Motion compensation in spectral computed tomography
By generating and evaluating multiple motion vector fields in spectral CT imaging, the final motion vector field is constructed for motion compensation, and the problem of motion artifacts in spectral X-ray tomography is solved, and the accuracy and fidelity of imaging data are improved. More valuable clinical imaging results are provided.
Patent Information
- Application Number
- CN202380029819.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2022-05-25
- Filing Date
- 2023-03-13
- Publication Date
- 2024-11-08
AI Technical Summary
In spectral X-ray computed tomography, motion artifacts lead to blur or distortion of the characteristics of interest, and existing motion compensation technologies are difficult to effectively solve, especially in cardiac CT imaging, where motion compensation does not provide clinical value.
By receiving spectral CT imaging data across the imaging time period, multiple subsets of imaging data are generated, each subset corresponds to different spectral or material components, identify the structural features of interest and determine their motion vector field, and perform quality based on multiple motion vector fields Evaluate and weighted combinations to construct the final motion vector field for motion compensation.
Improves the accuracy and fidelity of motion compensation, enables clearer display of motor characteristics of different tissues and anatomical structures, reduces artifacts, and provides more clinically valuable imaging data.
Smart Images

Figure CN118922862A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of tomography, in particular to spectral X-ray computed tomography. Background Art
[0002] When acquiring image data of moving anatomical objects, motion may introduce artifacts. Specifically, features of interest may become blurred or distorted if the imaged object to which the features belong moves during the acquisition of the CT projection data.
[0003] As is known, in a procedure in which motion of a structural feature of interest is detected between frames of an image data set, motion compensation is applied to the acquired CT data and then the compensation is applied to the image data to compensate for the motion. The motion compensation procedure involves determining a motion vector field for the image data that represents the motion of the structural feature of interest.
[0004] For example, motion compensated CT reconstruction is a known technique for spiral scanning and step scanning. It is based on reconstruction of multiple images at different time points, object detection and elastic registration to determine the motion vector field, and subsequent motion compensated reconstruction. For example, this procedure has advantageous applications in the field of cardiac CT imaging, where the heart undergoes periodic motion. An image data set spanning at least a portion of the cardiac cycle can be acquired, and the above procedure can be applied to compensate for the motion of at least one vessel of the heart. Although one may wish to capture images of multiple different points during the cardiac cycle, motion when acquiring the projection data can result in undesirable distortion and does not provide clinical value.
[0005] A subcategory of CT imaging is spectral CT imaging. This involves acquiring CT projection data at multiple different energy levels and using this data to derive additional information about the imaged structure, such as based on the known different spectral absorption properties of different materials. For example, because different tissues have different energy dependencies on x-ray attenuation, this type of imaging can be used to obtain higher-level material differentiation in the image. In spectral CT imaging, multiple different imaging data sets can be generated, each corresponding to a different component of spectral imaging, such as Compton scattering and photoelectric images, or images corresponding to different energy levels or bands of the x-ray spectrum, or images corresponding to different materials. The multi-energy projection data can be decomposed into different possible basis component projection data sets (sometimes called basis line integrals) by a decomposition algorithm, where the different basis components correspond to different contributions to x-ray attenuation by the body.
[0006] A motion compensation procedure known in the field of spectral CT imaging is described in US2019 / 282184 A1. Summary of the invention
[0007] The present invention relates to a proposed method for facilitating improved accuracy or fidelity of motion compensation using multiple imaging data subsets provided as output of spectral CT image reconstruction.
[0008] Motion compensation can be applied to spectral CT imaging in an improved manner compared to conventional CT. Specifically, in spectral CT, multiple spectral images can be reconstructed from spectral projection data. Different structural features with different material and structural properties will be visible with different degrees of contrast in the various spectral images. Therefore, the inventors have realized that spectral CT allows the possibility to select the best spectral or material image data set for a given structural feature of interest, to perform motion tracking on this feature, or to synthesize data from multiple spectral or material image data sets in order to construct motion data with the best reliability for a given feature of interest.
[0009] The invention is defined by the claims.
[0010] According to an example of one aspect of the invention, a computer-implemented method for motion compensation in computed tomography (CT) imaging is provided.
[0011] The method includes receiving spectral CT imaging data spanning an imaging time period, the spectral CT imaging data including a plurality of imaging data subsets, each imaging data subset corresponding to a different spectral or material component or channel of spectral CT imaging, and each imaging data subset including a series of image frames spanning the imaging time period.
[0012] For each subset of imaging data, the method includes: processing the image data of the subset to identify at least one structural feature of interest, and determining at least one motion vector field for the at least one structural feature of interest, the motion vector field representing the motion of the at least one feature over the time period (relative to the image field of view). Optionally, a corresponding motion vector field may be generated for each image frame (or at least a subset thereof), the motion vector field representing the motion of the feature relative to a temporally preceding image frame. The complete set of these frame-specific motion vector fields may form an overall motion vector field having both temporal and spatial dimensions.
[0013] The method also includes determining a single final motion vector field based on the motion vector fields of the plurality of imaging data subsets.
[0014] The method also includes applying motion compensation to at least one of the spectral or material CT imaging data subsets based on the final motion vector field to generate at least one motion compensated imaging data subset. The motion compensated imaging data subset may include all image frames of the original imaging data subset or only a subset of image frames of the original imaging data subset, and each frame is a motion compensated frame.
[0015] The method also includes generating a data output representing at least one motion compensated subset of the imaging data.
[0016] The general approach of the proposed method is to generate one or more motion vector fields for each of the subsets of spectral imaging data (corresponding to each spectral or material component of the CT imaging data), and then use this multiple motion information to obtain one or more final motion vector fields with improved reliability. For different types of tissues and different anatomical structures or features, different spectral or material image sets will depict the relevant features with different contrasts, and therefore may display the features more or less clearly. Therefore, it can be expected that the motion vector field of the relevant structural features has a quality and reliability that depends on the clarity with which the features are represented in the underlying image data. Therefore, the proposed method utilizes multiple imaging data sets available in spectral CT imaging to select or configure motion vector fields in order to obtain motion vector fields with improved reliability.
