Reduction of artifacts in spectral computed tomography image data

Abstract

Embodiments of the present disclosure relate to reduction of artifacts in spectral computed tomography image data. A method for reducing artifacts in spectral computed tomography image data, the method comprising: (a) receiving computed tomography measurement data having sets of measured values, where each set corresponds to a detector element readout of the measurement data, the measurement data comprising measured values from at least two different x-ray spectra; (b) determining an error estimate for each set of measurements; (c) defining a replacement reference value for each set of measured values; (d) applying a function that maps the measured value to a correction value, the correction value being between the mapped measured value and the replacement reference value of the corresponding set, such that for a larger error, the correction value is closer to or located at the replacement reference value, and for a smaller error, the correction value is closer to or located at the measured value; and (e) reconstructing computed tomography image data based on the correction value.

Description

Technical Field

[0001] The invention relates to a method for reducing artifacts in spectral computed tomography image data, a method for providing spectral computed tomography image data, a computer program, a computer-readable storage medium and a computed tomography system. Background Art

[0002] In spectral computed tomography (CT), x-ray attenuation measurements from multiple x-ray spectra (e.g., from two different x-ray spectra) are collected from an object of interest. For example, to generate measurements with different x-ray spectra, multiple measurements with different x-ray source voltages or beam filtering can be used. Additionally or alternatively, energy-resolving detectors such as photon counting or dual-layer detectors can be deployed. The measured x-ray attenuation depends on the material being scanned and the x-ray spectrum being applied. Therefore, the difference between the attenuation measurements obtained for different x-ray spectra can be used to distinguish between materials that have the same x-ray attenuation for a single x-ray spectrum.

[0003] The mathematical problem of decomposing the x-ray attenuation measured for different spectra into a basis suitable for clinical application, such as water and iodine, is called material decomposition. Errors that occur before material decomposition may be caused by physical effects such as scattered radiation, extrafocal radiation, or very low signal counts in low-dose situations. Errors may also be caused by imperfect data correction steps that are intended to correct for physical effects that occur during data acquisition. Typically, material decomposition does not reduce these errors. On the contrary, material decomposition may even increase the errors. Therefore, any errors in the data or image before material decomposition will be amplified and may cause image artifacts. Image artifacts can include, for example, streak artifacts, cupping artifacts, and a reduction in the contrast-to-noise ratio of the image.

[0004] In state-of-the-art CT systems, to minimize errors in the data before material decomposition, significant effort is typically invested in calibration and computational steps to correct for physical effects as accurately as possible.

[0005] Furthermore, the data acquisition modes available for spectral acquisition can be limited to reduce physical effects that contribute to image artifacts. For example, X-ray beam collimation can be reduced to reduce scattered radiation. However, this can have a direct negative impact on clinical application: less collimation typically results in longer examination times and a higher risk of motion artifacts. Longer examination times can also increase the amount of contrast agent required for the examination. Summary of the Invention

[0006] Therefore, an object of the present invention is to provide a means, in particular a method, to reduce the occurrence of image artifacts.

[0007] This object is met or exceeded by a method according to claim 1, a method according to claim 12, a computer program according to claim 13, a computer-readable storage medium according to claim 14, and a computed tomography system according to claim 15. Further advantages and features result from the dependent claims, the description, and the drawings.

[0008] According to a first aspect of the present invention, there is provided a method for reducing artifacts in spectral computed tomography image data. The method comprises:

[0009] (a) receiving computed tomography measurement data having sets of measurement values, wherein one set corresponds to measurement data read out from one detector element, the measurement data comprising measurement values from at least two different x-ray spectra;

[0010] (b) determining and / or receiving an error estimate for each set of measurement values;

[0011] (c) defining, for at least some of the sets, a replacement reference value for each set of measurement values;

[0012] (d) applying a function that maps the measured values to correction values, the correction values being between the mapped measured values and the corresponding set of alternate reference values, the distance of the correction values from each value being dependent on the error estimate, such that for larger errors the correction values are closer to or at the alternate reference value and for smaller errors the correction values are closer to or at the measured value;

[0013] (e) Reconstructing the computed tomography image data based on the correction values.

