CT-projection-based breathing binning

Abstract

System (SYS) and associated method for processing projection data for X-ray tomographic imaging. The system comprises an input interface (IN) for receiving projection images acquired by an X-ray tomographic imaging device (IA) along various projection directions of an anatomical feature (AF) of a moving object as it passes through several motion states. An in-image detection component (In-Image DC) detects, with respect to at least one reference path, a motion state of the anatomical feature (AF) based on image information in the projection images. A gating component (GC) assigns a subset of the projection images to a motion state bin based on the detected motion state. An output interface (OUT) provides the assigned subset of projection images for reconstruction.

Description

AREA OF INVENTION

[0001] The invention relates to a system for processing projection data for tomographic imaging, an imaging arrangement including such a system, an associated method, a computer program element and a computer-readable medium. BACKGROUND OF THE INVENTION

[0002] 4D CT imaging is routinely used in radiation oncology and potentially in other applications, such as radiology. With respiratory-phase-based gating, a tumor or organ (lungs, liver, kidneys, etc.) can be tracked throughout the respiratory cycle. Gating refers to the process of extracting a subset of projection data from a stream, representing the imaged organ or anatomy in a specific state of motion. A more accurate, motion-compensated reconstruction can then be calculated from this subset of projection data.

[0003] Some respiratory gating techniques may require additional hardware (“device-based”), such as a pressure sensor belt attached to the patient to measure current respiratory status. Such a device-based setup is disadvantageous because it complicates the clinical workflow. For example, the device must be acquired and cleaned, maintenance costs are incurred, and lengthy preparation is required before imaging, etc. BRIEF SUMMARY OF THE INVENTION

[0004] Therefore, there may be a need for more efficient imaging of moving regions of interest, particularly in transmission-based tomographic imaging, such as X-ray tomographic imaging.

[0005] One object of the present invention is fulfilled by the subject matter of the independent claims, with further embodiments being contained in the dependent claims. It should be noted that the aspect of the invention described below relates equally to the imaging arrangement, the associated method, the computer program element, and the computer-readable medium.

[0006] According to a first aspect of the invention, a system for processing projection data (projection images) for X-ray tomographic imaging is provided, comprising: an input interface for receiving projection images (comprising individual images) that are captured by means of an X-ray tomographic imaging device along different projection directions of an anatomical feature of a moving object as it passes through several states of motion; an in-image recognition component configured to detect a motion state of the anatomical feature with respect to at least one reference path based on image information in the projection images; a gating component configured to assign a subset of the projection images to a motion state bin based on the detected motion state, and an output interface for providing the assigned subset of projection images for reconstruction.

[0007] In general, the system is configured to facilitate gating in tomographic imaging, particularly when imaging a moving object of interest. Specifically, the image information can be fully contained within one or more adjacent projection images acquired over time. For example, the image information might consist of image features, such as an edge structure, within the projection images (as in a single image).

[0008] The gating component is configured to perform a gating operation. The gating component can be configured to interpolate motion states for other frames among the projection images that were not used by the in-image detection component. Thus, if needed, any (or at least most) of the frames can be assigned as a motion state bin.

[0009] In embodiments, the path is selected based on an analysis of an overview image of the anatomical feature or on an analysis of the projection images.

[0010] The overview image is preferably a 2D X-ray image that covers the anatomical feature and other structures of interest. Preferably, a lower X-ray dose is used when acquiring the overview image than when acquiring the projection images in a diagnostic scan. The overview image is preferably used with an axial scan method using "step and shoot" ("S&S"). The use of a helical scan method is not usually necessary, but can also be performed.

[0011] In some embodiments, the detected state of movement is that of an anatomical feature.

[0012] For example, in a medical context, such as the one primarily intended here, the “object” may refer to the patient or a part of it, such as the chest wall, while the anatomical feature may be, for example, its anterior border.

[0013] In some embodiments, the detection of the state of motion is carried out with respect to a multitude of such reference paths.

[0014] In embodiments, one plane of the at least one such reference path runs essentially perpendicular to a projection direction in which at least some of the projection images are captured.

[0015] In embodiments, the anatomical feature is completely and at all times within the field of view of the tomographic imaging device during such a movement.

[0016] In embodiments, the system includes a tomographic reconstructor to reconstruct, for at least one motion state and based on projection images associated with that bin and / or one or more bin(s) of similar states, a reconstructed cross-sectional image volume that is representative of the object in that motion state. The reconstructor can use any of a number of tomographic reconstruction algorithms, such as filtered backprojection or iterative-type algorithms, etc.

[0017] In some embodiments, the projected images are images in which a second imaging geometry is obtained from a first imaging geometry used by the imaging device by means of an imaging geometry binner (IG: Imaging Geometry, "IG" binner). However, this IG-based binning differs from motion-state-based binning, which is the main focus here.

[0018] In embodiments, the first imaging geometry is of the cone-beam and / or helical type, and the second geometry is a parallel or quasi-parallel, such as a wedge geometry.

[0019] In some embodiments, the system includes a display device for showing the reconstructed image and / or the state of motion.

[0020] In some embodiments, the tomographic imaging device is configured for axial or helical imaging.

[0021] In embodiments, the anatomical feature includes at least one section of one or more of a patient's diaphragm, chest wall, and abdominal wall.

[0022] In some embodiments, the said movement is periodic.

[0023] In embodiments, the anatomical feature is the anatomy of a patient, wherein such movement is caused by: respiration, cardiac activity or any other movement caused by a biological activity.

[0024] In embodiments, the system may include a path definer configured to calculate such a reference path based on an overview image, or based on the projection images or other data.

[0025] The path definer can be based on a trained machine learning model. Alternatively, the reference path can be provided by a user via a user interface. Thus, the provision of the reference path can be automated, as with such a path definer, or manual by the user. The implementation of the path definer's machine learning model is optional. Instead, or additionally, the path definer can use analytical / geometric methods.

[0026] In some embodiments, the movement occurs along the reference path or perpendicular to it.

[0027] In embodiments, the reference path is a line, and this is preferred and provided for in most embodiments herein. The path / line is located in the image domain and not in the projection domain. For in-image detection in the projection domain, a projection of the path / line as recorded in the projection images is used. In particular, the projection of the reference line in the projection images defines a relatively narrowly defined spatial area in which the pixel value distribution can be analyzed for gating purposes, thus keeping the computational effort low, as it is limited to this narrowly defined area. The reference line / reference path can be defined in a specific, dedicated image processing operation, either based on overview images (preferably 2D) or based on the projection images themselves.

[0028] In some embodiments, the movement occurs along or across the reference path. For example, in some embodiments, the reference path is chosen such that the anatomical feature can cross the reference path once or multiple times (e.g., repeatedly), causing a changing pattern of attenuation values ​​that is recorded in the projection images. This attenuation value change pattern (e.g., pulsating or otherwise) can be detected by the in-image detection component to identify the state of motion and can be used by the gating component to binn the state of motion. For example, the diaphragm is such a feature, with the reference line running in the AP direction and through the diaphragm.

