Method for correcting an X-ray image, processing device, X-ray equipment, computer program and data carrier

Abstract

Computer-implemented method for correcting an X-ray image (1) based on imaging by an X-ray device (2) with an X-ray source (3) and an X-ray detector (4) during a first time interval (6) and specifying a respective X-ray image value (8) for at least one image point (7), comprising the steps: - Receiving the X-ray image (1), - Receiving several dark images (9), wherein the respective dark image (9) is based on an image data acquisition by the X-ray detector (4) during a respective sub-interval of a second time interval (10) preceding the first time interval (6), during which no X-ray radiation is irradiated onto the X-ray detector (4), and indicates a respective dark image value (11) for the respective image point (7), - Predicting a respective night luminescence value (12) for the respective pixel (7) in the X-ray image (1), depending on the dark frame values ​​(11) of several of the dark frames (9) for the respective pixel (7), - Correcting the X-ray image (1) by determining a respective corrected X-ray image value (14) for the respective image point (7) depending on the respective X-ray image value (8) and the respective afterglow value (12).

Description

[0001] The invention relates to a computer-implemented method for correcting an X-ray image, which is based on imaging by an X-ray device with an X-ray source and an imaging X-ray detector during a first time interval and specifies a respective X-ray image value for at least one pixel. The invention also relates to a processing device, an X-ray device, a computer program, and a data carrier.

[0002] In X-ray-based medical imaging, digital flat panel detectors are most commonly used according to current technology. Depending on their design, these can exhibit varying degrees of afterglow. This results from charges being deposited by prior X-ray irradiation, for example, in defects of the scintillator or the sensor's photodiode. These charges are only released over time, thus influencing image values ​​at later points in time. The effect is particularly pronounced when a low-dose image follows a high-dose image. In this case, for example, strongly exposed structures from the high-dose image can be depicted in the low-dose image even if they are not actually present in the imaged area.

[0003] Currently, for example in fluoroscopy applications, ongoing exposure series are repeatedly interrupted to capture a dark frame, which is then used to correct the X-ray images taken in the time interval surrounding it. However, this approach has several disadvantages.

[0004] Firstly, the dark images contain electrical noise, which means that subtracting the dark image worsens the signal-to-noise ratio in the corrected images, potentially requiring higher X-ray doses and thus increasing the patient's exposure.

[0005] Secondly, acquiring dark frames requires time that is lost for the actual exposure, or necessitates a temporary reduction in the pulse frequency of the exposure pulses. Therefore, applying this correction approach leads to lower dose efficiency and / or a reduction in the temporal resolution of the imaging.

[0006] From the publication Starman, J., et al. (2012), A nonlinear lag correction algorithm for a-Si flat-panel x-ray detectors. Med. Phys., 39: 6035-6047, https: / / doi.org / 10.1118 / 1.4752087, it is known to correct for afterglow from previous images during an X-ray sequence by unfolding the measurement data. It is proposed to use a nonlinear model of afterglow, where the weighting and decay rate of previous image values ​​depend on the respective X-ray dose applied. This approach can at least largely compensate for afterglow on a short timescale, as typically occurs with low to medium X-ray doses. However, longer afterglow on a timescale of several minutes or even more than an hour often occurs in conjunction with saturation of the detector element and thus of the image value.Since the actual radiation dose cannot be clearly reconstructed during such saturation, long-lasting afterglow effects cannot usually be compensated for by the described unfolding, or at best only to a small extent.

[0007] The invention is therefore based on the objective of providing an improved approach to the correction of afterglow in X-ray images.

[0008] The task is solved by a computer-implemented method of the type mentioned above, which comprises the following steps: - Receiving the X-ray image, - Receiving multiple dark images, wherein the respective dark image is based on a respective image data acquisition by the X-ray detector during a respective sub-interval of a second time interval preceding the first time interval, during which no X-ray radiation is emitted by the X-ray source onto the X-ray detector, and indicates a respective dark image value for the respective pixel. - Predicting a respective night luminance value for the respective pixel in the X-ray image, which is expected as the X-ray image value in the case that no X-ray radiation from the X-ray source falls on a respective detector element of the X-ray detector assigned to the respective pixel even after the second time interval, depending on the dark frame values ​​of several of the dark frames for the respective pixel, - Correcting the X-ray image by determining a respective corrected X-ray image value for the respective image point depending on the respective X-ray image value and the respective afterglow value.

[0009] By considering several dark frames preceding the same X-ray image, the temporal progression of the afterglow decay, and thus the afterglow value at the time or within the time interval of the next image acquisition, can be predicted, for example by software, and used for correction. This approach is particularly useful because the X-ray source is often deactivated for a certain period of time between imaging sequences or between separate sub-sequences of a longer imaging sequence—for example, several seconds, several tens of seconds, or even more than a minute.During such an interval, a large number of dark frames can be acquired, allowing the temporal evolution of the image values ​​of individual pixels within the dark frames, and thus the decay of the afterglow, to be characterized with high accuracy without requiring any modification of the imaging sequence itself. Although, as will be discussed later, various processes are often involved in the afterglow of the X-ray detector, meaning that the decay of the afterglow depends on the detector's history or the amount of X-ray radiation applied to it, the prediction can still be made with good accuracy.