[0014] Advantageously, this method can reduce artifacts, particularly due to the step of mapping the measured values to the correction values. For example, artifacts including streak artifacts can be reduced. This allows, for example, more challenging modes such as dual-source acquisition or large collimation to be used without introducing excessive artifacts. This method can be particularly advantageous if data errors occur in a relatively small portion of the total measurement data. Therefore, preferably, only some of the data, particularly a small number of data, contain significant errors. For example, this method can be particularly advantageous if less than 10% of the measurement data contains significant data errors. A small portion of the data may be so erroneous that, without correction, attempting to perform material decomposition may only introduce artifacts to the image. It has been found that even a very small number of measurement values affected by errors can significantly affect image artifacts across a large portion of the image. In particular, even if far less than 10% of the data may be affected by errors, significant artifacts may still appear in the image. For example, only measurement data associated with specific x-rays from a specific direction may be affected. Advantageously, this method can avoid the effect of a few erroneous measurement values negatively impacting the entire image. On the other hand, since these very small amounts of data are relatively rare, they may be relatively insignificant for the spectral decomposition itself, and therefore sufficient spectral information may still be retained.

[0015] The term "computed tomography (CT) measurement data" is to be understood broadly in the context of the present invention. It generally describes data comprising measurement values acquired using a computed tomography system. In the context of the present invention, X-ray intensity measurement data may specifically be X-ray intensity measurement data and / or attenuation measurement data. For example, attenuation measurement data can be derived from X-ray intensity measurements by applying a negative arithmetic, such as μ = -logI / I0, where μ is the attenuation, I is the X-ray intensity, and I0 is the X-ray intensity measured in the absence of an attenuating object. The method according to the present invention is generally applicable to both X-ray intensity measurement data and attenuation measurement data. The measurement values can be acquired by transmitting x-rays from an x-ray source, particularly an x-ray tube, to at least one x-ray detector. For example, the at least one x-ray detector may comprise a row and / or an array of x-ray detectors. Typically, an object is positioned between the x-ray source and the at least one x-ray detector, such that the x-rays are attenuated before reaching the detector, depending on the material of the object, such as tissue material. For example, the object may be a portion of a patient. Typically, the x-rays generated by the x-ray source comprise a spectrum of x-ray frequencies. Depending on the specific spectrum of the x-rays and the materials in their path, the x-rays can be attenuated differently. Therefore, recording the attenuation of x-rays for different spectra can help distinguish between different materials.

[0016] In the context of the present invention, measurement values can be grouped together to form measurement value sets. A set corresponds to a detector element readout of measurement data that includes measurement values from at least two different x-ray spectra. The term "detector element readout" should be understood broadly in the context of the present invention. It generally describes a portion of at least one x-ray detector of a CT system. A detector element can correspond to a group of detector pixels, or preferably, to a single detector pixel. The detector element readout can particularly be a detector pixel readout. For example, at least one x-ray detector can be an energy-resolving detector, such as a photon counting detector or a dual-layer detector. The at least one energy-resolving detector can be used to detect different x-ray spectra. Additionally or alternatively, different x-ray spectra can be generated, for example, by applying different x-ray source voltages and / or different beam filtering. Because these sets include measurement values from at least two different x-ray spectra, spectral data can be reconstructed, which can enable better differentiation of certain materials. For example, receiving computed tomography measurement data can include retrieving data from a database and / or directly receiving data during a CT scan.

[0017] For a set of measurement values, an error estimate is determined and / or received. In particular, an error estimate can be determined and / or received for each set of measurement values. Thus, each detector element readout may be associated with an error. For example, an error estimate can be determined and / or selected such that the error estimate for one of the measurement values of the set is selected, which error estimate yields the highest error estimate among the error estimates for the measurement values of the set. For example, the error estimate can be stored in a database, such as together with the measurement data. In this case, the error can be retrieved together with the measurement data. For example, the error estimate can be estimated by an external processor and provided to the method. Alternatively, the error estimate can be determined during the method itself. The error estimate can be an error estimate relative to an error metric that defines the severity of the error. The error estimate can be an estimate that defines whether an error exists or whether there is no error. The error estimate can be defined based on an error metric. If the error metric is above a defined threshold, it can be defined that a substantial error exists.