[0017] Each data subset corresponds to a specific spectrum or material component or channel of the imaging data. To explain further, spectral CT scanning involves acquiring projection data readings at multiple energy levels, and then processing the data through a decomposition and reconstruction algorithm to obtain multiple reconstructed images, which can correspond to different contributions to the x-ray attenuation of the imaging volume. For example, the reconstructed images can correspond to different material components of the imaging volume, different virtual monochromatic images, photoelectric and Compton scattering images, etc. For example, a dual-energy scanner measures high-energy channels and low-energy channels through a dual-layer detector, and recalculates the acquired projection data into basic components (sometimes called basic line integrals) with the help of material decomposition, such as light and scattering projections, which can be converted into various images after reconstruction, such as: single E images, Zeff, iodine, Ca suppression, etc.
[0018] Each imaging data subset may include at least reconstructed image data per frame for each of the spectrum or material components or channels, such as photoelectric images and scatter images, or images of different X-ray energy levels. Optionally, each imaging data subset may additionally include projection data per frame corresponding to each of the spectrum or material components.
[0019] Generating corresponding final motion vector fields from a plurality of initial motion vector fields can be done in different ways.
[0020] One way is to simply select the "best motion vector field" from the multiple initial motion vector fields based on an analysis applied to the multiple motion vector fields, for example, the analysis performs a quality or reliability assessment on each motion vector field. Another way is to construct a composite motion vector field formed by a weighted combination of selected ones of the multiple motion vector fields. This option can also be based on a quality assessment performed on each motion vector field.
[0021] Thus, in at least a subset of the embodiments, the step of determining a final motion vector field comprises applying a quality assessment to each of the motion vector fields derived from the subset of imaging data.
[0022] There are different options for quality assessment. One common approach is to assess the consistency of each motion vector field with another one of the motion vector fields, or with a predetermined expected motion vector field for a given structural feature of interest. A second common approach is to compute one or more internal quality factors, e.g. related to noise, temporal smoothness, outlier detection, non-physiological motion vector detection, etc. A combination of these two approaches may be used.
[0023] Further expanding on this, in some embodiments, quality assessment of a given motion vector field may include comparing the motion vector field with a reference motion vector field, wherein the reference motion vector field may be another one of the motion vector fields derived from different imaging data subsets, or may be an average motion vector field calculated from multiple motion vector fields derived from the imaging data subsets, or may be another predetermined reference motion vector field. The output of the quality assessment of the given motion vector field may depend on the degree of consistency with the reference motion vector field.
[0024] In case the reference motion vector field is another predetermined motion vector field, the reference motion vector field may be a predetermined expected or typical motion vector field for the structural features of interest.The quality of a given motion vector field determined by the quality assessment may depend on the degree of consistency with the reference motion vector field.
[0025] As mentioned above, an alternative to comparing motion vector fields to an external reference is to compare motion vector fields to each other. To this end, an average or baseline field can be calculated, and the quality assessment can determine the consistency with this average field. In other words, a measure of the deviation or variance of a given motion vector field from the average is calculated. Higher quality can depend on smaller deviations from the average.
[0026] As mentioned above, the quality assessment applied to the motion vector field may additionally or alternatively comprise using one or more predetermined quality analysis algorithms to calculate one or more (internal) quality factors, such as local noise, spatial or temporal smoothness or outlier detection.
[0027] In some examples, quality assessment of a given motion vector field is performed by applying a pre-trained artificial neural network to the motion vector field, the artificial neural network being trained to receive the motion vector field as input and to generate one or more quality factors as output.
[0028] As described above, deriving the final motion vector field can be accomplished in at least two main ways: selecting one of the motion vector fields, or constructing a composite motion vector field based on a combination of multiple motion vector fields.
[0029] Thus, in some embodiments, determining the final motion vector field comprises selecting one of a plurality of motion vector fields determined from the subset of imaging data based on a quality assessment applied to the plurality of motion vector fields.
[0030] Optionally, the selecting comprises selecting one of the imaging data subsets based on a comparison with a reference motion vector field. Optionally, the selecting comprises selecting the imaging data subset whose corresponding motion vector field is most consistent with the reference motion vector field. Optionally, the selecting may be based on a quality factor of each motion vector field.
[0031] In some embodiments, the final motion vector field is a constructed motion vector field generated based on a combination of multiple motion vector fields derived from multiple imaging data subsets.
[0032] The final motion vector field may, for example, be generated as a weighted sum of motion vector fields derived from multiple subsets of CT imaging data.
[0033] Consistent with at least any one of the above embodiments, the received CT imaging data may be image data of at least a portion of a heart of the subject, and wherein the structural feature of interest is a blood vessel included in the heart.
[0034] In some embodiments, for each image frame of each imaging data subset, the received spectral CT imaging data includes CT projection data associated with the image frame, and a reconstructed image associated with the image frame, and wherein the steps of identifying at least one structural feature of interest and determining at least one motion vector field are applied to the reconstructed image data of at least one image frame, and the step of performing motion compensation includes applying a reconstruction algorithm to the projection data of the at least one image frame, wherein one or more parameters of the reconstruction algorithm are configured based on the derived motion vector field.
[0035] Another aspect of the present invention is a computer program product, which includes a computer program code, which is configured to, when executed by a processor, cause the processor to perform a method according to any embodiment described in the present disclosure or according to any claim of the present application.
[0036] Another aspect of the present invention is a processing device comprising an input / output and one or more processors. The one or more processors are adapted to perform the following steps:
[0037] receiving at an input / output spectral CT imaging data spanning an imaging time period, the spectral CT imaging data comprising a plurality of imaging data subsets, each imaging data subset corresponding to a different spectrum or material component of spectral CT imaging, and each imaging data subset comprising a series of image frames spanning the imaging time period;
[0038] For each imaging data subset:
[0039] processing the imaging data to identify at least one structural feature of interest, and
[0040] determining at least one motion vector field for at least one structural feature of interest, the motion vector field representing motion of the at least one feature during at least a portion of an imaging time period;
[0041] determining at least one final motion vector field based on the motion vector fields of the plurality of imaging data subsets;
[0042] Applying motion compensation to at least one of the spectral CT imaging data subsets based on the at least one final motion vector field to generate at least one motion compensated imaging data subset; and generating a data output at an input / output representing the at least one motion compensated imaging data subset.