[0010] X-ray image correction can be achieved, for example, by subtracting the respective afterglow value from the corresponding image value. However, it is also possible to scale the image value based on the afterglow value, or to determine the corrected image value as a linear or non-linear function of the image value and the afterglow value, for example, to account for non-linearities in the imaging process. Alternatively, an iterative correction method can be used. It is also possible to consider the afterglow value, or afterglow values ​​determined for successive X-ray images, within a deconstruction to calculate the actual radiation dose delivered to each detector element during the acquisition of the respective image. For this purpose, the method described in the publication by Starman, J., et al., can be used, as will be explained later., known algorithm can be extended to take into account the dark frame values ​​or information about the state of the X-ray detector determined on the basis of the dark frame values.

[0011] If multiple X-ray images within an X-ray sequence need to be corrected, in the simplest case, the influence of the X-ray dose emitted during this sequence on the afterglow of the X-ray detector can be neglected. This allows the afterglow value to decay further according to the decay behavior already determined for its initial calculation. Such an approximation can be sufficient and appropriate, for example, when relatively low X-ray doses are emitted within the sequence, such as during fluoroscopy, while the afterglow of a preceding image acquired with higher doses needs to be compensated for. However, as will be explained in more detail later, X-ray doses emitted during a subsequent X-ray sequence can also be considered to further improve the correction.

[0012] In particular, the X-ray image and its corresponding dark frame comprise multiple pixels, for example, 128x128 or 256x256 pixels. Specifically, each pixel of the X-ray image or dark frame is assigned to a corresponding detector element or pixel of the X-ray detector. The X-ray image is acquired while X-rays from the X-ray source are directed onto the X-ray detector or at least onto parts of the detector elements of the X-ray detector. The X-ray radiation incident on the X-ray detector may be attenuated by the object being imaged, such as a patient. The X-ray image can be, in particular, a medical imaging image, such as a single X-ray image, an image from a fluoroscopy sequence, or a projection image from a computed tomography scan.

[0013] X-ray radiation can always reach the X-ray detector when the X-ray source is active. However, it is also possible for the radiation to reach the X-ray detector to be interrupted, for example by a shutter and / or by a relative tilting and / or shifting of the X-ray source and X-ray detector, while the X-ray source is active.

[0014] The night luminance value can be determined for the respective pixel as a function of at least one respective derivative value of a first and / or second time derivative of a temporal image value profile, whereby the respective temporal image value profile is specified as a function of the dark frame values ​​for the respective pixel.

[0015] It was recognized that a particularly robust prediction is possible when first and / or second time derivatives of the temporal image value history, which describes the dark frame values ​​of the respective pixel, are used. This results from the physical principles of afterglow, which are explained below using a silicon detector as an example. However, other detectors or scintillators also exhibit corresponding charge-trapping mechanisms, so the following discussion can be applied to these systems, at least with regard to its relevance to the described method.

[0016] In crystalline semiconductors (CMOS), defects form monoenergetic states within the semiconductor's band gap. These states are either occupied or unoccupied at equilibrium, corresponding to the Fermi level. Charging and discharging processes, such as photon irradiation of the detector element, cause defect states to enter a non-equilibrium state, before returning to equilibrium, usually thermally (i.e., via phonon-assisted processes). In crystalline silicon, the number of defects is very low, resulting in CMOS-based detectors exhibiting a small afterglow, also known as lag. In contrast, amorphous silicon detectors exhibit pronounced sidebands, which can lead to a longer afterglow. Because these sidebands are formed from local lattice variations, the states are localized, and the charge carriers are therefore immobile.

[0017] Furthermore, the non-binding states of the bonding sp are typically located approximately in the middle of the band gap. 3 Hybrid orbitals (dangling bonds).

[0018] Since the defects in the center of the band gap have a significantly higher time constant than the localized sidebands or defects due to their higher energy levels, the decay of the afterglow cannot usually be described by a single, time-constant decay coefficient. Instead, depending on the irradiated dose and the initial occupancy state of the defects, different decay profiles result, each with a time-varying decay rate.

[0019] As will be explained in more detail later, by evaluating the multiple dark images, especially by considering the different time derivatives of the image value progression in the second time interval in which no X-ray irradiation occurs, it is possible to determine where and on which of the possible decay curves the current state of the detector element is located, so that future decay states and thus also the afterglow value at the time of the X-ray exposure can be robustly predicted.

[0020] In the simplest case, the dark frames can be acquired at fixed time intervals, so that the temporal image value progression for each pixel can be directly determined by a sequence of dark frame values ​​ordered according to the temporal order of the acquisitions. The time derivatives can then be calculated as the respective difference quotient. However, it can also be useful to determine the image value progression as an analytical function, for example by a fit, or by means of resampling.

[0021] As will be explained later, the temporal image value profile can include not only the dark frame values ​​but also one or more additional image values, for example, a respective reference image value from at least one reference image acquired before the second time interval, the acquisition of which will be explained later. If a second time derivative is to be evaluated, the temporal image value profile should include at least three dark frame values ​​or at least two dark frame values ​​and one additional image value.

[0022] Several possible decay behaviors can each describe a model image value profile for the dark frame values ​​of the respective pixel and / or a first and / or second time derivative of the respective model image value profile, whereby depending on the derivative value or at least one of the derivative values ​​for the respective pixel, one of the decay behaviors is selected and depending on the selected decay behavior, the afterglow value is determined.