[0018] For at least some of the measurement value sets, replacement reference values are defined. The replacement reference values can be defined based on the values of the measurement values in the set. Optionally, the definition of replacement reference values can be omitted for some of the sets, particularly for sets with an error estimate below a threshold or determined to have no substantial error. Thus, the following step of applying the function can optionally be applied only to the sets for which replacement reference values have been defined. The replacement reference values can be used as a reference for mapping the measurement values to the correction values.

[0019] In particular, a function is applied that maps measured values to correction values based on an error estimate and an alternative reference value. The output of the function is, in particular, a corresponding correction value. The correction value is between the mapped measured value and the alternative reference value. In the context of this function, "between" can also include the end values of the described range, i.e., the alternative reference value and the measured value. In other words, at least in some cases, the correction value can be the alternative reference value or the actual measured value. In particular, when the estimated error is low or essentially non-existent, the correction value can be set to the measured value. On the other hand, relatively large errors may result in the measured value being corrected to a different correction value. In particular, for a set with relatively large errors, each measured value in the set can be mapped to a correction value so that the correction value is the same as or more similar to the original measured value for the entire set. Since spectral decomposition is typically determined based on (relatively small) differences in measured values from different spectra, the differences are typically enhanced in the reconstructed image data, for example by multiplying by a relatively large factor. Therefore, errors in the measured values may result in large artifacts because the errors are greater than or much greater than the differences caused by applying different spectra. Advantageously, this method can therefore allow the impact of these errors to be reduced. On the other hand, since the mapping relies on errors, it is still possible to perform spectral reconstruction of regions associated with reliable data. While some of the true spectral content associated with high errors can be reduced in this way, it has been found that avoiding the introduction of artifacts is generally better for the final result than attempting to use spectral information from erroneous measurements. For larger errors, any true spectral content will generally be obscured by artifacts.

[0020] The correction values are then used as a basis for reconstructing the computed tomography image data. In particular, spectral computed tomography image data can be reconstructed. Thus, the method according to the present invention can provide a relatively simple method for treating unreliable data with less weight during spectral reconstruction, while still allowing for spectral reconstruction of regions associated with reliable data.

[0021] According to one embodiment, the replacement reference value is defined as the order of magnitude of the measured values of the corresponding set of measured values. The order of magnitude can be defined, for example, as a decimal or binary order of magnitude. Thus, for example, the order of magnitude can correspond to a power of 10 of the corresponding value. Advantageously, using the same order of magnitude can be a simple way to ensure that the correction value between the replacement reference value and the measured value also has the same order of magnitude as the measured value. As a result, the correction value does not deviate too much from the measured value.

[0022] According to one embodiment, the replacement reference value is defined as a linear combination of the measurement values of the set of measurement values.Using a linear combination may provide a relatively simple measure to ensure that the replacement reference value corresponds well to the measurement values of the set.

[0023] According to one embodiment, the coefficients of the linear combination are the same within each mapping set. Using the same coefficients can be a particularly simple way to determine the replacement reference value, without having to determine the coefficients for each set separately. Additionally or alternatively, the coefficients of the linear combination sum to substantially 1. In other words, the linear combination can be a convex combination. The convex combination can allow for a simple method to ensure that the order of magnitude of the measured value is the order of magnitude of the measured value of the corresponding measurement value set.

[0024] For example, the corresponding linear combinations can be defined as explained below. N} can represent the attenuation or X-ray intensity measurements of N different spectra from a single detector element readout (in particular a detector pixel readout). Thus, {S1, S2, ..., S N} are measurements from a set. For example, in the case of dual-energy CT, the list may be of the form {S1, S2}. The corresponding error estimate may be defined as w. The error may be defined within an error range, with different parts of the range corresponding to different severity of the error. For example, the error estimate may be in the range [0, 1], where 0 means no error and 1 means the error is so high that the corresponding set is not suitable for spectral processing. However, other error measures may also be applied in general. The reference value S may then be replaced by * Defined as from the set {S1,S2,…,S N} is a linear combination of the measurements in the measurement list. * In particular, the measured values {S1, S2, ..., S N}. So, for example, S * Can have the following forms

[0025]