[0029] In another aspect, an imaging arrangement is provided, comprising the system according to one of the embodiments described above and the imaging device.

[0030] In embodiments of the arrangement, the system is at least partially integrated into or at least connected to an operator console or workstation as part of the imaging device.

[0031] In another aspect, a method for processing projection data for X-ray tomographic imaging is provided, including: Receiving projection images taken by means of an X-ray tomographic imaging device along different projection directions of an anatomical feature of a moving object as it passes through several states of motion; In-image recognition, with respect to at least one reference path, of a motion state of the anatomical feature based on image information in the projection images; Assigning a subset of the projection images to a motion state bin based on the detected motion state and Providing the assigned subset of projection images for reconstruction.

[0032] The method can further include reconstructing an image based on a subset of projection images. Different such images can be reconstructed per bin, one or more for each state.

[0033] In yet another aspect, a computer program element is provided which, when executed by at least one processing unit, is adapted to cause the processing unit to carry out the procedure.

[0034] In another embodiment, at least one computer-readable medium is provided on which the program element is stored.

[0035] The proposed system and method enable gating without additional hardware, such as an ECG (electrocardiogram) device for cardiac gating or the aforementioned pressure belt for respiratory gating. The proposed method is "device-free" because it performs gating solely based on analyzing information in the captured projection image.

[0036] Additionally, the proposed approach is more responsive and requires less CPU and memory, as it eliminates the need for additional reconstruction, which can otherwise be computationally intensive. The proposed approach avoids a "detour into the image domain" beyond the reconstruction required based on the projection data limited by gating.

[0037] For example, it is conceivable to achieve image-based, device-less gating by first performing a reconstruction that creates a set of (partial) image volumes, followed by an image-based estimation of the associated breathing states in the image area. Based on the breathing state of each partial volume, the projection data corresponding to the partial volumes belonging to the same breathing bin can be merged and passed to a second reconstruction, from which the (complete) image volumes can be obtained. However, thanks to the proposed system and method, the computational effort can be reduced. Thus, the proposed system and method are not only device-less but also "reconstruction-less," since the first reconstruction step can be omitted, enabling a responsive return with lower computational and memory requirements.The method and system can be run, at least partially, on a mobile computing device (smartphone, tablet, etc.), which is typically less powerful than a server. Only the final reconstruction based on the gated data can then be offloaded to one or more such servers, creating a highly responsive system that can even run on thin clients at the edge, with only the single reconstruction phase performed in the cloud if necessary. Therefore, as provided herein in preferred embodiments, the gating takes place entirely within the projection domain, without reconstruction.

[0038] The proposed method is also robust against interference from objects that appear and reappear in the field of view (FOV) during acquisition, as objects larger than the scan FOV are very common in practice. Therefore, such objects, like covers, cables, and even the patient table itself, appear to "move" in and out of the detector's FOV. The proposed approach is more robust because the in-image (projection) detection is preferably limited to the analysis of projection footprints of features (preferably anatomical features) that always remain within the beam, e.g., in the helical embodiment, when footprints of the chest wall or similar features are detected, and are used as the basis for binning / gating.Particularly during respiratory movements, the chest wall can be used as a substitute for the phase / motion state, and it has been shown to be readily identifiable, for example, in lateral projections. However, for other applications, other anatomical features, reference lines, and / or projection directions can be used to ensure this advantage of additional robustness against transient in-field (FOV) image features.

[0039] In some embodiments, respiratory trigger points are determined based on the attenuation of a single reference line / beam (or alternatively, a set of such reference lines / beams) and used for binning to specific states, such as respiratory states. A similar approach can be used in imaging for cardiac states or for other anatomical regions of interest (ROIs) that are subject to movement during projection data acquisition. Thus, the proposed system can be used for either respiratory gating, cardiac gating, both, or gating with respect to another type(s) of movement.

[0040] "User" refers to a person, such as medical or other personnel, who operates the imaging equipment or supervises the imaging procedure. In other words, the user is generally not the patient. It is the patient being imaged.

[0041] "Object" refers to a part of the patient, such as an anatomical structure, an organ, or a group of anatomical structures or organs. An object may include a region of interest or an anatomical feature.

[0042] In this context, "state of movement" refers to a (momentary) position / orientation of at least one part of the anatomical feature.

[0043] "Bin" includes a specification of the motion state. States that are sufficiently similar are in the same bin. Bins can be defined as one or more intervals of angular or positional data (length, deformation) used to track motion states during in-image recognition.

[0044] “Binning / Gating” is the process by which a single image (part of the projection images) is assigned to a specific bin.

[0045] “Anatomical feature”: any anatomical part of anatomy, tissue, etc., which is appropriately influenced by movement and can serve to detect states of movement. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Exemplary embodiments of the invention are now described with reference to the following drawings, which, unless otherwise indicated, are not to scale, wherein: Fig. Figure 1 shows a schematic representation of a tomographic imaging setup; Fig. 2 shows the recording of projection images that can be recorded using an axial scan method; Fig. Three projection images are shown that can be recorded using a helical scan method; Fig. 4 shows a periodic movement that can be extracted from recorded projection images; Fig. 5 A schematic block diagram of a motion compensator shows how it can be used in one embodiment in a tomographic imaging system; Fig. Figure 6 shows a flowchart of a computer-implemented procedure for processing projection data for tomographic imaging and Fig. 7 an illustration of the procedure of Fig. 6 is. DETAILED DESCRIPTION OF EXECUTION FORMS

[0047] First, we will look at the block diagram of Fig. Figure 1 refers to a medical imaging setup MIA. The setup MIA preferably includes an X-ray tomographic imaging device IA capable of acquiring projection data λ of a patient PAT located in an examination region ER. The longitudinal axis of the patient is preferably aligned with the rotation axis Z of the imaging device IA (“imager” / “scanner”), which is located in Fig. 1 is specified as axis Z. Therefore, axis Z is used here to denote both the longitudinal axis of the patient and the rotational axis of the imaging device IA.

[0048] The projection data λ generated by the imaging device IA, referred to herein as projection images, can be transmitted wirelessly or via a wired connection to a computing system SYS, which processes the projection images λ to generate tomographic cross-sectional or volume image data V. This data can be stored in memory, visualized by a visualizer VIZ on a display device DD to provide information for therapy or diagnosis, or used for planning if necessary, or such image data V can be processed in other ways. This mainly concerns a time series of 3D volumes, sometimes referred to as "4D" imaging. A stream or video feed Vt can be obtained by reconstructions from a stream of projection images.However, a reconstruction at a single point in time, as a still image, is not excluded here, and in some applications it is actually provided for.

[0049] The SYS computing system can be integrated into an operator console OC of the imaging device IA or can be integrated into another computing device PU, such as a workstation connected to the imaging device IA.

