected with the difference records that belong to Artificially generated partial scans with the same starting angle belong. The method of claim 1, wherein for each Virtual partial scan set a differential record between this record and the artificially formed record is calculated for the full-scale scan and the raw data the real partial scans each with difference records be corrected, which leads to artificially generated partial scans belong with the same starting angle. The method of claim 1, wherein the correction This is done by missing for a scan range of 2π Data of the real partial scan with data of the artificially generated Partialscans be supplemented, one opposite the real partial scan by π shifted starting angle has. Method according to one of claims 1 to 4, in which the artificial formation of the data set for the full-scan is done by over all available projections of real partial scans is averaged at the same projection angle to an average Projection, and the averaged projections to education of the full scan scan data set become. Device for reducing partial scan artifacts in image data of phase-correlated computed tomography, which are generated from raw data of a plurality of real partial scans at different starting angles, with at least one arithmetic unit ( 1 ), a data memory ( 2 ), an image reconstruction module ( 3 ) and a module ( 4 ) for the reduction of partial scan artifacts, wherein the module ( 4 ) is designed to reduce partial scan artifacts such that it first artificially calculates a data set for a full-rotation scan from the raw data of all real partial scans or one of a subset thereof, extracts data sets for virtual partial scans with different starting angles from the artificially formed data record for the full-scale scan; and - the raw data of the real partial scans are respectively corrected with the data sets belonging to artificially generated partial scans with the same or π shifted starting angle, and wherein the image reconstruction module ( 3 ) is designed such that it reconstructs an image from the corrected raw data of the real partial scans. Apparatus for reducing partial scan artifacts in image data of phase-correlated computer tomography, which are generated from raw data of several real partial scans at different starting angles, with at least one arithmetic unit ( 1 ), a data memory ( 2 ), an image reconstruction module ( 3 ) and a module ( 4 ) for the reduction of partial scan artifacts, wherein the module ( 4 ) is designed to reduce partial scan artifacts such that it first artificially calculates a data set for a full-rotation scan from the raw data of all real partial scans or one of a subset thereof, extracts data sets for virtual partial scans with different starting angles from the artificially formed data record for the full-scale scan; For each image data set of a virtual partial scan, calculating a difference data set between this image data set and the image data set of the full-rotation scan; and correcting images reconstructed from the raw partial scans with the difference data sets corresponding to artificially generated partial scans having the same or similar starting angle; and wherein the image reconstruction module ( 3 ) is designed so that it - each reconstructs the image data sets from the artificially formed data set for the Vollum laufscan and the data sets for virtual partial scans and - reconstructs an image from the raw data of the real partial scans. Device according to Claim 6, in which the module ( 4 ) is designed for the reduction of partial scan artifacts such that it calculates a differential data set for each virtual partial scan set between this data set and the artificially formed data set for the full scan and corrects the raw data of the real partial scans respectively with differential data sets corresponding to artificially generated partial scans with the same starting angle belong. Device according to Claim 6, in which the module ( 4 ) is designed for the reduction of partial scan artifacts such that, for the correction for a scan area of 2π, it replenishes missing data of the real partial scan with data of the artificially generated partial scan which has a starting angle shifted by π relative to the real partial scan. Device according to one of claims 6 to 9, in which the module ( 4 ) is adapted to reduce partial scan artifacts such that it averages the artificial data set for the full scan across all available projections of the real partial scans at the same projection angle to obtain an average projection and the averaged projections to form the data set for the full scan then composed.