[0026] where c k ≥0,k=1,…,N and

[0027] According to one embodiment, the function that maps the measured values to the correction values is a monotonic function. Applying a monotonic function can be a particularly useful method for selectively reducing the spectral content of a set of measured values with high errors. Preferably, the function is a smooth monotonic function. For example, the measured value S k The mapping can be done by a smooth monotonic function that transforms the measured value S k Mapping to S k With S * As defined above. For example, the function can convert the measured value S k Mapping to Correction Values The form is

[0028]

[0029] Where f(0;S k )=S k and f(1;S k )=S * Therefore, in this example and for an error w=1, the function can be replaced by the reference value S * Replace the measured value S k , so that the correction value Set to the corresponding replacement reference value S * On the other hand, in this example, for a high error w, the measured value S k is mapped closer to S * For low error w, the measured value S k is mapped closer to S k Therefore, for detector element measurements associated with high error estimates, the value of N The difference between the measured values of} is compared with the corrected value of the set The reduction or even complete elimination of this difference results in the associated errors in the measured values having less influence on the reconstructed image.

[0030] According to one embodiment, the replacement reference value is defined as one of the measured values of the corresponding measurement value set. Using one of the measured values as the replacement reference value may be a particularly simple method of generating the replacement reference value. This embodiment may correspond to a linear combination in which one coefficient is 1 and the other coefficients are set to zero. The replacement reference value can be selected so that the most reliable measurement value is selected. For example, the most reliable measurement value can be selected based on the estimated error. For example, a measurement value corresponding to a spectrum that has fewer artifacts or provides better statistical properties can be selected. Alternatively, the measurement value to be used as the replacement reference value can be randomly selected. This may be particularly simple while still achieving the main advantage of avoiding the increase in error caused by the difference between the measurement values of the calculation set. For example, in the case of large errors, all measurement values can be set to the value of one of the measurement values in the set. Therefore, in practice, certain unreliable data, i.e. data with large error estimates, can be regarded as non-spectral data because all correction values have the same value and therefore there are no spectral differences.

[0031] According to one embodiment, error estimation is based on determining errors in the data due to scattered radiation. Errors due to scattered radiation can be estimated using a scatter correction algorithm. For example, errors may be due to forward scatter or cross scatter reaching the x-ray detector. In a setup with two x-ray sources and detectors located on opposite sides of each x-ray source, due to cross scatter, x-rays may be detected by the wrong detector, i.e., a detector that is not opposite the corresponding x-ray source. Forward scatter is generally known to be present in all x-ray devices. Scattered radiation may result in additional signals in the detector elements.

[0032] According to one embodiment, an error estimate is determined based on the ratio, optionally the maximum ratio, between the estimated scattered signal and the total signal intensity of the corresponding measurement value. Maximum error can be understood as using the error of the measurement value in the set that produces the highest error estimate for the error estimate. Alternatively, other ratios besides the maximum ratio can be used. For example, a minimum ratio can be used. For example, a ratio corresponding to a specific energy and / or a specific spectrum in the set can be used. For example, two spectra can be acquired. In some cases, it may be advantageous to reconstruct an image that corresponds almost exactly to one of the spectra. In this case, it is best to use only the scattered / total ratio corresponding to that spectrum. The scattered signal can be understood as describing the portion of the total signal that is due to scattered X-ray radiation. For example, scattered X-ray radiation can include forward scatter and / or cross scatter. When the total signal is low and the scattered radiation is high, the corresponding error estimate is typically particularly high. This is often observed, for example, in X-rays passing through a patient's shoulder blades. Attempting to measure using the affected spectrum may result in severe streak artifacts along the X-rays passing through the shoulder blades. By applying artifact reduction methods, the severity of such artifacts can be reduced while leaving areas with smaller errors unaffected.

[0033] For example, the diffuse radiation can be estimated The ratio between the estimated scatter and total signal can be given by Definition, k = 1, ..., N, N is the number of spectra. This definition can be used to construct the error estimate

[0034]

[0035] Therefore, the maximum error is determined by selecting the k with the largest error, ie the error of the measurement values of the set with the largest error is thus determined.

[0036] Methods for estimating scattered radiation are generally known in the prior art. For example, scatter can be estimated by model-based profiling and / or by direct measurement of scattered radiation. The estimation based on the model-based profiling can include a model of the scattering process, which is used to estimate the scatter based on the measured signal. For example, the direct measurement can be performed by a dedicated scatter sensor for measuring scattered radiation. Examples of model-based estimation and direct measurement of scattered radiation are disclosed in the prior art, such as by Petersilka et al. in "Strategies for scatter correction in dual source CT" (Med. Phys. 37(11), November 2010, 5971-5992).