[0043] Another aspect of the present invention is a system comprising: a CT imaging device for acquiring spectral CT imaging data; and a processing device as described above, or according to any embodiment described in the present disclosure, or according to any claim of the present application, and wherein the input / output of the processing device is communicatively coupled to the CT imaging device and is arranged to receive the spectral CT imaging data acquired thereby.
[0044] These and other aspects of the invention will be apparent from and elucidated with reference to the embodiments described hereinafter. BRIEF DESCRIPTION OF THE DRAWINGS
[0045] For a better understanding of the invention, and to show more clearly how the same may be carried into effect, reference will now be made, by way of example only, to the accompanying drawings, in which:
[0046] Figure 1 A block diagram showing an example method according to one or more embodiments of the present invention is shown;
[0047] Figure 2 An example CT imaging device is shown;
[0048] Figure 3 Example results of applying motion compensation to CT images obtained from conventional and spectral CT imaging are shown;
[0049] Figure 4Example images of anatomical regions corresponding to different spectral or material compositions are shown to demonstrate the varying contrast of features of interest depicted in each image; and
[0050] Figure 5 Example processing devices and systems according to one or more embodiments of the invention are shown. DETAILED DESCRIPTION
[0051] The present invention will be described with reference to the accompanying drawings.
[0052] It should be understood that although the detailed description and specific examples show exemplary embodiments of the device, system and method, these detailed descriptions and specific examples are for illustrative purposes only and are not intended to limit the scope of the invention. These and other features, aspects and advantages of the device, system and method of the present invention will be better understood from the following description, the attached claims and the accompanying drawings. It should be understood that the drawings are only schematic and are not drawn to scale. It should also be understood that the same reference numerals are used throughout the drawings to indicate the same or similar components.
[0053] The present invention provides a method for performing motion compensation in spectral CT imaging to compensate for the motion of at least one structural feature. The method includes receiving or generating reconstructed imaging data of multiple spectra or material components of a spectral CT imaging data set, and generating at least one motion vector field for each subset of the spectral or material component data, the motion vector field corresponding to the motion of the feature of interest detected during at least a portion of the imaging time period. A quality assessment can be performed on each of the multiple motion vector fields, and motion compensation is applied using a final motion vector field selected or constructed from the multiple motion vector fields, and the selection or construction is optionally based on the quality assessment. The quality assessment can be based in part on an assessment of the consistency of each motion vector field with other motion vector fields in the motion vector field or with a predefined reference motion vector field, and / or it can be based on internal quality characteristics, such as noise level or temporal smoothness.
[0054] Figure 1 A block diagram showing the steps of a method according to one or more embodiments of the present invention is shown. These steps will be explained in more detail later.
[0055] The method is computer-implemented. The method includes receiving 10 spectral CT imaging data spanning an imaging time period, the spectral CT imaging data including a plurality of imaging data subsets. Each imaging data subset corresponds to a different spectral or material component of the spectral CT imaging (e.g., corresponding to a different decomposed basis component of the acquired projection data and / or a different component of the source x-ray energy spectrum), and each imaging data subset includes a series of image frames spanning the imaging time period. In other words, each image frame corresponds to a different time point.
[0056] Each image frame may be a volumetric image frame containing reconstructed image data of a volumetric region.
[0057] Those skilled in the art will appreciate that in spectral CT, for a given set of acquired spectral projection data, the data may be reconstructed to form a plurality of different images corresponding to different spectral or material aspects or components of the image data, which images are generally referred to as "spectral images". For example, different reconstructed spectral images may correspond to different contributions of the imaged volume to X-ray attenuation. For example, a common implementation includes reconstructing dual-energy CT projection data into a pair of spectral images corresponding to contributions of the photoelectric effect and the Compton scattering effect to X-ray attenuation, which images may be referred to as photoelectric and scattering images, respectively. For example, following this example, the spectral components or channels of the two imaging data subsets would be photoelectric and scattering components, and one of the imaging data subsets would include a time series of photoelectric images, while the other would include a time series of scattering images. This is merely an example of illustrating what is meant by reference to spectral or material components or channels, and those skilled in the art will appreciate that there are a variety of options for spectral CT reconstruction.
[0058] Typically, more than two spectral images can be reconstructed from the source spectral CT projection data by applying different decomposition and reconstruction algorithms. In some cases, different spectral components or channels correspond to different single energy levels (single keV reconstruction). In some cases, different spectral images can correspond to different specific materials, such as iodine or calcium or water, and thus different imaging data subsets may correspond to different materials.
[0059] The received imaging data may be image data that has been reconstructed, or projection data may be received, and the method may further include the step of reconstructing multiple imaging data subsets of different spectral or material components. In some cases, the received data may include reconstructed image data and projection data for each frame of each spectral or material component. In some cases, the received data may be multi-energy projection data, and the method may include decomposing the projection data into a set of two or more basis component projection data sets (sometimes referred to as basis line integrals), wherein these data sets form two or more CT imaging data subsets. The method may also include reconstructing the decomposed projection data to form reconstructed data subsets, each data subset corresponding to a different basis component of spectral imaging.
[0060] For each subset of imaging data, the method includes processing 12 the imaging data to identify at least one structural feature of interest. This can be done using segmentation or another object detection method. For example, to detect blood vessels, the well-known Frangi vessel filter can be applied.