[0050] Patient movements (PAT) during image acquisition can significantly impair the quality of the cross-sectional or volume images (V) reconstructed from the projection images (λ) in the image area. Therefore, as proposed herein, the computing system (SYS) can include a motion compensator module (MC) ("motion compensator") capable of compensating for patient movements (PAT) through gating. Before explaining the operation of the motion compensator (MC) in more detail, reference is first made to the components of the imaging device (IA) to facilitate the explanation of the motion compensator and provide more context.

[0051] The tomographic imaging system IA is configured for multidirectional acquisition of projection images λ. The imaging system IA is therefore capable of acquiring projection images λ along various projection directions α with respect to the examination region ER and thus with respect to the anatomical region of interest (“ROI”) of the patient.

[0052] In some embodiments, the image acquisition can be performed by a rotation system in which at least the X-ray source XS is arranged in a movable gantry MG. The movable gantry (and in some embodiments, the X-ray source XS with it) is rotatable within a stationary gantry SG around the examination region ER, in which the patient / ROI is located during imaging. Opposite the X-ray source, the X-ray detector XD is arranged in the movable gantry. The detector XD can rotate with the gantry and the X-ray source around the examination region ER and around the axis of rotation Z to achieve different projection directions α. The source XS can follow a scan path during the acquisition, while in some embodiments, the detector XS can remain in an opposite spatial relationship across the examination region ER and thus the ROI.

[0053] As in Fig. As shown schematically in Figure 1, the patient's longitudinal axis, or the imaging axis Z, can extend into the examination region ER during imaging. The patient PAT can lie on a patient support PS, such as a bed / couch, which is positioned at least partially within the examination region ER during imaging.

[0054] The imaging device IA can be a CT scanner, as in Fig. Figure 1 illustrates this and can be used for diagnostic purposes. However, other tomographic imaging equipment, such as C-arm or U-arm scanners or similar, or cone-beam CT arrays, etc., are not excluded herein. C-arm cone-beam imaging devices may be used in some configurations, such as during therapeutic sessions in catheterization labs (cathlabs), trauma bays, or elsewhere.

[0055] The ability to perform multidirectional data acquisition does not necessarily result from a rotation system, as is the case in Fig. Figure 1 shows that non-rotating imaging systems (IA) are also provided, such as those found in 4th or 5th generation CT scanners, where multiple X-ray sources are arranged, for example, in a source ring around the examination region. Additionally or instead, the detector (XD) can be arranged as a detector ring around the examination region. Thus, in such systems, neither the X-ray source (XS) nor the detector (XD), nor both, rotate. In such a scanner, there is only apparent rotation, not physical rotation. Nevertheless, for each generation of CT scanner, reference is made here to the "axis of rotation" or "rotation," assuming that the rotation of the source (XS) can be both physical and apparent.

[0056] There may be an operating console (OC) through which a user, such as medical personnel, controls the imaging process. For example, the user can request the initiation of image acquisition, or request reconstruction or other operations, or initiate data transmission to the computing system (PU) or stop such transmission as needed, or set imaging parameters, such as pitch, or settings of the tube (XS), such as current and / or voltage, etc.

[0057] During imaging, and while the source XS is rotating or between phases of such rotation, an X-ray beam XB is emitted along the various projection directions α from a focal spot of the X-ray source(s) XS. The beam XB traverses the examination region containing the patient. The X-rays interact with the patient's tissue. This interaction modifies the X-ray beam XB. In general, such modification of the X-ray beam XB involves attenuation and scattering of the original, incident X-ray beam. The modified X-rays are then detected as a spatial distribution of varying intensities at X-ray-sensitive pixels of the detector XD. The projection direction α can denote a mean projection direction, i.e.,The angle formed by the imaginary line extending from the focal spot of source XS to the center of detector XD, with respect to a zero position, such as 12 o'clock or another position. Particularly in divergent imaging geometries, a given projection direction α can be represented by a geometric line extending from the focal spot of an X-ray source XS of imaging device IA to a specific detector pixel of detector XD of imaging device IA. Thus, in a divergent imaging geometry, different pixels can belong to different projection directions.

[0058] It is not necessary to acquire projection images λ over a full 360° angular range around the examination region ER. Acquisition over a partial angular range, such as 270°, 180°, or even less, may suffice. The X-ray detector is preferably configured for acquiring 2D projection images with rows and columns of intensity values ​​registered by the detector pixels. That is, the detector pixels themselves may be arranged in a matrix configuration. Such a 2D configuration can be used with divergent imaging geometries, such as a cone or fan beam, or other geometries. However, one-dimensional detector pixel configurations (such as along a single detector row) are not excluded, nor are parallel beam geometries. The 1D or 2D detector pixel configuration may be curved or planar. Digital 2D detectors are preferred.The pixel arrangement defines the data space (“projection area”) in which the projection images are defined. The detector pixel arrangement can be referred to here as the (projection) image plane or, more generally, as is mainly used here, as the (projection) image surface. By convention, detector pixel columns generally run parallel to the rotation axis Z, while detector pixel rows run perpendicular to the rotation axis Z.

[0059] The RECON reconstructor implements one or more reconstruction algorithms to process the projection images. Specifically, the RECON reconstructor can calculate a cross-sectional image V of the examination region (with the patient within it) for diagnostic, therapeutic, or other purposes. The RECON reconstructor can generate section-by-section volumetric image data (“image volume(s)”) V. However, this does not preclude the creation of a single image slice within the examination region ER, if required. Therefore, the reconstructed images can be designated V, which can encompass the entire volume, a partial volume, or a specific section within it. Volumetric reconstruction can be facilitated by the helical motion and / or the 2D arrangement of the X-ray detector XD.

[0060] In general, reconstruction algorithms implement a mapping that maps projection images λ, located in the projection area, to the image area. The image area is a section of 3D space and is located in the examination region ER of the imaging device, while the projection area is located in a different 3D space defined by the 2D detector and an angular measure (the projection direction) and is located at a (X-ray) detector XD of the imaging device IA. Conceptually, the image area consists of points of a 3D grid (voxels). The reconstruction algorithm, whether of type FBP (filtered backprojection), algebraic, or iterative, fills the voxels with image values ​​(such as HU values) to construct the image volume V or image slice at a given time.

[0061] Regarding the patient's movements, these can be, upon closer examination, voluntary / controllable by the patient or involuntary / uncontrollable. For example, the patient may be fidgety, like a child, or may be unable to comply with the requirements for other reasons. Other patient-controllable actions are desirable, such as planned movements or holding their breath during chest imaging. Involuntary / uncontrollable movements can result from biological activity, such as movements caused by respiration. Such activity causes movements within the body that can be transmitted to other anatomical structures, so that the movement may even be perceptible from outside the patient's body. An example of this is respiratory-related movements, which can include movements of the chest, rib cage, parts of the abdomen, the diaphragm dome, and so on.For example, the diaphragmatic dome is repeatedly raised and lowered during the respiratory cycle. Another example of involuntary, uncontrollable movements is the cardiac cycle, since the pumping action of the heart sets not only the heart itself in motion, but also potentially surrounding tissues and organs. Other movements that are uncontrollable by the patient can include those caused by peristalsis, tremors, etc.