[0037] According to one embodiment, an error estimate is determined based on data errors in the low signal domain, such that a high error is assumed for low detector signals. For example, for low detector signals, statistical errors may make the measurement too unreliable to be used for spectral reconstruction. For low signals, the error estimate may be high, while for high signals, the error estimate may be low. For example, a particularly large patient may have a low signal domain due to a longer path of the X-rays through the patient's body. For example, some X-ray paths through the object being examined may result in complete or almost complete attenuation of the X-rays in this direction. For example, if the number of photons of a measurement is below a threshold, the measurement may be marked as erroneous. For example, a measurement based on a signal below the noise level may be determined to be unusable. For example, an error estimate for data errors in the low signal domain may be constructed in the following form

[0038]

[0039] The parameters and Defines the signal level range within which the measurement is considered unreliable, i.e., erroneous.

[0040] According to one embodiment, the mapping is defined based on at least two different error ranges of the estimate, such that a different partial mapping function is applied for each error range. For example, one error range may correspond to a measured value that is unchanged during the mapping, i.e., the corrected value is the same as the measured value, and a second error range may correspond to a measured value that is mapped closer to or located at a corresponding alternative reference value.

[0041] According to one embodiment, the error estimation includes applying a mapping based on at least three error ranges, the error ranges comprising: a first error range of low error, for which a correction value is set to the measured value; a second error range of medium error, for which the measured value is mapped so that the correction value is between the corresponding measured value and the corresponding replacement reference value; and a third error range of high error, for which the correction value is set to the replacement reference value of its corresponding set. The first error range may be below a first error threshold. The third error range may be above a second error threshold. The second error range may be between the first and second error thresholds. Specifically, the first error range may correspond to a lower error than the second error range, while the second error range may correspond to a lower error than the third error range. The first error range and the third error range may each be directly adjacent to the second error range. Optionally, there may be more than three error ranges. For example, there may be different types of media error ranges, such that different mappings are applied to different types of media error ranges.

[0042] Unless otherwise stated, the embodiments described herein may be combined with each other. For example, a replacement reference value may be defined via a linear combination, while a mapping may be defined via an error range.

[0043] According to another aspect of the present invention, a method for providing spectral computed tomography image data is provided. The method comprises

[0044] - Acquisition of measurement data by performing spectral computed tomography,

[0045] - performing the method described herein for reducing artifacts in spectral computed tomography image data.

[0046] All features and advantages of the method for reducing artifacts may be applied to the method for providing spectral computed tomography image data and vice versa.

[0047] According to another aspect of the present invention, a computer program includes instructions that, when executed by a computer, cause the computer to perform the method described herein, in particular the method for reducing artifacts in spectral computed tomography image data. All features and advantages of the method for reducing artifacts and the method for providing spectral computed tomography image data may apply to the computer program, and vice versa.

[0048] According to another aspect of the present invention, a computer-readable storage medium, in particular a non-transitory storage medium, is provided, comprising instructions that, when executed by a computer, cause the computer to perform the method described herein, in particular, a method for reducing artifacts in spectral computed tomography image data. All features and advantages of the method for reducing artifacts, the method for providing spectral computed tomography image data, and the computer program may be applied to the computer-readable storage medium, and vice versa. The storage medium may be any computer-readable storage medium. For example, the storage medium may be an optical storage medium or a solid-state storage medium. The storage medium may be a hard disk, a solid-state disk, a flash memory, an online server, or the like.

[0049] According to another aspect of the present invention, a computed tomography system is provided, comprising a processing unit configured to perform any of the methods described herein. All features and advantages of the method for reducing artifacts, the method for providing spectral computed tomography image data, the computer program, and the computer-readable storage medium may be applied to the computed tomography system, and vice versa. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The drawings illustrate various exemplary embodiments and methods of various aspects of the present invention.