[0061] The method further comprises, for each imaging data subset, determining 14 at least one motion vector field of at least one structural feature of interest, the motion vector field representing the motion of the at least one feature (relative to the image field of view) over at least a portion of the time period (i.e., between consecutive image frames in each imaging data subset). This may be achieved, for example, by performing a registration between temporally consecutive frames in each imaging data subset, and wherein the motion of the object between frames may be derived from the net displacement detected by the registration algorithm. A variety of methods may be used to detect the displacement between consecutive frames of the feature of interest, and if the method is performed for each pair of consecutive frames, a motion vector field or set of motion vector fields may be established for the entire imaging period. For example, a (spatial) motion vector field may be generated for each image frame of each imaging data subset, the motion vector field representing the motion of the feature of interest relative to the immediately preceding image frame. The motion vector field at each time point (i.e., each frame) may span multiple points belonging to the feature of interest, such as a contour around the feature, or a volume, area, or line occupied by the feature. The set of motion vector fields for each frame may form a complete 4D motion vector field, spanning 3D space and spanning the time period covered by the frame set. Alternatively, a set of one or more motion vector fields may be generated for only a subset of image frames, covering only a portion of the imaging time period. At a minimum, the claimed method may correspond to motion compensation applied only between two temporally consecutive image frames, and wherein the motion vector field is calculated based on the motion of a feature of interest between the pair of frames, and motion compensation is applied to at least only one of the frames. However, more generally, motion compensation will be applied to more than just two frames, and pointing this out is merely to clarify the scope of the claimed invention.
[0062] The method further comprises determining 16 a single final motion vector field based on the motion vector fields of the plurality of imaging data subsets.
[0063] The method further comprises applying 18 motion compensation to at least one of the spectral CT imaging data subsets based on the final motion vector field to generate at least one motion compensated imaging data subset. The motion compensation may be a procedure applied to a single image frame at a time, wherein the image frame may be a 3D (volume) image frame. The procedure may be applied to one or more image frames of at least one of the imaging data subsets. In this case, the output of each execution of the motion compensation procedure may be a reconstructed image frame in which the motion represented in the motion vector field is compensated, thereby reducing or eliminating image artifacts associated with the feature of interest.
[0064] The method further comprises generating 20 a data output representing at least one subset of the motion compensated imaging data. This simply means generating a data signal carrying the motion compensated data to allow the data to be output to another processing unit, or a data storage, such as to another computer. In some examples, the method may comprise transmitting or transferring the data to a defined target address, wherein the target address may be user configurable.
[0065] In order to further illustrate the features of the present invention, Figure 2 An example CT imaging device for acquiring spectral CT imaging data is shown. It should be noted that the present invention does not require the steps of actually acquiring CT projection data or reconstructing projection data to form image data, although these steps may optionally be included as another part of the method. The method at least requires receiving reconstructed spectral CT image data, which may come from a CT scanner or may come from a data storage device. Figure 2 The description is provided by way of a useful context to aid in understanding the features encompassed by embodiments of the present invention.
[0066] Reference Figure 2 , a computed tomography (CT) scanner 100 includes a substantially stationary gantry portion 102 and a rotating gantry portion 104. The rotating gantry portion 104 is rotatably supported by the substantially stationary gantry portion 102 via bearings or the like.
[0067] A radiation source 106, such as an x-ray tube, is supported by the rotating gantry portion 104 and rotates therewith about a longitudinal axis or z-axis 110 about an examination region 108. A source collimator 114 or the like collimates radiation emitted by the radiation source 106 to produce a generally conical, fan-shaped, wedge-shaped, or other shaped radiation beam that traverses the examination region 108.
[0068] The radiation source voltage determiner 112 selectively determines the (average) emission voltage. In one example, the radiation source voltage determiner 112 switches or changes the emission voltage between consecutive scans of the same subject / object (this is called kVp switching). Alternatively, the radiation source voltage determiner 112 switches the emission voltage during the same scan, for example, from view to view, within a view, and / or otherwise switches the emission voltage. Therefore, a radiation beam with different energy spectra can be used to scan the subject / object. Since the absorption of photons by the material depends on the photon energy, the data from the two scans can be used to determine the information indicating the elemental composition (such as atomic number) of the tissue / material in the scanned subject / object. In this way, data related to two different spectral channels can be obtained.
[0069] By non-limiting illustration, the radiation source voltage determiner 112 can switch the emission voltage between about 80kV and about 140kV between scans, between views, within a view and / or otherwise. Under higher emission voltages, filters can be used to filter low-energy photons, which can improve the spectral sensitivity of the system. The result of switching the emission voltage is that the radiation source 106 produces a first radiation beam with a first energy spectrum and a second radiation beam with different second energy spectra. Of course, other emission voltages can also be considered, and the radiation source voltage determiner 112 can switch between more than two different emission voltages.
[0070] The detector array 116 forms an angular arc opposite the examination region 108 relative to the radiation source 106 and detects radiation passing through the examination region 108. The illustrated detector array 116 may include: a photosensor array 118 having photosensors such as photodiodes; and a scintillator array 120 optically coupled to the photosensor array 118 at a photosensitive side of the photosensor array 118. The detector array 116 is arranged in the scanner 100 so that the scintillator 120 receives incident radiation. Although only a single row of energy-resolving detector arrays 116 is shown, a two-dimensional energy-resolving detector array is also contemplated herein, where rows extend in the z-axis direction and columns extend in the lateral direction.
[0071] As an alternative to or in addition to using multiple emission voltages, an energy-resolving detector or a photon counting detector may be used. The energy-resolving detector may include a scintillator array 120 that includes two or more regions 122, 124 with different spectral sensitivities. The spectral sensitivities of the photosensors in the photosensor array 118 are matched to the emission spectra of the scintillation regions 122, 124. For example, some photosensors in the photosensor array 118 detect light emitted by the scintillation region 122, while other photosensors in the photosensor array 118 detect light emitted by the scintillation region 124. Typically, lower energy photons are absorbed in the scintillation region 122, while photons that pass through the scintillation region 122 are absorbed in the scintillation region 124. Alternatively, the energy-resolving detector array may be of a direct conversion type.
[0072] The energy-resolving detector array 116 outputs a signal or projection data indicative of the detected radiation. The array 116 may be a dual-layer detector array, an energy-sensitive photon counting detector array, or another energy-resolving detector array. Because the emission voltage may vary during a scan and the energy-resolving detector array 116 is sensitive to photon energy, the energy-resolving detector array 116 generates energy-resolved projection data d n, where n represents the energy resolved data for the nth energy range. For example, where the emission voltage is switched between two different emission voltages during scanning and the detector array 116 includes two sets of photosensors with two different spectral sensitivities, the resulting projection data includes four (4) independent energy resolved measurements representing different combinations of the two emission voltages and the two photosensor spectral sensitivities.