[0062] If the motion occurring during projection data acquisition is not taken into account, image artifacts can appear in the image volume or tomographic image V reconstructed from the projection images λ. This is because the reconstruction algorithm is confronted with inconsistent data, as the same point of tissue appears at different voxel positions during movement. This inconsistency in the acquired projection data λ can cause the reconstruction algorithm, particularly an iterative type, to converge to suboptimal solutions when populating the voxel grid with image values ​​to create the tomographic image V. This, in turn, can lead to false image structures and artifacts that may obscure important clinical image information, rendering the image diagnostically unusable or unsuitable for the intended medical purpose in some cases.This can in turn lead to repeat imaging, which is expensive and slows down image throughput, a disadvantage for busy clinics. Additionally, unnecessary repeat imaging incurs dosage costs for both the patient (PAT) and the medical user (staff), thus posing a health risk.

[0063] Therefore, motion artifacts in medical images are something that must be avoided or mitigated, and the proposed Motion Compensator MC is configured precisely for this purpose. The Motion Compensator MC can be implemented as hardware, software, or both. For on-site / online use during the imaging session in the examination room, the Motion Compensator MC can be integrated into the operating console (OC) of the imaging system. Alternatively, the Motion Compensator MC can also be used later, for example, in a workstation (WS / PU), where images can be motion-compensated at any time after the imaging session, such as when a radiologist, as the user, reviews the image data as a basis for their findings.

[0064] While the anatomical feature AF moves in some way, projection images λ are acquired. The projection images λ comprise individual images captured at different times and with different projection directions during the acquisition. In simplified terms, the motion compensator MC uses a post-acquisition (retrospective) gating mechanism in which a gating component GC analyzes the acquired projection images λ into groups or "bins" (subsets of individual images). Thus, each of these bins contains a specific subset of individual images of the projection images. Each bin, and therefore its individual images, is / are associated with a specific motion state of an anatomical feature AF, as explained in more detail below.In particular, the projection images are analyzed with respect to the movement of the anatomical feature AF relative to a reference line ℓ, or at least relative to one such line ℓ. The reference line (or any of a plurality of such lines) is a geometric line in the image area. It can be predefined or user-defined. For example, the reference line ℓ can be defined by a user via a user interface (UI), such as a graphical user interface (GUI), where the user can draw the line during a simulation of the study region. The line can also be provided in other ways, such as being calculated automatically, as explained in more detail below.

[0065] The reference line ℓ is generally located within the examination region ER and is preferably chosen such that, during patient movement (particularly of the anatomical feature AF), an edge (a gradient) is detectable in the projection images. Such an edge represents a sufficiently pronounced attenuation gradient that is movable with respect to the reference line ℓ, for example, along the reference line ℓ. The anatomical feature AF may include the diaphragmatic dome or the chest wall. In such or other cases, an example of such a reference line ℓ, as provided herein, includes a line passing at least temporarily through the diaphragmatic dome AF parallel to the axis Z. Another example includes a reference line passing at least temporarily in the AP direction through the patient's chest wall AF.

[0066] For a specific such reference line ℓ, there is at least one corresponding projection angle α on the scan path. O, This allows the edge moving along the reference line to be tracked in the associated projection λ(ℓ) of reference line ℓ, since it can be recorded in the corresponding single-frame projection area. In the example of reference line ℓ passing through the diaphragmatic dome, this projection angle α can correspond to the AP projection direction. In the example reference line covering the chest wall, the corresponding projection direction α is lateral. The reference lines ℓ and their associated projection directions α, as provided herein, enable the detection of such high-contrast edges, thus allowing robust tracking and therefore robust binning. It is understood that any reference herein to such an angle α includes a reference to its complement at α+180°, as may be the case with full-angle scans.

[0067] Motion states can correspond to a specific configuration that the anatomical feature AF assumes at a given time during the movement. The movement, as considered herein, can be periodic and extend over several cycles. A motion state, or "bin," to which a respective subset of the projection images is assigned, can thus correspond to a particular phase of a periodic movement. Individual frames of the projection images form the subset for this motion state, or bin. Preferably, some or all bins include individual frames recorded in different projection directions α. The number (≥ 2) of bins can be predefined by the user via a suitable user interface. Thus, the individual frames of the projection images for a given bin represent the region of interest to be imaged, such as the patient's lungs, thorax, or heart, in a given motion state.The RECON reconstructor can then reconstruct a version of the volume for that given state of motion based solely on the individual images from the bin associated with that state of motion. Thus, if any number of states / bins exist, each with its own set of projection images, a different version of the tomographic image can be reconstructed, representing the region of interest in a different state. A user, such as a radiologist evaluating the results, can then select the state that best serves the medical purpose at hand.

[0068] It should be noted that the gating principles described herein, which are based on the projection area, are not necessarily limited to periodic movements. Cases of non-periodic movement are also provided for. Such a case of non-periodic movement may occur, for example, if the patient is unable to hold their breath for the entire recording.

[0069] This document proposes that the motion compensator component, and in particular the gating component GC (hereinafter referred to as ), Fig. (explained in more detail in section 5) is only operational in the projection domain and therefore works exclusively with structural image information in projection images λ. For the proposed motion compensator MC, no image information from the image domain, in particular no partial or preliminary reconstructions of the images, are required. This significantly reduces the load on CPU time and / or memory, because the execution of reconstruction algorithms, especially of the iterative type, can be demanding, whereas the proposed setup requires only one reconstruction per state, and not two as in previous approaches.Thus, the proposed motion compensator module MC enables the avoidance of unnecessary reconstructions and is therefore faster and more responsive, which is an important consideration especially for online / on-site / in-room applications, particularly for busy medical facilities and situations where near real-time image output is required, such as in trauma operations.

[0070] In general, the motion compensator MC uses only image projection information encoded in projection images acquired over time to identify the anatomical feature AF and its motion states as projections. For example, in some embodiments, the anatomical feature may include the diaphragm or another suitable landmark that is affected by the motion and thus serves as a suitable substitute. Therefore, the anatomical feature AF is not necessarily the feature that causes the motion, but rather a feature that is affected by the motion and functionally related to such motion, which may originate elsewhere. The manner in which AF is affected serves as an indicator or substitute for the motion.Therefore, the anatomical feature AF, as used here, primarily serves to identify motion information in order to compensate for it. The anatomical feature AF is generally distinct from the region of interest (ROI), the latter being the purpose and object of interest of the imaging. For example, the anatomical feature AF might be the diaphragm, while the ROI encompasses one or more of the lungs, as in a chest scan. The ROI can define a space within the 3D image area that includes other parts (anatomical structures, organs, etc.) of the patient, not just the AF.