[0023] Different X-ray doses applied to the individual detector elements, or different time profiles of the dose application, lead to different population distributions of the various defects, such as the defects in the center of the band gap and the localized sidebands described above. However, it has been recognized that it is typically sufficient to consider the decay behavior for different initial doses as possible decay behaviors, since for conventional X-ray detectors all relevant excitation histories result in a decay behavior that can also be modeled by an excitation profile resulting from a single excitation.As has already been explained, starting from a certain excitation state, this results in superimposed decay processes of varying speeds, which can be summarized into a time-varying decay rate and thus a specific decay behavior.

[0024] The decay rate over time, and thus the overall decay behavior, depends on the initial dose and the time course of the dose irradiation. This can be accounted for by considering different possible decay behaviors. Since the varying occupancy of different defect types at different energies leads to different shapes in the model image value curves, comparing the derived values ​​with the derivatives of the various model image value curves allows the decay behavior that most closely corresponds to the actual decay behavior resulting from the respective detector element to be identified and selected.

[0025] Since the image value progression may be subject to a constant offset, at least approximately, it is advantageous to rely exclusively or at least primarily on the derivative values ​​when determining the actual current state of the detector element. For example, the first time derivative can be used to determine which point in time on the respective model image value progression corresponds to the current state of the detector element, while the selected model image value progression can be determined based on the second time derivative.

[0026] The respective model image value progression, or the progression of its first and / or second time derivatives, can be explicitly defined by the decay behavior, for example as a table of values, or implicitly, for example by an analytical function that may depend on one or more parameters, such as the applied X-ray dose. In the simplest case, a possible decay behavior assigns a model value for the image value to each time point, or a derivative value to each time derivative. However, it is also possible that a respective probability distribution of the model values ​​or the derivative values ​​is assigned to the different times, for example by specifying a mean and a standard deviation for each.

[0027] It is theoretically possible to use the same possible decay behaviors for all pixels. However, since the properties of the different detector elements can vary, it can be advantageous to define separate possible decay behaviors for individual pixels or at least for several subgroups of pixels.

[0028] The potential decay behaviors can be determined, particularly through preliminary tests. For example, measurement data relating to the decay behavior of the image value of each pixel can be collected for different irradiated X-ray doses. The measured values ​​can then be used directly, for example, as a table of values, to describe the possible decay behaviors. However, it is also possible, for instance, to determine the probability distributions described above based on the measurement data from several measurements, or to parameterize a predefined model through one or more measurements, thus arriving at a semi-empirical model that specifies the possible decay behaviors. Alternatively, it would also be possible to provide such a model based solely on theoretical considerations without experimental data.

[0029] Using a parameterized model can be particularly useful for utilizing a quasi-continuous field of model image value curves, between which one or more parameters, especially dose, can be blended quasi-continuously. In this case, the number of decay behaviors considered is potentially limited only by the resolution of the at least one parameter that parameterizes the model.

[0030] In principle, the possible decay behaviors for each detector can be determined separately, for example, during calibration at or after its manufacture. However, it is typically sufficient to determine the possible decay behaviors for a specific type of X-ray detector and apply them to all X-ray detectors of that type, i.e., all X-ray detectors of a particular production series.

[0031] Optionally, aging of the X-ray detector can be taken into account, for example by updating the possible decay behaviors after a certain operating period, for example during maintenance, by remeasuring and / or by taking into account a measure of the operating period, for example the value of an operating hours counter, in order to select between different possible decay behaviors for specifying the time derivatives or to adjust a parameterization of the evaluation of the respective decay behavior.

[0032] The respective decay behavior can be based on a sequence of reference images acquired sequentially within a third time interval following the application of a predetermined X-ray dose to the first or subsequent X-ray detector(s). The sequence of reference images thus describes, for at least one detector element of the second or subsequent X-ray detector(s), a temporal change in a reference image value of the respective detector element due to the decay of the excitation of the detector element by the applied X-ray dose. Different predetermined X-ray doses can be applied to define separate decay behaviors.

[0033] By analyzing the data from the X-ray detector itself, the potential decay behavior for an individual X-ray detector can be determined. However, as already mentioned, it may be sufficient to determine the potential decay behavior for a production series of X-ray detectors or for a specific type of X-ray detector. Therefore, data from another X-ray detector can also be analyzed to determine potential decay behavior for the X-ray detector.

[0034] During the third time interval, no further X-ray dose is applied to the other X-ray detector(s). Thus, the decay behavior of the excitation is recorded separately for different applied X-ray doses and therefore for different excitation states. As explained above, this allows us to identify, based on at least one of the time derivatives, the decay curve and thus the model image value progression, which decay behavior most closely corresponds to the temporal image value progression in order to predict the night luminance value.

[0035] In the procedure described above, the sequence of reference images can also be determined multiple times for the same given X-ray dose, so that the temporal course of the reference image value or its derivatives can be statistically evaluated in order to parameterize a stochastic model or to determine a respective probability distribution for the derivative values ​​and / or other quantities for different times of the respective decay behavior.

[0036] The respective night luminance value can be determined as a function of at least one decay parameter, which is determined by optimizing a cost function for the respective pixel, whereby the respective derivative value for the first and / or the second time derivative of the temporal image value progression is determined for several time points in the second time interval, wherein a decay model as a function of at least one decay parameter - on the one hand, specifies a forecast value for the respective derivative value at the respective time, whereby the cost function depends on a measure for the deviations of the forecast values ​​from the derivative values, or - on the other hand, specifies a probability distribution for the respective derivative value at the respective time, whereby the cost function depends on the result of a likelihood function that indicates the probability of a joint occurrence of the respective derivative values ​​at the several times according to the specified probability distributions.