[0073] The reconstructor 126 reconstructs the projection data from the detector array and generates volumetric image data indicative of the examination region 108. As described above, the reconstructor 126 receives energy-resolved detector signals d indicating energies detected within n energy ranges. n The reconstructor 126 employs one or more spectral decomposition algorithms 128 and / or one or more spectral reconstruction algorithms 130, such as a maximum likelihood (ML) reconstruction algorithm to reconstruct spectral data that takes into account the detected spectral information. The resulting reconstructed image data includes a plurality of image data subsets, each image data subset corresponding to a different spectral or material composition.
[0074] For example, in one embodiment, the decomposition algorithm 128 may include an algorithm that models the data as a combination of a photoelectric effect having an attenuation spectrum P(E) and a Compton effect having an attenuation spectrum C(E). n In the equation, the density-length product of these components, namely the density-length product of the photoelectric effect component p and the Compton effect component c, can be modeled as a nonlinear system according to the following relationship:
[0075] Formula 1: d n =∫dET(E)D n (E)exp(-(pP(E)+cC(E)))
[0076] Where T(E) is the emission spectrum of the radiation source 106, D n (E) is the spectral sensitivity of the nth measurement.
[0077] When there are at least two detection signals d1, d2 in at least two energy ranges, at least two sets of equations with two unknowns will be formed, which can be solved by known numerical methods. The results p and c are called spectral line integrals. They can be used alone or in combination to reconstruct images of the desired components using conventional reconstruction methods. These images are called reconstructed spectral images. For example, in the above example, a Compton scattering image and a photoelectric image can be generated for the imaging volume.
[0078] Reconstruction can be performed on a series of image frames, which can be captured in, for example, a spiral scan or a step scan. Another term for a step scan is sequential scan; it is an imaging protocol in which projection data is acquired one axial position at a time. For each axial position, the scanner's rotating gantry rotates once around the patient to acquire the projection data, then the patient is moved axially to position the gantry at the new axial position relative to the patient, and so on. This is in contrast to a spiral scanning protocol, in which the patient is continuously moved in the axial direction while the scanner is continuously rotated.
[0079] As described above, the method generally includes receiving a plurality of subsets of reconstructed spectral image data, each subset corresponding to a separate spectrum or material component. Each data subset includes image frames at a series of time points.
[0080] The method also includes, for each subset of imaging data, detecting a structural feature of interest. For example, one advantageous application of the method is for cardiac imaging. The structural feature of interest may be at least one blood vessel or valve of the heart. As another example, the structural feature of interest may be a pacemaker lead or another implanted structure. In some examples, segmentation may be used to achieve detection of the feature of interest. Those skilled in the art will be aware of various other object detection methods. For example, to detect blood vessels, the well-known Frangi vessel filter may be applied.
[0081] At least one detected feature of interest in each imaging data subset may be tracked over at least a subset of the plurality of frames to identify its location in each image frame. At a minimum, its location should be detected in at least two consecutive frames in each imaging data subset.
[0082] At least one motion vector field may be constructed for each subset of imaging data. In general, a corresponding motion vector field may be generated for each corresponding image frame of each subset of imaging data or at least a subset of image frames. The motion vector field for a given frame may represent the motion of a feature of interest between a previous frame and the given frame.
[0083] There are several specific ways to generate a motion vector field. One method is based on elastic registration of images. Another method is to use model-based segmentation, which is applied to the two image frames of a given pair of consecutive image frames, and wherein the location of the feature of interest in each frame is then detected based on the segmentation, from which a motion vector field from one frame to the next can be derived. Another method is to use landmark tracking. Another method is to train an AI algorithm (such as an artificial neural network) to generate a motion vector field for an input image frame pair, which represents the movement of the feature of interest between the two image frames.
[0084] The method of generating a motion vector field is described in detail by way of example in the following paper: van Stevendaal, U., von Berg, J., Lorenz, C. and Grass, M. (2008), A motion-compensated scheme for helical cone-beam reconstruction in cardiac CTangiography. Med. Phys., 35: 3239-3251. Part II.B of this paper is particularly relevant.
[0085] The motion vector fields of the plurality of imaging data subsets may be referred to as initial motion vector fields or candidate motion vector fields. From these motion vector fields, a motion vector field is obtained, which is used to perform motion compensation for a feature of interest in at least one received imaging data subset.
[0086] Once the final motion vector field is determined, motion compensation can be performed as follows.
[0087] Motion compensation can be applied to one of the subsets of imaging data, or only one of the subsets of imaging data. Motion compensation can be a process applied individually on a frame-by-frame basis. It can be applied to only one frame, or to multiple frames.
[0088] As an example embodiment, motion compensation can be accomplished by reprocessing the underlying projection data associated with a given image frame, but where one or more parameters of the reconstruction algorithm have been adjusted based on the detected motion represented by the motion vector field calculated between that frame and an adjacent frame so that the reconstruction takes into account the motion. In other words, the reconstruction algorithm is configured to assume the motion of the imaged feature of interest, rather than (as is the standard case) assuming that the feature of interest is static (or, in the case of helical imaging, that the feature of interest moves in some fixed motion determined by the fixed axial velocity of the patient support through the gantry). The method assumes that the motion of the feature detected between the first frame and the second frame represents the motion of the feature during the acquisition of the projection data of the second frame, and then compensates for the motion by repeating the reconstruction of the second frame by assuming the motion vector field. This can be done for any number of frames for which motion compensation is desired. Other methods are also possible. For example, the projection data itself can be processed to compensate for the motion. Alternatively, the motion compensation procedure can be applied as a whole to the reconstructed image data using a suitable algorithm. For example, a trained AI algorithm may be a suitable method to implement this example.
[0089] Methods for performing motion compensation based on motion vector fields are described in detail by way of example in the following paper: van Stevendaal, U., von Berg, J., Lorenz, C. and Grass, M. (2008), A motion-compensated scheme for helical cone-beam reconstruction in cardiac CTangiography. Med. Phys., 35: 3239-3251. Part II.C of this paper is particularly relevant.