[0071] The time-based analysis of the projected images, performed using the AF orientation aid, can be based on segmentation, optical flow, machine learning, intensity thresholding, or any other image processing algorithm that allows tracking the movement of the anatomical feature AF over the time series of images acquired over time within a specific field of view. This time-based tracking of the footprint (shadow) of the anatomical feature AF across movement cycles allows for the identification of the different phases / states of movement. Since each phase corresponds to a given state of movement, this enables the mapping of a given single image representative of that phase to the corresponding state of movement / bin, as previously explained, thus performing binning.Time-based analysis enables the creation of a time profile that can be monitored for gating trigger events to assign a given frame to a phase / bin. Monitoring and defining the time profile are based solely on image information from the projection images. As explained in more detail below, the time-based analysis of the projection images is dependent on the reference line ℓ. Since the disclosure contained herein is not limited to periodic motions, "phase" (hereafter sometimes referred to as "φ") is occasionally used as another term for "(motion) state".

[0072] The embodiments of the motion compensator MC described herein may depend on the operating mode or configuration of the tomographic CT imaging system. One operating mode described herein includes axial scanning, sometimes referred to as step-and-shoot (“S&S”). Another operating mode includes helical scanning, preferably with a conical beam geometry, although such a conical beam geometry can also be used in S&S. The different operating modes define different scan orbits, which are traversed in at least one rotation by the source XS, as in CT scanners up to the fourth generation. However, such rotation can also be only apparent, as in fifth-generation CT scanners, where, as already mentioned, a source ring around the scan region is used instead of a rotating source, with a detector ring providing multidirectional imaging.The rotation can include one or more full rotations of 360° or at least 180°.

[0073] In embodiments of axial methods, the X-ray source rotates around the anatomical feature in a given plane, acquiring projection images in that plane. The scan paths in each plane can be circular, but other geometries are also possible. The rotation then stops, and a relative translational movement is performed along the Z-axis between the source XS and the patient (e.g., the patient table is translated along Z) until the next z-position is reached. The source then rotates again in a new plane that intersects the Z-axis at the new position z' to continue acquiring projection images, and so on, alternating between translation along the Z-axis and rotation around it.Individual images collected over a range of different z-positions then form the projection images, which are analyzed by the motion compensator MC. Helical methods differ in that the X-ray source follows a helical scan path with the Z-axis as its central axis. During the acquisition, the rotation of the source XS around the Z-axis and its relative translation along this Z-axis occur simultaneously.

[0074] The following Fig. 2, Fig. 3 to Fig. Figure 4 illustrates concepts underlying the operation of the MC motion compensator, while the block diagram of Fig. 5 and the flowchart of Fig. Provide 6 more details.

[0075] The illustrations in Fig. 2 refer to an axial scan method with S&S, while Fig. 3 and Fig. 4 refers to a helical scan method.

[0076] In general, and now with reference to Fig. 2, Fig. 3 to Fig. 4 In its entirety, the proposed motion compensator MC is configured to provide reliable binning according to motion phases or, more generally, in the case of non-periodic motions, according to motion states, using precisely defined trigger points.

[0077] More precisely, the operation of the motion compensator MC can be arranged as a two-stage processor: in a first stage, based, for example, on the reference line ℓ, the projection data selector SEL selects suitable frames from the stream of projection images λ. Specifically, each frame whose projection plane includes a projection λ(ℓ) (of the reference line ℓ) is selected, stored, or marked. This set of marked frames is then analyzed by an analyzer stage that includes an in-image detector component DC, configured to generate profiles over time with respect to the footprint of the anatomical feature AF, as encoded in the marked frames, relative to the projection λ(ℓ). The profile(s) can then be monitored / analyzed for trigger points to effect gating.

[0078] First, the S&S method will be discussed in more detail. Fig. Figure 2 is used to illustrate the selection process based on the reference line ℓ of the acquired projection images λ. The concept is illustrated for the anatomical feature AF of the diaphragm, which has proven to be a good motion substitute for thoracic or chest imaging in general. For other regions of interest, the selection of a different anatomical feature AF as a motion substitute or indicator may be more appropriate.

[0079] In the embodiments described herein, two options A) and B) are provided. In general, the reference line ℓ can be determined from the motion vector. v→ its movement depends on the intended anatomical feature AF, such as the diaphragm. Its movement can be characterized as a linear upward and downward motion along or parallel to the patient's longitudinal axis, which can be readily observed, for example, in the frontal or coronal projection plane. The schematic view of Fig. Line 2 runs perpendicular to the frontal plane along the anterior-posterior axis (AP axis). The two lungs (LS) and the diaphragm (AF) are shown schematically. The diaphragm (AF) alternates between two extreme states and passes through a number of intermediate states as it relaxes and contracts in cycles, with exhalation and inhalation represented by the solid and dotted lines and movement by the vector shown as a double arrow. v→ be specified.

[0080] The reference line can be set, for example, by a user via the user interface (UI) along the AP axis, perpendicular to the plane / axis of diaphragmatic movement. v→ be defined.

[0081] Preferably, however, the line selection is performed automatically. For this purpose, the SYS system can include a path definer PD configured for such automatic selection of a reference path, such as one (or more) of the lines ℓ. The operation of the path definer PD can be based on anatomy-guided image processing of an overview image. Additionally or instead, attenuation values ​​of possible lines can be determined based on the overview image S. The position of the reference line can be taken where a transition of attenuation values ​​across several lines is observed. Instead of the overview image, one or more of the projection images can also be used, particularly with the helical scan method. For S&S, the overview image S can be used by the path definer PD. The overview image is located in the projection area (2D) or in the image area (3D).More precisely, and with regard to some proposed embodiments of the path definer PD (which may also be referred to herein as the line definer, since the path in preferred embodiments is, as mentioned, a line), the approximate positions of the lung centers are detected in the overview image according to some, but not all, embodiments, followed by a search for possible lines in the caudal direction of the patient. Other anatomical directions are also provided, such as the AP direction or others. For some or each of these possible lines, an associated attenuation value is determined. If a possible line is located above the diaphragm, the corresponding attenuation value remains below a certain threshold. If a possible line passes through or at least comes into contact with the diaphragm or liver tissue, its attenuation value is higher. The position at which this transition occurs is recorded.If the respiratory state in which the overview image is acquired is known, an adapted position can be determined by a shift, for example, cranially (if the overview image is acquired at or near the end of inhalation) or caudally (if the overview image is acquired at or near the end of exhalation). If the respiratory state is unknown, such a shift is applied in both cranial and caudal directions, and two adapted positions are determined. The principles outlined above—observing attenuation values ​​and shifting along an anatomical line to find the reference line—can also be applied to other anatomical features of respiratory function if necessary.