[0051] Figure 1 A flow chart illustrating a method for reducing artifacts in spectral computed tomography image data according to one embodiment of the present invention is shown;

[0052] Figure 2 shows a mapping function according to one embodiment of the present invention;

[0053] Figure 3 A reconstructed spectral computed tomography image is shown, wherein the measured values have not yet been corrected according to the invention;

[0054] Figure 4 shows a reconstructed spectral computed tomography image based on the correction values according to the present invention;

[0055] Figure 5 A flow chart illustrating a method for providing spectral computed tomography image data according to one embodiment of the present invention; and

[0056] Figure 6 A computed tomography system according to one embodiment of the present invention is shown. DETAILED DESCRIPTION

[0057] Figure 1A flow chart of a method for reducing artifacts in spectral computed tomography image data according to an embodiment of the present invention is shown. In a first step 101, computed tomography measurement data having a set of measurement values is received. A set {S1, S2, ..., S N corresponds to one detector element readout of the measurement data and includes measurement values S from N different x-ray spectra k (k = 1, 2, ... N). N is at least 2, so there are at least 2 different x-ray spectra. In the case where each set has exactly two different x-ray spectra, the set has the form {S1, S2}. This specific example of two different x-ray spectra per set can correspond to dual-energy CT.

[0058] In a further step 102, an error estimate w is determined and / or received for each set of measurement values. For example, the error estimate may be based on determining the error of the data that is erroneous due to scattered radiation. For example, the estimated scattered signal may be used to estimate the error of the data. The total signal strength S of the corresponding measurement value k The error estimate of the scattered radiation is determined by the maximum ratio between . For example, such an error w can be determined according to the following terms:

[0059]

[0060] Additionally or alternatively, the error estimate w may be determined based on the data error in the low signal domain, such that a high error is assumed for low detector signals, such as based on the parameter and k = 1,…,N defines the range of signal levels within which the measurement is considered unreliable, for example:

[0061]

[0062] In a further step 103, for at least some of the sets, a replacement reference value S is defined for each set of measured values. * Alternatively, the replacement reference value can be defined such that it is a linear combination of the measured values of a set of measured values, for example as a convex combination

[0063]

[0064] Preferably, the coefficient c of the linear combination k The sum is basically 1, that is In one example, the coefficient c k One of the coefficients is 1, while the other coefficients are 0. Therefore, in this example, the replacement reference value is defined as one of the measured values of the corresponding set of measured values.

[0065] In a further step 104, the application converts the measured value Sk Mapping to Correction Values Correction value Between the mapped measurement value S k Replace the reference value S with the corresponding set * Between. Correction value and S * and S k The distance depends on the error estimate w, so that for larger errors, the correction value is closer to or located at the replacement reference value S * For smaller errors, the correction value Closer to or at the measured value S k For example, the function f(w; S k ), especially monotonic functions, can be used to convert the measured value S k Mapping to Correction Values For example, assuming that the error estimate ranges from 0 (indicating no error) to 1 (indicating excessive error), the function can produce f(0; S k )=S k and f(1;S k )=S * Alternatively, the mapping may be defined according to multiple error ranges of the error estimate, such that different partial mapping functions are applied for different error ranges.

[0066] Figure 2 An example of a mapping function with different error ranges 11, 12, 13 is shown in FIG. Figure 2 In the embodiment shown in , the error estimates range from 0% estimated error to 100% estimated error, where 100% error means that the corresponding set of measured values is completely unsuitable for spectral decomposition. Although in this representation the replacement reference value is higher than the measured value, in principle it can be higher or lower than the measured value. Function f(w; S k ) is divided into three ranges 11-13, a first range 11 below 80%, a second range 12 between 80% and 90%, and a third range 13 above 90%. In the first range 11, i.e. below 80% of the error estimate, the function maps the measured value so that the correction value is the measured value The measured value therefore does not change in this first range 11. In the second range 12, ie between the error estimates of 80% and 90%, the function maps the measured value so that the correction value At the measured value S k and replace the reference value S * Thus, in the second range 12, the measured values are corrected to a certain extent. In the third range 13, i.e. above an error estimate of 90%, the function maps the measured values so that the corrected values replace the reference values. Therefore, in the third range 13 the measured values are corrected to the greatest extent.

[0067] See also Figure 1 In a further step 105 , the computed tomography image data are reconstructed based on the correction values.