[0090] With regard to the step of determining the final motion vector field from a plurality of initial motion vector fields (determined from a subset of the received imaging data), there are different options which will now be outlined in detail. It should be noted that this step may be applied to each motion vector field calculated. In the case where one field is calculated for each frame, then determining the final motion vector field may be a step that is done for each motion vector field (i.e. for each frame).
[0091] In general, the step of determining the final motion vector field may be based on applying a predetermined quality assessment to each initial motion vector field. This may comprise one or more programs, encoded in the form of algorithms, which may be stored in a memory of a processing device executing the method.
[0092] There are different options for quality assessment. One common approach is to assess the consistency of each motion vector field with another one of the motion vector fields, or with a predetermined expected motion vector field for a given structural feature of interest. A second common approach is to calculate one or more internal quality factors, e.g. related to noise, temporal smoothness, etc. A combination of these two approaches may be used.
[0093] Turning to the first way of assessing consistency, the quality assessment of a given motion vector field may include comparing the motion vector field to a reference motion vector field. The output of this comparison may be a numerical score indicating the degree of consistency with the reference motion vector field.
[0094] In a first possible implementation thereof, the reference motion vector field may be a pre-stored reference motion vector field, which may, for example, represent an expected or typical motion vector field for a given feature. Comparison with this reference may identify motion vector fields that are inconsistent with the tracked feature (which may indicate a lack of reliability). This predetermined motion vector field may, for example, be stored in advance in a local memory and retrieved as part of the method for comparison purposes. For example, it may be determined in advance based on averaging or otherwise synthesizing motion vector fields of the feature obtained in a plurality of previous real-world scans.
[0095] In a second possible implementation, the reference motion vector field may be an average motion vector field derived from a plurality of initial motion vector fields. By comparing each motion vector field with the average field, a deviation or variance from the average may be derived, which may indicate whether and to what extent a given motion vector field is an abnormal result. In some examples, a higher consistency with the average field may indicate a higher quality / reliability. The average field may be a mean or median field, or any other type of average field. In some examples, it may be a weighted average field.
[0096] In another possible implementation, the reference motion vector field may simply be one selected from the other motion vector fields. This may be selected based on closeness to the expected motion field, or based on satisfying some other predefined threshold or criteria.
[0097] In either of these cases, the output of the quality assessment for a given motion vector field may depend on the degree of consistency with a reference motion vector field.
[0098] As mentioned above, the quality assessment applied to the motion vector field may additionally or alternatively comprise using one or more predetermined quality analysis algorithms to calculate one or more (internal) quality factors, such as local noise, spatial or temporal smoothness or outlier detection.
[0099] With regard to local noise, this can be understood as follows. For a given motion vector field, the pattern of motion vectors as a function of position in the image frame can be analyzed to detect abnormal vectors or motion spikes. For example, the degree of motion vector variation between two spatially adjacent points, or between two points with a certain proximity to each other, can be assumed, and the motion variation between these two points beyond these boundaries can be assumed to be noise. A noise metric for a given motion vector field can be derived, for example, based on the number and degree of these abnormal motion vectors.
[0100] With respect to temporal smoothness, an analysis similar to that for noise detection can be performed between motion vectors corresponding to the space of a series of temporally continuous motion vector fields (for temporally continuous frames). Here, the motion vectors for a given spatial point or region can be assumed to have a certain degree of temporal smoothness, and variations in the motion vectors between two frames outside of certain predefined boundaries or parameters can be assumed to be non-physiological and indicative of errors. The number and degree of such temporally anomalous vectors can be analyzed for a given motion vector field and used to determine a measure of temporal smoothness for the motion vector field.
[0101] With regard to outlier detection, this may simply consist of defining upper and lower limits for what is considered physiological motion, and motion vectors outside the range defined by these thresholds are classified as non-physiological motion.
[0102] Another type of internal quality assessment can be to assess the degree of spatial correspondence between motion vectors and underlying physiological or anatomical structures. In other words, it can be assumed that the pattern of motion vectors should be spatially mapped onto the underlying anatomical structures, so motion vectors outside the boundaries of detected anatomical objects or features can be classified as abnormal.
[0103] In some examples, quality assessment of a given motion vector field is performed by applying a pre-trained artificial neural network to the motion vector field, the artificial neural network being trained to receive the motion vector field as input and to generate one or more quality factors as output. The neural network can be trained in advance in a supervised training procedure using training data including sample images annotated with one or more quality indicator scores, and wherein the neural network is trained to output the quality indicator scores.
[0104] In some examples, quality assessment may include a combination of the above approaches.
[0105] Once multiple initial motion vector fields are derived, a final motion vector field may be determined based thereon. Optionally, this may be based on a quality assessment applied to the initial motion vector fields (as described above), but this is not required. As described above, deriving the final motion vector field may be done in at least two main ways: selecting one of the motion vector fields; or constructing a composite motion vector field based on a combination of multiple motion vector fields.
[0106] According to a first approach, the final motion vector field is obtained by simply selecting one of the multiple initial motion vector fields. This can be based on their respective quality scores derived from the quality assessment. For example, the motion vector field with the highest quality score is selected.
[0107] According to a second approach, the final motion vector field is constructed from a composite of at least some of the motion vector fields selected from the initial motion vector fields. It can be formed as a weighted sum of the initial motion vector fields. The weights can be determined in different ways. In the simplest case, the weights are all the same, so that the composite field is simply the average motion vector field. In other cases, the weights may be different. As an example, the weight of each motion vector field may be set based on a quality score output from an optional quality assessment applied to the individual motion vector fields. Instead of a weighted sum, a spatial composite of multiple motion vector fields may be constructed so that spatial portions of multiple of the initial motion vector fields are stitched together.
[0108] Once the final motion vector field is derived, motion compensation may be applied in the manner discussed previously in this disclosure. As previously described, in the case where one motion vector field is derived for each frame, the process of deriving the final motion vector field may be applied multiple times to obtain multiple final motion vector fields, such as one for each frame.