[0082] As stated, the position and / or direction of the reference line in the 3D image area generally depends on the motion geometry of the anatomical feature AF. If the motion of the anatomical feature is confined to a plane, as assumed herein in some embodiments, the projection direction α for the reference line ℓ can be, at least to a good approximation, substantially perpendicular to such a plane. If the target motion of feature AF is not confined to such a plane, it may be sufficient to consider the projection of one component of the motion onto such a plane, since for the purposes presented here, such consideration may be sufficient to determine motion states / phases.

[0083] The path definer PD can include a trained machine learning model, such as an artificial neural network or something else. The model can be trained based on training data. The training data can include patterns of different reference lines as training targets, each associated with specific training input data patterns. For example, the training targets can be provided as annotated training overview images. The annotations can specify specific patterns of reference lines. Alternatively, the targets can specify patterns of such reference lines, either directly in the projection space or in the image space. The respective annotation can be provided by a human expert in the form of the pattern reference lines marked (e.g., drawn) in the training overview images or in the image / projection space.The training data may also include the respective training input data patterns, such as descriptions of the imaging objectives and / or overview images (but not yet annotated). The imaging objectives may include any one or more of the intended type of imaging or gating purpose, the organ of interest for which the imaging or gating is to be performed, etc. After training, in use, the user can provide input data for use via the user interface (UI) or another user interface. "Use" is the application of the trained model for its intended purpose, for example, in clinical practice. The input data for use may include a description of the imaging objective and / or the overview image.The input data for the application is passed via the user interface to the trained machine learning (ML) model of the path definer (PD). The ML model can process the user-provided input data to calculate the corresponding reference line in the image area or, preferably, directly in the projection area. For example, the calculated reference line can be defined in the overview image. The calculated reference line can be displayed for quality control by the user, but this display is optional.

[0084] Regardless of how the reference line is provided, it allows the selection of the captured projection images, as already mentioned. Now, referring to option A), which is on the right in Fig. As illustrated in Figure 2, preferably some or all of the individual frames whose image area / plane intersects the projection direction α associated with the reference line ℓ can be marked for motion analysis of motion profiles over time. In embodiments, a change in the intensity of the individual frames at the interaction point is used to trigger a gating event. An intensity profile is monitored for events to trigger gating events and perform binning of the current individual frame. For example, line ℓ may lie above the diaphragm AF during the inhalation phase. Consequently, the diaphragm will cross this line until the exhalation phase, thus causing an intensity change as indicated by the measurement of a periodic intensity profile I(t).This intensity change is periodic due to the oscillating movement of the diaphragm and is schematically indicated by the varying shading of the intersection point, which alternates between black and light. Therefore, monitoring the intensity curve I(t) at the intersection point during such a diaphragmatic crossing event can be used to determine the trigger point for binning the individual frame for which such an intensity change is detected. Repeatedly triggering the frames as they cross the line binns the projection data and thus effectively provides the gating mechanism described here. Alternatively, the reference line can be set below the exhalation phase.

[0085] This principle can be extended to define multiple reference lines for triggering different phases. Additionally or instead, a multitude of such lines can be used for the same phase, with triggering based on combining the respective intensity measurements, averaging, or similar methods to more robustly determine the correct gating trigger point. However, it has been observed that a single such reference line offers high computational responsiveness and memory savings. Therefore, for the present purposes, a single line is sufficient, and the position of the diaphragm or other anatomical features is estimated based on transitions in attenuation values ​​(e.g., dark-to-light transitions) along the projection λ(ℓ) of line ℓ located in the (selected) projection images.Thus, a single reference line ℓ is sufficient to determine the temporal position that represents the same respiratory, cardiac, or other anatomical state of motion.

[0086] The correct placement of the line (e.g., above the inhalation phase to ensure that crossing events actually occur) must be considered. For this purpose, a 2D or 3D overview scan can be performed on the patient during a preparation phase (before diagnostic imaging at a higher dose), using a user interface (UI), such as a GIU, which allows the user to mark the projection λ(ℓ) of the reference line ℓ in the overview scan S. This can be done based on the user's medical / anatomical knowledge. Alternatively, or in conjunction with such user-defined placement, a suitably trained machine learning algorithm can be used for this purpose, taking the intended anatomical feature AF as input, and the algorithm automatically placing one or more such lines in the overview scan.The machine learning algorithm can be based on a regression model, for example, an artificial neural network (ANN), particularly in a convolutional configuration, to better utilize spatial correlations expected in image data. Alternatively, or additionally, the previously described image processing based on "possible reference lines" can be used.

[0087] Now, referring to option B), which is on the left in Fig. As illustrated in Figure 2, this can be based on tracking the projection footprint of the diaphragm AF as a geometric shape across the projection images, particularly individual images where the AP direction is perpendicular to the plane of motion in which the projection λ(ℓ) of the reference line ℓ lies, and monitoring the displacement profile Δ(t) of the footprint to establish trigger points. The projection footprint of the diaphragm AF can be located in a specific, a priori known section of the detector XD image area, such as a particular detector column or group of columns, or otherwise. In embodiments such as those illustrated in Figure B), the projection footprint of the anatomical feature AF (highly schematically indicated by a black dot) can be determined, for example, by segmentation, optical flow methods, etc.The footprint can be located and tracked across multiple frames selected based on the reference line ℓ, as described above in A). The apparent movement of the footprint across the selected frames can be graphically represented as the displacement profile Δ(t) over time. Since this is periodic, phases can be determined, which in turn can be used as trigger points for gating.

[0088] In A) and B), only parts of the selected individual images are shown. The other parts of the individual images are at 180° and are preferably also taken into account in the motion analysis.

[0089] Now, referring to Fig. Three options for a helical scan path H are shown, as on the left in Fig. 3 specified over time, with the source XS rotating around the imaging axis Z.

[0090] The upper part of Fig. Figure 3 offers a perspective along the imaging axis, which extends perpendicularly into the drawing plane. Fig. 3 extends, where the rotation for image acquisition is indicated by a curved line around an isocenter IS located within the patient PAT. (X, Y) generally denotes image planes in the image area, with the imaging axis Z perpendicular to them generally being aligned with the longitudinal axis of the patient.

[0091] A scout scan S in the sagittal view of an anthropomorphic imaging phantom is illustrated in B). The gating procedure for helical scans is similar to that for axial scans as described above in 2), but is preferably repeated at multiple positions z along the imaging / rotation axis Z. Such a z-position is indicated by the line in the AP direction (drawn horizontally in section B). Subsequently, motion analyses are performed for each of a multitude of such z-positions, and the results are then combined to obtain such good, robust gating results as shown in Fig. Figure 4 illustrates this. The selection of z-positions could be random, as long as all selected points lie within half the collimation width. A selection of 3 or more points, such as 4, 5, or any other number, is preferred. Such computational processing based on multiple z-positions is useful for helical acquisition mode because both the changing respiratory state and the changing cone angle contribute to the attenuation measured for a single beam (or for multiple beams) around the anatomical feature, such as the diaphragmatic dome or the chest wall. Therefore, a more robust method for extracting the respiratory signal is proposed here for helical acquisitions. To this end, the fact that in helical mode the pitch is usually very small, for example, on the order of 0.04 to 0.10, is exploited.Thus, in this example, each voxel remains within the radiation cone for 10 to 25 rotations. "Pitch" is an imaging parameter of the imaging device and refers to the effective translation along the Z-axis of rotation per 360° rotation divided by the collimation width.