[0037] If, in the first case, the cost function comprises only the measure of the deviation and is minimized, this corresponds to minimizing the deviation of the time course of the forecast values ​​from the time course of the derivative values. If quasi-continuous decay parameters are used, this ultimately corresponds to fitting the decay model to the derivative values.

[0038] In the second case, using only the likelihood function as the cost function when maximizing the cost function corresponds to using the maximum likelihood method known from other application areas.

[0039] Additional information can be taken into account by incorporating additional dependencies of the cost function, for example by using a weighted sum where one summand depends on the measure of deviation or the likelihood function, and at least one further summand on other influencing factors, in particular, as will be explained in more detail later, image values ​​from an X-ray taken before the second time interval.

[0040] The decay model can be defined, for example, by the various possible decay behaviors described above. In this case, one of the decay parameters can select one of the possible decay behaviors, while another decay parameter selects a time offset on the model image value progression defined by the selected decay behavior.

[0041] The respective corrected X-ray image value can additionally be determined for the respective image point depending on a respective reference image value of a reference image, whereby the reference image is based on an imaging by the X-ray detector that takes place before the second time interval and is irradiated onto the X-ray detector by the X-ray source during the X-ray radiation.

[0042] The reference image value can be considered an approximate measure of the last X-ray dose applied before the start of the second time interval. Assuming that the preceding X-ray imaging did not involve excessively high doses and that the X-ray detector is not saturated, the reference image value can be used to determine the applied dose and thus select the corresponding decay behavior to predict the afterglow value. However, an afterglow value predicted using this relatively simple approach is typically subject to significant errors, as it neglects the detector element's history—that is, the X-ray dose applied before the reference image and the resulting presence of defects—as well as any potential saturation in the reference image.

[0043] The described procedure can therefore be modified such that, as explained above, one of several possible decay behaviors is first selected based on the reference image value. As soon as at least one dark frame has been acquired, the current first time derivative can be determined, in the simplest case as the difference quotient between the reference image value and the subsequent dark frame value, or between successive dark frame values. This first time derivative is then used to determine which point in time on the model image value progression defined by the selected decay behavior most closely corresponds to the current state of the detector element.In this process, the adjustment of the timing can be iteratively repeated with each recorded dark frame value, whereby the correction of the timing based on the first time derivatives can, for example, be scaled to align the assumed decay behavior with the actual decay behavior, for example, in the manner of a proportional control.

[0044] Once at least two dark frames have been acquired, the second time derivative of the image value can be used to potentially select a different decay behavior than the one previously assumed. As explained above, the different decay behaviors can be associated with different initial X-ray doses. For example, a deviation in the initial dose between the previously assumed decay behavior and the initial dose of the decay behavior selected based on the second time derivative can be scaled and added to the previously assumed initial dose to determine a new assumed initial dose and a corresponding decay behavior.

[0045] The described iterative approach can be particularly useful when an afterglow value needs to be available within a few dark frames with minimal effort to correct subsequent X-ray imaging. However, in many applications, dark frames can be acquired over relatively long periods, resulting in a large number of dark frame values ​​per pixel. In these cases, it may be advantageous to use one of the optimization approaches described above instead of the iterative approach. The robustness and accuracy of these optimization approaches can potentially be further improved if the cost function also depends on the reference image value. For example, the reference image value can be incorporated into the time-dependent image value profile.Alternatively or additionally, for example, a measure for the deviation of an initial dose assigned to the reference image value from an initial dose assigned to the decay behavior chosen within the optimization can be considered as an additional term in a cost function to be optimized which is formed by a weighted sum.

[0046] Within the framework of the inventive method, an additional X-ray image can be corrected, which is based on imaging by the X-ray device after acquisition of the X-ray image and which specifies a further X-ray image value for at least one pixel. The further X-ray image can be corrected by determining a further corrected X-ray image value for the respective pixel in the further X-ray image as a function of the respective further X-ray image value, the X-ray image value at the same pixel of the X-ray image, and on the one hand the respective afterglow value for this pixel of the X-ray image and / or on the other hand the dark frame values ​​of several dark frames for this pixel.

[0047] The X-ray image and the subsequent X-ray image can be part of an imaging sequence and, in particular, acquired in rapid succession, for example, as part of a fluoroscopy sequence. By additionally considering the X-ray image value at the same pixel of the X-ray image, afterglow due to a dose received during the preceding X-ray sequence can also be taken into account. To achieve this consideration even in an X-ray sequence with more than two X-ray images, the X-ray image values ​​of all previously acquired X-ray images, or a predetermined number of previously acquired X-ray images, can be considered for each subsequent X-ray image.

[0048] In the simplest case, correction based on the afterglow value or dark frame values ​​and correction based on previously acquired X-ray image values ​​within the same X-ray sequence can be performed independently. For example, a separate afterglow value can first be determined for each X-ray image in the sequence, depending on the dark frame values ​​and the temporal position of the respective X-ray image relative to the dark frames, and a correction can be performed based on this, as explained above for X-ray images. Subsequently, a correction of the afterglow can be performed based on a dose irradiated within the sequence, for which known algorithms can be used, such as the deconvolution method disclosed in the publication by Starman, J., et al. cited above.