[0109] Embodiments of the methods described in the present disclosure may be advantageously applied to imaging any anatomical region of the body and to motion compensate for any structural feature of interest.
[0110] One area of application of particular interest is in cardiac imaging and for compensating for movements of blood vessels or heart valves.
[0111] In some embodiments, the motion compensation method may be performed multiple times (i.e., multiple times) for each of a plurality of different structural features (e.g., each blood vessel or each valve). Alternatively, an object detection method that detects a plurality of features and generates a motion vector field covering a plurality of different features at once may be used, in which a plurality of structural features may be tracked and motion compensated in a single iteration of the method.
[0112] Figure 3 Examples of standard and motion compensated reconstruction of a conventional CT image series and a spectral CT image series using the same motion vector field estimated based on conventional images are shown. Figure 3 (Top left) Frames from a conventional CT image series without motion compensation applied are shown. Figure 3 (Top right) Frames from the same conventional CT image series with motion compensation applied are shown. Figure 3 (Bottom left) shows a frame corresponding to one spectral component (corresponding to the 60 keV energy channel) of a spectral CT image series without motion compensation. Figure 4 (Bottom right) shows frames corresponding to the same spectral component (corresponding to the 60 keV energy channel) of the CT image series with motion compensation applied, but using the same motion vector field derived from the conventional images. This demonstrates the principle that using image data of a selected spectral or material component from a spectral CT imaging dataset (even without changing the motion compensation algorithm) can produce superior results if the selected spectral or material basis component represents the feature of interest with maximum contrast. In this case, the feature of interest is the coronary artery, which extends from the top of the frame to the bottom of the frame, as roughly indicated by the white arrow in the upper left image.
[0113] Figure 4 A set of different images of the same anatomical region are shown, which have been reconstructed to represent different spectra or material compositions of the spectral CT image dataset. Figure 4 (Top left) A conventional (non-spectral) image is shown, meaning it was reconstructed from conventional CT projection data corresponding to a single x-ray energy band. Figure 4 The remaining images in are spectral CT images. Figure 4 (Top right) shows the spectral CT image corresponding to the 60keV spectral component, Figure 4(lower left) Spectral CT image corresponding to the 70keV spectral component, and Figure 4 (Bottom right) Spectral CT image showing the 120 keV spectral component.
[0114] Figure 4 The images of exemplify the basic premise discussed earlier in this disclosure, namely that for a given feature of interest, it can be expected that the feature will be depicted with greater or lesser contrast in images corresponding to different spectra or material compositions derived from the spectral CT image dataset. Figure 4 The blood vessels pointed by the white arrows can be clearly seen in the figure. It can be seen that the blood vessels have higher contrast at lower energies, with the highest contrast appearing in the 60keV image ( Figure 4 (top right), while in the 120keV image ( Figure 4 (lower right)) Poor contrast.
[0115] It can be appreciated that a motion vector field for a depicted vessel derived from image data of different spectral channels can be expected to have a fidelity or reliability that also varies depending on the relative contrast of the features depicted in such image data. Thus, the method of the present invention advantageously provides a way to obtain an optimal motion vector field by analyzing motion vector fields from images corresponding to a plurality of different spectral or material basis components of the image data.
[0116] Although examples related to imaging of the heart or parts thereof are given above, the methods described in the present disclosure can be applied to imaging a variety of different anatomies or regions with advantageous effects. By way of non-exhaustive example, the methods can be applied to motion compensation of spectral images depicting lung motion (breathing), head motion, intestinal motion, or involuntary motion of a patient.
[0117] Furthermore, it is worth noting that the above methods are applicable to any type of spectral tomographic imaging, not just spectral CT imaging. Moreover, these methods can be applied even more broadly to any type of spectral medical imaging, such as X-ray imaging, such as spectral C-arm imaging or airborne imaging. Within the field of CT imaging, it can be applied to any type of CT imaging, including cone beam CT.
[0118] Another aspect of the present invention provides a computer program product, which includes computer program code, and the computer program code is configured to, when executed by a processor, cause the processor to perform a method according to any embodiment or example described in the present disclosure or according to any claim of the present application.
[0119] Another aspect of the present invention provides a processing device, which includes one or more processors, wherein the processors are suitable for executing the method according to any embodiment described in this disclosure or according to any claim of this application.
[0120] Figure 5 An example processing device 30 is depicted, which includes: an input / output (I / O) 32; and one or more processors 34. The input / output 32 is operably coupled to the CT imaging device 100 to receive spectral CT imaging data 42 spanning an imaging time period at the input / output, the spectral CT imaging data including a plurality of imaging data subsets, each imaging data subset corresponding to a different spectrum or material component of the spectral CT imaging data, and each imaging data subset including a series of image frames spanning the imaging time period. However, the image data 42 may be received from a different source, such as from a data storage.
[0121] One or more processors are programmed to perform the method according to any embodiment described in the present disclosure or according to any claim of the present application to achieve motion compensation of image data. At the end of the program, a data output 44 representing the motion compensated data is provided at the input / output 32.
[0122] Another aspect of the present invention may provide a system 50, comprising: a CT imaging device 100 for acquiring spectral CT imaging data; and a processing device 30 according to any embodiment or any claim of the present application, wherein the input / output 32 is communicatively coupled to the CT imaging device and is arranged to receive the spectral CT imaging data 42 acquired thereby.
[0123] The embodiments of the present invention described above employ processing means. The processing means may generally include a single processor or multiple processors. It may be located in a single containing device, structure or unit, or it may be distributed between multiple different devices, structures or units. Therefore, reference to a processing means being adapted or configured to perform a particular step or task may correspond to the step or task being performed by any one or more of a plurality of processing components, either individually or in combination. A skilled person will understand how such a distributed processing means may be implemented. The processing means includes a communication module or input / output for receiving data and outputting data to other components.
[0124] The one or more processors of the processing device can be implemented in software and / or hardware in a variety of ways to perform the various functions required. The processor usually adopts one or more microprocessors, which can be programmed using software (e.g., microcode) to perform the required functions. The processor can be implemented as a combination of dedicated hardware to perform certain functions and one or more programmed microprocessors and associated circuits to perform other functions.