[0068] Figure 3 A reconstructed spectral computed tomography image is shown, wherein the measured values have not been corrected according to the invention. Figure 4 Spectral computed tomography images are shown, wherein correction values corrected according to the invention have been used for image reconstruction. Figure 3 In , an axial CT image from a dual-source chest examination is depicted, where several (white) streak artifacts are clearly seen horizontally across one scapula, ribs, and vertebrae. Figure 4 In the example shown in FIG. 5 , these artifacts have been largely eliminated due to the application of the method according to the invention.

[0069] Figure 5 1 shows a flow chart of a method for providing spectral computed tomography image data according to an embodiment of the present invention. In a first step 200, measurement data is acquired by performing spectral computed tomography. The subsequent steps 201-205 can be similar to those described in relation to Figure 1 The described steps 101 - 105 are equivalently applied.

[0070] Figure 6 A computer tomography system according to an embodiment of the present invention is shown. In this embodiment, the computer tomography system comprises a gantry 1, a movable patient bed 2 and a processing unit 3, which is configured to perform the method as described herein, for example, as described with respect to Figure 5 The method described.

Claims

1. A computer-implemented method for reducing artifacts in spectral computed tomography image data, the method comprising: (a) receiving computed tomography measurement data having sets of measurement values, wherein each set corresponds to a respective detector element readout and comprises measurement values of at least two different x-ray spectra; (b) determining and / or receiving an error estimate for each set of measurement values; (c) defining a replacement reference value for each set of measurement values; (d) applying a function that maps measured values to correction values, wherein for each measured value, a corresponding correction value is between the measured value and an alternative reference value of the set comprising the measured values, the distance of the corresponding correction value from each value depending on the error estimate, such that for larger errors, the corresponding correction value is closer to or located at the alternative reference value of the set comprising the measured values, and for smaller errors, the corresponding correction value is closer to or located at the measured value; (e) reconstructing the computed tomography image data based on the correction value.

2. The method according to claim 1, The replacement reference value of the set comprising the measurement values is defined to be of the same order of magnitude as the measurement values of the set comprising the measurement values.

3. The method according to any one of the preceding claims, The replacement reference value of the set comprising the measured values is defined such that the replacement reference value is a linear combination of the measured values of the set comprising the measured values.

4. The method according to claim 3, wherein the coefficients of the linear combinations are the same within each set of mappings, and / or The sum of the coefficients of the linear combination is substantially 1.

5. The method according to any one of the preceding claims, The replacement reference value of the set comprising the measurement values is defined as one measurement value among a plurality of measurement values comprising the set of measurement values.

6. The method according to any one of the preceding claims, The error estimate is based on determining the error in the data that is erroneous due to scattered radiation.

7. The method according to claim 6, Therein the error estimate is determined based on a ratio, optionally a maximum ratio, between the estimated scattered signal and the total signal strength of the measurement values of the corresponding set.

8. The method according to any of the preceding claims, wherein the error estimate is determined based on data errors in a low signal domain, such that a high error is assumed for low detector signals.

9. The method according to any one of the preceding claims, Wherein the function mapping the measured value to the correction value is a monotonic function.

10. The method according to any of the preceding claims, wherein the mapping is defined according to at least two different error ranges of the estimate, such that for each error range a different partial mapping function is applied.

11. The method according to any one of the preceding claims, wherein the error estimate comprises at least three error ranges (11, 12, 13) according to which the mapping is applied, the error ranges (11, 12, 13) comprising: a first error range (11) of low error, for which the corresponding correction value is set to the measured value; a second error range (12) of medium error, for which the measured values are mapped such that corresponding correction values are between the measured values and the replacement reference value of the set comprising the measured values; as well as A third error range (13) of high error, for which a corresponding correction value is set as the replacement reference value of the set comprising the measurement values.

12. A method for providing spectral computed tomography image data, the method comprising: - Acquisition of measurement data by performing spectral computed tomography, - performing a method according to any one of the preceding claims.

13. A computer program comprising instructions which, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 12.

14. A computer-readable storage medium, in particular a non-transitory storage medium, comprising instructions which, when executed by a computer, cause the computer to perform the method of any one of claims 1 to 12.

15. A computed tomography system comprising a processing unit (3) configured to perform the method of any one of claims 1 to 12.

Images