[0092] To better account for the effects observed in helical scans in this embodiment, the reference line ℓ can run along the lateral-medial axis (LM axis). Two sets of lateral images λ L,R , each 180° apart, can be defined as follows: a left set (“L”) of frames at α° and a right set (“R”) of frames at α' = 180° + α, as illustrated in sections A) and C). Figures A) and C) illustrate different frames over time. In the example of Fig. 3. The anatomical feature of interest AF is the chest wall, but it could just as easily be the diaphragm, the upper abdominal wall, or something else. The same applies to Fig. 2, since the chest wall or upper abdominal wall may have been chosen instead of the diaphragm. Here too, as already mentioned in Fig. 2, that the projection direction α, which is connected to the reference line ℓ, is essentially perpendicular to the motion vector v→ runs, but now at 90° to the direction α of the reference line in Fig. 2 is. The projection λ(ℓ) of reference lines ℓ for each individual image is shown as a dashed line. Fig. 3A), C) illustrate projection images from left to right, λ L , or right, λ R The dashed line indicates the respective position at which the chest wall position should be measured.

[0093] For the projection images, a re-binning of the image geometry from the native geometry of the divergent cone beam into a parallel geometry or another non-divergent geometry, such as wedge geometries, is preferably performed. Such wedge geometries are discussed, for example, by P. Koken and M. Grass in Phys. Med. Biol. 51 (2006), pp. 3433 to 3448.

[0094] Here too, as in Fig. 2, Option B), segmentation, optical flow, etc., are used to track the chest wall AF footprint across frames and to define the displacement profile Δ(t) to establish trigger points for gating / binning based on phases of the periodic displacement profile. Maxima and minima or other points can be used for gating. The above is repeated for multiple z-positions, as in Fig. Figure 4 illustrates this, and will now be discussed in more detail. As an example, five displacement profiles Δz(t) for five different z-positions z1 to z5 are shown. In particular, Fig. Figure 4 shows a schematic representation of chest wall positions at some selected locations as a function of scan time. Since the distance between the z-positions is half the collimation, the measurements overlap by 50%. A periodic respiratory signal is obtained at each z-position. Minima and maxima of overlapping signals align, but their absolute positions differ because a different chest wall position is being analyzed. It should be noted that, due to the small pitch, each curve is interpolated from typically 10 to 25 individual measurements, which provides additional robustness.

[0095] We will now turn to the block diagram of Fig. 5 Reference is made to the components of the motion compensator MC.

[0096] At input port IN, projection images λ' acquired by the tomographic imaging unit IA and detected by its detector XD are received. These projection images are acquired as individual images from multiple directions around the feature of interest AF. For images with a divergent imaging geometry, such as a helical geometry and / or a cone-beam geometry, an imaging geometry rebinner RB is used, which rebinns the projection images into a parallel or at least approximately parallel geometry. Therefore, it is the projection images λ, in which rebinning of the imaging geometry has been performed, that are preferably processed here.

[0097] A projection data selector (SEL) selects individual images from the projection images (λ) for processing based on the reference line (ℓ). The reference line can be selected by the user. For example, a scout image (S) of a volume that includes the region of interest (ROI) and / or the feature of interest (AF) can be rendered. The user can interactively select one or more reference lines via a user interface (UI), such as a graphical user interface (GUI). The reference line can, for example, be perpendicular to the AP direction. Thus, the reference line (ℓ) can be defined perpendicular to the movement of the anatomical feature, although other angles are also possible. Preferably, but not necessarily, reference lines do not run parallel to the motion vector. For gating, the movement of the feature of interest (AF) is used as a substitute to account for patient movements.The anatomical feature of interest AF may be identical to the region of interest, but it may also differ from it.

[0098] An in-image detection component (DC) performs motion analysis in the selected individual frames, as described above. Motion profiles are obtained over time, representing the preferably periodic motion and its phases.

[0099] For example, an in-image detection component (DC) can detect projection footprints of an anatomical feature across pre-selected frames acquired over time. The DC can employ segmentation, optical flow techniques, machine learning, model-based segmentation (MBS), and other methods for this detection process. Alternatively, at intersections of one or more reference lines with the anatomical feature, changes in intensity value λ(ℓ) occurring over time within the projection area of ​​the projected line are recorded to obtain an intensity profile over time. A single reference line can be used, defining a single intersection point. Multiple lines can be used, and the above process is performed for each line, with the results being combined.

[0100] The detection component DC therefore performs a motion analysis with respect to the reference line and compiles a motion profile over time (time series) based on intensity values ​​I(t) or the displacement Δ(t) of projection footprints of the anatomical feature AF, as determined, for example, by segmentation. The above can preferably be used for axial scans. For helical scans, the above process of the in-image detection component DC can be repeated for a number of z-positions (preferably all within a collimation width used for the scan, such as half the collimation width), and the results can then be combined to obtain a more robust overall motion profile that includes a sufficient number of phases to allow for robust gating.The gating component GC uses phase points in such periodic profiles over time as trigger points for gating.

[0101] Suitable substitute anatomical features AF for motion have been found to be the chest wall, abdominal wall, diaphragm, or any other suitable anatomy or part thereof that preferably moves periodically; however, the processing of non-periodic motion is not excluded. Contrast-enhanced X-ray imaging is also provided, which can be used to account for cardiac motion, in which case segmentation is performed for specific segments of the coronary vessels that can be recorded in the corresponding projection images. The rotation time and pitch of the imaging device IA may need to be adjusted to the generally higher motion frequency induced by cardiac activity.

[0102] Preferably, but not necessarily, the motion is periodic, as is the case with movements caused by respiration or cardiac activity. The periodic profile over time for the motion states of the anatomical feature includes various phases. The phases of the periodic motion can be used to define the motion states, with different phases representing different motion states. Since, for better stability, the motion profile is preferably recorded over a multitude of cycles by the in-image component DC, the gating component allows for binning multiple projection frames into the same subset λ for a given phase / motion state. φThis is carried out. Different subsets contain individual images for different phases, while individual images within a given subset represent the same phase. These subsets of individual images λ φ These can then be output as gated subsets of projection images via the output interface OUT for further processing, such as storage or display on the display device DD by a visualizer, or for other processing. As mainly intended herein, the processing primarily involves reconstructed images V (volume or layer), and these reconstructed images V are displayed on the display device DD. In particular, for each φ, the respective gated projection images λ can be output. φ They can be reconstructed separately to create different image volumes V φto obtain images, each of which represents the anatomical feature and its surroundings in different states without motion artifacts. The user can then specify φ, and the system outputs the respective reconstructed images V. φ back (for example, it displays it).