[0125] Examples of circuits that may be employed in various embodiments of the present disclosure include, but are not limited to, conventional microprocessors, application specific integrated circuits (ASICs), and field programmable gate arrays (FPGAs).
[0126] In various embodiments, the processor may be associated with one or more storage media, such as volatile and non-volatile computer memory, such as RAM, PROM, EPROM, and EEPROM. The storage media may be encoded with one or more programs that, when executed on one or more processors and / or controllers, perform the desired functions. The various storage media may be fixed within the processor or controller, or may be removable so that one or more programs stored thereon may be loaded into the processor.
[0127] Those skilled in the art can understand and implement various modifications to the disclosed embodiments when practicing the claimed invention by studying the drawings, the disclosure and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality.
[0128] A single processor or other unit may fulfill the functions of several items recited in the claims.
[0129] The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage.
[0130] The computer program may be stored / distributed on suitable media (such as optical storage media or solid-state media provided together with or as part of other hardware), but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems.
[0131] If the term "suitable for" is used in the claims or the description, it should be noted that the term "suitable for" is intended to be equivalent to the term "configured to".
[0132] Any reference signs in the claims should not be construed as limiting the scope.
Claims
1. A computer-implemented method for motion compensation in computed tomography (CT) imaging, the method comprising: receiving (10) spectral CT imaging data spanning an imaging time period, the spectral CT imaging data comprising a plurality of imaging data subsets, each imaging data subset corresponding to a different spectrum or material component of spectral CT imaging, and each imaging data subset comprising data of a series of image frames spanning the imaging time period; processing (12) each imaging data subset to identify at least one structural feature of interest in each respective imaging data subset, and determining (14) for each imaging data subset at least one motion vector field for the at least one structural feature of interest in the respective imaging data subset, the motion vector field representing motion of the at least one feature over at least a portion of the imaging time period; determining (16) a final motion vector field based on the motion vector fields of the plurality of imaging data subsets; applying (18) motion compensation to at least one of the spectral CT imaging data subsets based on the final motion vector field to generate at least one motion compensated imaging data subset; as well as A data output representing the at least one motion compensated imaging data subset is generated (20).
2. The method according to claim 1, wherein: Determining the final motion vector field comprises applying a quality assessment to each of the motion vector fields derived from the imaging data subset.
3. The method according to claim 2, wherein: Quality assessment of a given motion vector field comprises comparing the motion vector field with a reference motion vector field, wherein the reference motion vector field is another one of the motion vector fields derived from different subsets of imaging data, an average motion vector field calculated from a plurality of derived motion vector fields, or another predetermined reference motion vector field.
4. The method according to claim 3, wherein: The reference motion vector field is another predetermined motion vector field, and wherein the quality of a given motion vector field determined by the quality assessment depends on the degree of consistency with the reference motion vector field.
5. The method according to any one of claims 2 to 4, wherein: The quality assessment applied to the motion vector field comprises calculating one or more quality factors, such as local noise, spatial or temporal smoothness, or outlier detection, using one or more predetermined quality analysis algorithms.
6. The method according to claim 5, wherein: Quality assessment of a given motion vector field is performed by applying a pre-trained artificial neural network to the motion vector field, the artificial neural network being trained to receive the motion vector field as input and to generate one or more quality factors as output.
7. The method according to any one of claims 2 to 6, wherein: Determining the final motion vector field includes selecting one of a plurality of motion vector fields determined from a subset of the imaging data based on a quality assessment applied to the plurality of motion vector fields.
8. The method according to any one of claims 1 to 6, wherein: The final motion vector field is a constructed motion vector field generated based on a combination of multiple motion vector fields derived from multiple imaging data subsets.
9. The method according to claim 8, wherein: The final motion vector field is generated as a weighted sum of the motion vector fields of the plurality of CT imaging data subsets.
10. The method according to claim 9 in combination with claim 2, wherein: The weight of each corresponding motion vector field in the weighted sum is determined according to the result of the quality evaluation of the corresponding motion vector field.
11. The method according to any one of claims 1 to 10, wherein: The received CT imaging data is imaging data of at least a portion of a heart of a subject, and wherein the structural features of interest are blood vessels contained within the heart.
12. The method according to any one of claims 1 to 11, wherein: The received spectral CT imaging data comprises, for each image frame of each imaging data subset, CT projection data associated with the image frame, and a reconstructed image associated with the image frame, and wherein the steps of identifying at least one structural feature of interest and determining at least one motion vector field are applied to the reconstructed image data of at least one image frame, and the step of performing motion compensation comprises applying a reconstruction algorithm to the projection data of at least one image frame, wherein one or more parameters of the reconstruction algorithm are configured based on the derived motion vector field.
13. A computer program product, comprising a computer program code, which, when executed by a processor, is configured to cause the processor to perform the method according to any one of claims 1 to 12.
14. A processing device (30), comprising: Input / Output (32); as well as One or more processors (34), the one or more processors (34) being adapted to: receiving at the input / output spectral CT imaging data (42) spanning an imaging time period, the spectral CT imaging data comprising a plurality of imaging data subsets, each imaging data subset corresponding to a different spectrum or material component of spectral CT imaging, and each imaging data subset comprising a series of image frames spanning the imaging time period; processing each imaging data subset to identify at least one structural feature of interest in each respective imaging data subset, and determining, for each imaging data subset, at least one motion vector field for at least one structural feature of interest in the corresponding imaging data subset, the motion vector field representing motion of the at least one feature over the time period; determining a final motion vector field based on the motion vector fields of the plurality of imaging data subsets; applying motion compensation to at least one of the spectral CT imaging data subsets based on the final motion vector field to generate at least one motion compensated imaging data subset; as well as A data output (44) representing the at least one motion compensated imaging data subset is generated at the input / output (32).
15. A system (50), comprising: A CT imaging device (100) for acquiring spectral CT imaging data; as well as The processing device (30) of claim 14, wherein the input / output is communicatively coupled to the CT imaging device and is arranged to receive spectral CT imaging data acquired thereby.