[0049] However, it may be advantageous to perform a joint correction of both afterglow effects. For this purpose, one can, for example, start with the deconvolution algorithm used according to equations (12) to (16) in Starman, J., et al., which, when deconvolution of a respective X-ray image, takes into account the charge state q that existed before the acquisition of that X-ray image. n,kThe charge state of each detector element is taken into account. Since, as explained above, the evaluation of the multiple dark frames allows us to determine where and on which of the possible decay curves the current state of each detector element lies, the charge state is also at least approximately known or correlates in a known manner with the determined afterglow value. Thus, the initial charge state before the acquisition of the first X-ray image of the sequence can be determined by evaluating the dark frames, after which the deconvolution algorithm used in Starman, J., et al., can be used with this initial charge for the subsequent X-ray images.

[0050] During the correction of an X-ray image, or the creation of an X-ray image from raw data of the X-ray detector, and / or the correction or creation of a subsequent X-ray image, an offset correction of the respective X-ray image value or corrected X-ray image value can be performed based on a predefined offset value for the respective pixel. The offset value can be set to an updated value, particularly when an update condition is met. This updated value is determined based on the dark frame values ​​of a subset of dark frames for the respective pixel. The update condition for the subset can only be met if a further night luminance value, determined for the respective pixel for the earliest dark frame acquired in the subset based on the dark frame values ​​of several previously acquired dark frames, reaches or falls below a predefined threshold.

[0051] Offset correction of X-ray images, which can correct temperature-dependent and / or age-related offsets, for example, is a known technique. To determine or update the offset value, dark frame values ​​of dark images are already being evaluated. However, the described procedure ensures that only those dark frames are considered when determining the offset value in which, at least for the respective pixel under consideration, no or only a very slight impairment of the dark frame value due to afterglow from a previously applied X-ray dose is to be expected. The remaining afterglow value can be determined in the same way as explained above for the afterglow value of the X-ray image.

[0052] In addition to the method according to the invention, the invention relates to a processing device configured for carrying out the computer-implemented method according to the invention. The processing device can, for example, be designed as suitably programmed data processing equipment, or the aforementioned functionality can alternatively be implemented, at least partially, by hardwiring. The processing device can be integrated into a medical imaging device, in particular an X-ray device, or be designed separately from it. It can, for example, be implemented as a workstation, server, or cloud solution.

[0053] In addition, the invention relates to an X-ray device with an X-ray source and an imaging X-ray detector, which includes a processing device according to the invention.

[0054] By integrating a processing device according to the invention and thus implementing the method according to the invention in an X-ray device, the afterglow correction described above can be performed directly during data acquisition or visualization by the X-ray device for a user.

[0055] The invention also relates to a computer program with instructions designed to carry out the computer-implemented method according to the invention when executed on a data processing device.

[0056] Furthermore, the invention relates to a data carrier that includes the computer program according to the invention.

[0057] In the preceding description, regardless of the grammatical gender of a particular term, persons with male, female or other gender identities are to be included.

[0058] Further advantages and details of the invention will become apparent from the following exemplary embodiments and the accompanying drawings. These schematically illustrate: Fig. 1 a flowchart of an embodiment of the computer-implemented method according to the invention for correcting an X-ray image, Fig. 2 an embodiment of an X-ray device according to the invention, comprising an embodiment of a processing device according to the invention, Fig. 3. Determining various possible decay behaviors of the detector elements of the X-ray detector, which are located in the Fig. The methods shown in section 1 can be used, and Fig. 4 additional steps that are in the Fig. The procedure shown in 1 can be used to update an offset value used to correct the X-ray image or subsequent X-ray images.

[0059] Fig. Figure 1 shows a flowchart of a procedure for correcting an X-ray image. For example, it is assumed that the procedure is carried out by the [missing information]. Fig. The processing unit 36 ​​shown in Figure 2 is implemented and integrated into an X-ray unit 2. As already explained in the general section, such a processing unit 36 ​​can also be designed separately from the X-ray unit 2, for example as a server, cloud solution or workstation.

[0060] As detailed below using the exemplary examples, in Fig. As will be explained in the embodiment of the method shown in Figure 1, the X-ray image 1 to be corrected is based on an image acquired by an imaging X-ray detector 4 of the X-ray device 2 during a first time interval 6, during which X-rays were emitted onto the X-ray detector 4 or its detector elements 13 by an X-ray source 3 of the X-ray device 2. Several dark frames 9 are taken into account to correct the X-ray image 1. Each dark frame 9 is acquired by the X-ray detector 4 during a respective sub-interval of a second time interval 10, which precedes the first time interval 6 and during which no X-rays are emitted onto the X-ray detector 4.

[0061] For each pixel 7 in the X-ray image 1, a night luminescence value 12 is predicted based on the dark frames 9. This night luminescence value 8 is expected to be the X-ray image value if no X-ray radiation falls on the detector element 13 assigned to the respective pixel 7 even after the second time interval 10. Thus, the decay behavior of an afterglow due to a previously irradiated X-ray dose can be detected using the dark frames 9, and based on this decay behavior, an afterglow at the time of acquisition of the X-ray image 1 can be predicted. The X-ray image 1 can then be corrected based on the determined afterglow values ​​12.

[0062] As in Fig. As indicated by the vertical dashed line in Figure 1, the depicted procedure can be divided into two parts, which can, in principle, be implemented independently of each other, for example, at a considerable time interval and / or by separate devices. Steps S1 to S4 relate to data acquisition by the X-ray device 2 itself, while steps S5 to S14 implement the correction procedure and can, for example, be implemented by executing a computer program 38 implementing the procedure on a data processing device 37. However, it would also be possible, for example, for the X-ray images 1, 28 to be corrected immediately after their acquisition. In this case, steps S4, S8, and S14, which are carried out to acquire and correct the further X-ray images 28, could, for example, only be carried out after the other steps shown.

[0063] As in Fig. As shown schematically in Figure 2, the procedure is implemented in the example by means of a memory 39 storing a suitable computer program 38, whose instructions for implementing the procedure are executed by a processor 40.

[0064] In the Fig. In the example shown, step S1 first captures a reference image 27. This reference image 27 can be, in particular, the last X-ray image of an X-ray sequence completed before the acquisition of the X-ray image 1 to be corrected. The acquisition of the reference image 27, as well as the acquisition of X-ray images 1 and 28 (which will be explained later), takes place while the X-ray source 3 of the X-ray unit 2 is active, thus emitting an X-ray dose onto the X-ray detector 4 or its detector elements 13. In this example, an image of a patient 5 is created by radiation. The reference image 27 can depict the same patient 5 as the X-ray image 1 that is subsequently acquired and to be corrected; however, the reference image 27 and the X-ray image 1 can also depict different patients 5 or, more generally, different objects.

[0065] In step S2, several dark images 9 are then recorded during a second time interval 10, which lies before the first time interval 6, during which the X-ray image 1 is captured, whereby no X-ray radiation is emitted by the X-ray source 3 onto the X-ray detector 4 during the second time interval 6.

[0066] In step S3 the X-ray image 1 is then captured and in step S4 further X-ray images 28 to be corrected can be captured.

[0067] In steps S5 to S8, the data acquired in steps S1 to S4 is received. "Received" here can refer to making the data available to a specific function or algorithm within a computer program, but alternatively, for example, it can also refer to a transmission between different systems.

[0068] In step S5, the reference image 27 and thus the reference image values ​​26 for the individual pixels 7 of the reference image 27 are received. Similarly, in step S6, the dark frames 9 and thus the dark frame values ​​11 of the individual pixels 7 of the individual dark frames 9 are received. In step S7, the X-ray image 1 and thus the X-ray image values ​​8 of the individual pixels 7 of the X-ray image 1 are received, and in step S8, the further X-ray images 28 and thus the further X-ray image values ​​29 for the pixels 7 of the further X-ray images 28 are received.

[0069] In the example, the dark frame values ​​11 for each pixel 7 are arranged in the order in which their dark frames 9 were acquired, forming a temporal image value profile 17. In the example, the reference image value 26 for each pixel 7 is also placed before the dark frame values ​​11 in the temporal image value profile 17, so that the temporal image value profile 17 describes the decay of the afterglow after the acquisition of the reference image 27 or the exposure to the X-ray dose during a preceding imaging procedure.

[0070] As already discussed in the general section, significantly different image values ​​can occur due to constants or slowly changing offsets in otherwise identical decay processes. To enable better comparability of the temporal image value progression 17 with various possible decay behaviors 18, a first derivative value 15 (the first time derivative of the temporal image value progression 17) and a second derivative value 16 (the second time derivative of the image value progression 17) are determined in step S10 for several time points, in particular for the acquisition times of the dark frames 9.

[0071] In step S11, a decay model 25 is provided which, in the example, specifies a predictive value for the respective derivative value 15, 16 at the respective time point, depending on several decay parameters 23. In the example, the decay model includes several possible decay behaviors 18, each of which specifies a model image value progression 19 for the dark frame values ​​11 of the respective pixel 7. One way to determine such decay behaviors 18 will be discussed later with reference to Fig. 3 will be explained.

[0072] By means of first and second time derivatives, the respective model image value profile 18 can thus be compared with the determined profile of the derivative values ​​15, 16. In this configuration of the decay model 25, one of the decay behaviors 18 can be selected by one of the decay parameters 23, and a time offset between the respective model image value profile 19 and the temporal image value profile 17 can be set by another of the decay parameters 23.

[0073] In step S 12 of the example, a cost function 24 is minimized by varying the decay parameters 23, which depends on a measure of the deviations of the forecast values ​​from the derivative values ​​15, 16. In other words, by choosing one of the decay behaviors 18 and thus one of the possible model image value progressions 19, and by varying the time offset, a section of one of the model image value progressions 19 is sought that corresponds as closely as possible to the temporal image value progression 19 with respect to its derivatives.

[0074] The value of the selected model image value profile 19, which corresponds to the modified time of imaging of the X-ray image 1 according to the determined time offset, can then be used as the afterglow value 12.

[0075] As already described in the general section, it is also possible that the cost function 24 depends on other variables. Additionally or alternatively, instead of specifying model image values ​​or their derivatives based on the possible decay behaviors 18, a probability distribution can be specified for each time point in time, so that the decay parameters 23 can be determined, for example, using the maximum likelihood method. In a further alternative embodiment, the afterglow value 12 could be determined by iteratively selecting the decay behavior and adjusting the time offset after each acquisition or reception of a dark frame 9, as also already explained in the general section.

[0076] In step 13, the X-ray image 1 is then corrected by determining a respective corrected X-ray image value 14 for the respective image point 7 depending on the respective X-ray image value 8 and the respective afterglow value 12. For example, the respective afterglow value 12 can be subtracted from the respective X-ray image value 8 to determine the respective corrected X-ray image value 14.

[0077] In step S14, the subsequent X-ray images 28 are corrected by determining a further corrected X-ray image value 30 for each image point 7. This can be performed essentially according to a nonlinear, charge-consistent stored charge (NLCSC) deconvolution algorithm, as discussed in detail, for example, in the publication by Starman, J., et al. cited at the beginning. Based on the decay parameters 23 determined in step S12, it is possible to determine, at least approximately, the charge state in the detector element 13 assigned to each image point 7 before the acquisition of the X-ray image 1 and thus at the beginning of an X-ray sequence comprising the X-ray image 1 and the subsequent X-ray images 28.By taking this initial charge into account in the unfolding algorithm, medium-term and long-term afterglow effects, for example on a timescale of several minutes or even more than an hour, resulting from a dose exposure before the second time interval 10, can also be taken into account.

[0078] Fig. Figure 3 shows a possible procedure for determining the [relevant] in the procedure according to Fig. 1. Possible decay behavior used. 18. For the sake of simplicity, the example assumes that the determination takes place in the X-ray device 2 itself. However, as already discussed in the general part of the description, possible decay behaviors can also be determined for certain types or series of X-ray detectors 4 using an exemplary X-ray detector 4.

[0079] In step S15, a predetermined X-ray dose 22 is first emitted onto the X-ray detector 4 by the X-ray source 3. The X-ray dose 22 can be selected, for example, by setting an exposure time.

[0080] Within a time interval 21 following this irradiation, a sequence of reference images 20 is then acquired in steps S16 and S17. In step S16, each reference image 20 is acquired, and in step S17, it is checked whether a predetermined number of reference images 20 has already been acquired. After acquiring a sufficient number of reference images for the predetermined X-ray dose 22, the procedure continues in step S18.

[0081] In step S18, it is then checked whether reference images 20 have already been acquired for all desired X-ray doses 20. If this is not the case, the procedure continues from step S15 with the irradiation of a different X-ray dose 22.

[0082] Otherwise, in step S19, based on the respective sequence of reference images 20 for the respective specified X-ray dose 22, a respective possible decay behavior 18 is determined, in particular separately for each of the image points 7 or each of the detector elements 13. In particular, the respective decay behavior 18 for the respective image point 7 can be specified as a sequence of the reference values ​​that were recorded for this image point 7 in the respective sequence of reference images.

[0083] For the sake of clarity, a relatively simple procedure for determining the decay behavior 18 was discussed beforehand. To avoid or reduce the influence of image noise, the respective sequence of reference images 20 can, for example, be acquired multiple times, after which the decay behavior can be determined by statistical evaluation, for example by averaging, or, as already explained above, by which the different decay behaviors can define probability distributions for the different times. To avoid the influence of offsets, instead of a sequence of reference values, a first and / or second time derivative of these reference values ​​can be stored directly to describe the respective decay behavior 18.

[0084] As already mentioned, constant or slowly changing offsets can occur in the individual pixels when acquiring X-ray image data. These offsets can be compensated for by subtracting an offset value. It is advantageous to update the offset value during operation, for example, by averaging image values ​​from a large number of dark frames. One way to avoid the influence of medium- and long-term afterglow effects will be discussed below with reference to… Fig. 4 explained.

[0085] To check whether a subgroup 33 of the dark frames 9 selected in step S20 is suitable for updating the offset value 31 of a specific pixel 7, the following is performed in the Fig. The procedure shown in step S9 for this pixel 7 is as follows: Fig. 1. The temporal image value progression determined is used, whereby in particular only that part of the image value progression is evaluated which relates to the image value progression temporally before the recording of the first dark frame 9 of subgroup 33.

[0086] Based on this image value progression, for example, reference is made to the steps S9 to S12 already mentioned in Fig. 1 explained way, a night luminance value for the respective pixel is predicted at the time when the first dark frame 9 of subgroup 33 is captured.

[0087] Only if this afterglow value 34 is less than a predefined limit 35 is the update condition 22 fulfilled in step S21, allowing the offset value 31 to be updated in step S22. Otherwise, a different subgroup of dark frames 9 must be selected, or the update of the offset value 31 must be postponed, since the afterglow value 34 indicates that the dark frames 9 of subgroup 33 may be at least partially affected by afterglow, which would likely distort an offset value 31 determined based on these dark frame values ​​9. QUOTES INCLUDED IN THE DESCRIPTION

[0000] This list of documents cited by the applicant was automatically generated and is included solely for the reader's convenience. The list is not part of the German patent or utility model application. The DPMA accepts no liability for any errors or omissions. Cited non-patent literature

[0000] Starman, J., et al. (2012), A nonlinear lag correction algorithm for a-Si flat-panel x-ray detectors. Med. Phys., 39: 6035-6047, https: / / doi.org / 10.1118 / 1.4752087

[0006]

Claims

[1] Computer-implemented method for correcting an X-ray image (1) based on imaging by an X-ray device (2) with an X-ray source (3) and an imaging X-ray detector (4) during a first time interval (6) and specifying a respective X-ray image value (8) for at least one image point (7), comprising the steps: - Receiving the X-ray image (1), - Receiving several dark frames (9), wherein the respective dark frame (9) is based on a respective image data acquisition by the X-ray detector (4) during a respective sub-interval of a second time interval (10) preceding the first time interval (6), during which no X-ray radiation is emitted by the X-ray source (3) onto the X-ray detector (4), and indicates a respective dark frame value (11) for the respective image point (7), - Predicting a respective night luminance value (12) for the respective image point (7) in the X-ray image (1), which is expected as an X-ray image value (8) in the case that no X-ray radiation from the X-ray source (3) is incident on a respective detector element (13) of the X-ray detector (4) assigned to the respective image point (7) even after the second time interval (10), depending on the dark frame values ​​(11) of several of the dark frames (9) for the respective image point (7), - Correcting the X-ray image (1) by determining a respective corrected X-ray image value (14) for the respective image point (7) depending on the respective X-ray image value (8) and the respective afterglow value (12). [2] Computer-implemented method according to claim 1, characterized by, that the night luminance value (12) for the respective pixel (7) is determined as a function of at least one respective derivative value (15, 16) of a first and / or second time derivative of a temporal image value profile (17), wherein the respective temporal image value profile (17) is specified as a function of the dark frame values ​​(11) for the respective pixel (7). [3] Computer-implemented method according to claim 2, characterized by , that several possible decay behaviors (18) each describe a model image value profile (19) for the dark frame values ​​(11) of the respective pixel (7) and / or a first and / or second time derivative of the respective model image value profile (19), wherein depending on the derivative value (15, 16) or at least one of the derivative values ​​(15, 16) for the respective pixel (7) one of the decay behaviors (18) is selected and depending on the selected decay behavior (18) the afterglow value (12) is determined. [4] Computer-implemented method according to claim 3, characterized by , that the respective possible decay behavior (18) is based on a sequence of reference images (20) which are acquired successively within a third time interval (21) following the irradiation of a respective predetermined X-ray dose (22) onto the or a further X-ray detector (4), whereby the sequence of reference images (20) for at least one detector element (13) of the or the further X-ray detector (4) describes a respective temporal change of a reference image value of the respective detector element due to the decay of the excitation of the detector element (13) by the irradiated X-ray dose (22), wherein different predetermined X-ray doses (22) are irradiated for the specification of separate of the possible decay behaviors (18). [5] Computer-implemented method according to any one of claims 2 to 4, characterized by , that the respective night luminance value (14) is determined as a function of at least one respective decay parameter (23), which is determined by optimizing a cost function (24) for the respective pixel (7), wherein the respective derivative value (15, 16) for the first and / or the second time derivative of the temporal image value profile (17) is determined for several time points in the second time interval (10), wherein a decay model (25) as a function of at least one decay parameter (23) - on the one hand, specifies a forecast value for the respective derivative value (15, 16) at the respective time, whereby the cost function (24) depends on a measure for the deviations of the forecast values ​​from the derivative values ​​(15, 16), or - on the other hand, a probability distribution for the respective derivative value (15, 16) at the respective time point, wherein the cost function (24) depends on the result of a likelihood function that indicates the probability of a joint occurrence of the respective derivative values ​​(15, 16) at the several time points according to the given probability distributions. [6] Computer-implemented method according to any one of the preceding claims, characterized by , that the respective corrected X-ray image value (14) for the respective image point (7) is additionally determined as a function of a respective reference image value (26) of a reference image (27), wherein the reference image (27) is based on an imaging by the X-ray detector (4) which takes place before the second time interval (10) and is irradiated onto the X-ray detector (4) by the X-ray source (3) during the X-ray radiation. [7] Computer-implemented method according to any one of the preceding claims, characterized by , that a further X-ray image (28) is corrected, which is based on an imaging by the X-ray device (2) after acquisition of the X-ray image (1) and specifies a respective further X-ray image value (29) for at least one image point (7), wherein the further X-ray image (1) is corrected by determining a respective further corrected X-ray image value (30) for the respective image point (7) in the further X-ray image (28) depending on the respective further X-ray image value (29), the X-ray image value (8) in the same image point of the X-ray image (1) and on the one hand the respective afterglow value (12) for this image point of the X-ray image and / or on the other hand the dark image values ​​(11) of several dark images for this image point (7). [8] Computer-implemented method according to any one of the preceding claims, characterized by, that, within the framework of the correction of the X-ray image (1) or the creation of the X-ray image (1) from raw data of the X-ray detector (4) and / or the correction or creation of a subsequent X-ray image, an offset correction of the respective X-ray image value (8) or corrected X-ray image value (14) is carried out depending on a predetermined offset value (31) for the respective image point (7), wherein the offset value (31) is set to an updated value upon fulfillment of an update condition (32), which is determined depending on the dark frame values ​​(11) of a subgroup (33) of the dark frames (9) for the respective image point (7), wherein the update condition (32) for the subgroup (33) can only be fulfilled if a further night luminance value (34),the threshold value (35) is determined for the respective pixel (7) for the earliest recorded dark frame (9) of the subgroup (33) depending on the dark frame values ​​(11) of several previously recorded dark frames (9), whether it reaches or falls below a predetermined limit value (35). [9] Processing device, characterized by that it is equipped to carry out the computer-implemented method according to one of the preceding claims. [10] X-ray equipment comprising an X-ray source (3) and an imaging X-ray detector (4), characterized by that it comprises a processing device (36) according to claim 9. [11] Computer program with instructions configured to perform the computer-implemented method according to any one of claims 1 to 8 when executed on a data processing device (37). [12] Data storage devices, characterized by that it comprises the computer program (38) according to the invention.

Images