[0103] We will now turn to the flowchart of Fig. Reference is made to Figure 6, which shows the steps for the practical implementation of the motion compensation principles described above. However, it is understood that the steps of the computer-implemented motion compensation process described below are not necessarily limited to the architecture discussed above. In particular, it shows Fig. 6 a diagram of a procedure for processing projection data acquired during tomographic imaging.

[0104] In step S610, an X-ray tomographic imaging device is used in a helical or S&S type scan to acquire projection images λ'.

[0105] When a helical and / or cone beam scan is used, the projection data can be subjected to re-binning of the imaging geometry into at least approximately parallel imaging geometries, and these re-binned projection images, which can now be referred to as λ, are then used in the remaining steps.

[0106] In step S620, such projection images, in which a re-binning of the imaging geometry may have been carried out for a different, at least approximately parallel, imaging geometry, are received in step S620.

[0107] In an optional step S630, individual images are preselected based on one or more reference paths (preferably a single such path). Since, as described above, lines are primarily provided as such paths, the reference path will continue to be referred to as a reference line, without limiting the present disclosure. The reference line preferably runs perpendicular to a motion vector that describes the movement of the anatomical feature of interest. The reference line(s) can be specified by a user based on an overview scan or can be calculated based on the imaging protocol, for example, using machine learning or a lookup table in which suitable reference lines are recorded in comparison to the imaging protocol (purpose of imaging, ROI, etc.). Machine learning can be used to define suitable reference lines.In other embodiments, reference tables can be used, which can be compiled from an anatomical atlas or other model. Additionally or instead, the attenuation-value-based processing / analysis of the projection images described above in relation to the path definer PD is used.

[0108] In step S640, in-image processing, based exclusively on image information in the projection images and on individual frames thereof, is used to perform a motion analysis over time with respect to one of several reference lines of the anatomical feature of interest AF, as represented by its projection λ(ℓ) in the projection images. A profile of one aspect of the motion over time can be recorded. The profile is preferably periodic, like the motion, and represents phases / states of motion, preferably over a multitude of motion cycles. Additionally or instead, the in-image analysis yields different positions of the projection of the anatomical feature, and each such position represents a state of motion.

[0109] The in-image processing step S640 can include a segmentation operation and / or use optical flow techniques to identify projection footprints of an anatomical feature across frames and / or in some or all of the selected frames, thereby deriving a representation of the motion that allows the identification of different phases of the movement. Additionally, or alternatively, the profile over time is a profile of the variation in attenuation intensity caused by the motion. The intensity variations can be determined at the intersections of the projection direction α for the reference line(s) ℓ with frames. However, the motion need not be periodic; non-periodic motion is also possible.

[0110] The phases shown in the profiles and / or the positions determined in step SS640 can be used to define trigger points for gating / binning in step S650. For individual images that represent the same or a sufficiently similar phase or the same or a sufficiently similar state of motion, binning is performed into the same bin representing the same state of motion. More generally, the different positions of the projections of the anatomical feature are used to define the bins. Any position that falls within a predefined set of one or more intervals is considered representative of the same or at least sufficiently similar individual images.

[0111] The individual frames of the original projection images, which have undergone re-binning in this manner, can then be passed on in step S660 for further processing in step S670, such as storage, display, or, if necessary, other processing. In the intensity-based embodiment, thresholding of intensity values ​​can be used in step S660 to define trigger points for gating.

[0112] Preferably, the different bins of projection images are provided in step S670 for reconstruction in step S670 in order to reconstruct (i.e. per bin) a respective tomographic volume Vφ, each based exclusively on individual images from the respective bin that is representative of a particular state of motion φ.

[0113] The proposed method can be used for helical and / or cone-beam imaging. For helical scans, the preceding procedure steps S630 to S640 are performed for various z-positions along the rotational axis of the imaging device covered by the helical scan. The different profiles thus obtained over time, namely one for each z-position, are then combined to obtain an overall profile that includes a sufficient number of phases from multiple cycles.

[0114] More precisely, referring back to option A) of Fig. 2. As referenced, in an axial S&S scan, step S630 in embodiments may include receiving the reference line based on prior knowledge (e.g., scout scan), wherein the reference line is located immediately above the diaphragm in the AP direction during inhalation and immediately below it during exhalation. Alternatively, other anatomical features AF that perform a movement may be used, with the line of interest being appropriately defined, for example, perpendicular to the movement vector.

[0115] In step S640, the attenuation for the selected line of interest (a single beam) is monitored over the acquisition time and stored as a profile, and the profile is analyzed to define trigger points that can be used in step S650 for gating / binning to obtain the respective sets for each trigger point.

[0116] Only a single ray can be considered and tracked, allowing for a responsive implementation. Expiration depth correlates with the value of the line integral; that is, during expiration, the diaphragmatic dome moves upward, and the line integral increases. Optionally, multiple rays can be tracked to increase the dynamic range of the process. For example, if a patient inhales more deeply than expected during recording, this condition will not be detected because the attenuation does not decrease further once the diaphragm is below the reference line. Using multiple such reference lines at appropriate locations increases the probability of the condition actually being detected.

[0117] Now referring to option B) in Fig. Step S640 can include identifying, based on prior knowledge (e.g., Scout-Scan S) or using machine learning methods, a portion of the detector area, such as a detector slit, that covers the diaphragm in the direction α of the reference line, e.g., the caudal-cranial (CC) direction of the patient PAT. Thus, in this embodiment, the reference line ℓ is parallel to the axis of rotation Z, and it is the projection λ(ℓ) of the reference line ℓ that is observed / tracked, preferably in projections that, in embodiments, are along the AP (anterior-posterior) direction α of the patient PAT (for an illustration, see also Fig. 7 below) or along lateral projection directions. The position of the diaphragm is estimated as displacements by segmentation in the AP projections, thus allowing a direct measure of diaphragmatic position to be calculated. The displacement vector of positions (across the respiratory cycle) is analyzed to define gating / binning trigger points for respiratory status.

[0118] Now, referring to a helical and / or cone-beam CT scan, this may include re-binning for a wedge geometry or other non-divergent imaging geometry.

[0119] In step S630, horizontal (i.e., lateral) wedge projections are selected along the ML axis.

[0120] In a step preceding step S640, a set of z-positions along the rotation axis Z is selected for which a motion analysis is to be performed, as in a case where the position of an anatomical feature, preferably the chest wall, is to be detected. The z-positions should be no more apart than half the collimation width. For some or all of these selected z-positions, the projection footprint of the anatomical feature in the detector image area is detected for each lateral projection frame. This can include calculating, for each lateral projection frame, the portion of the radiation-sensitive detector area onto which the anatomical feature, such as the chest wall at the respective z-position, is projected. Based on the segmentation, the displacement Δ (position of the anatomical feature) is estimated.For example, the position of the chest wall is shown on the dashed lines in the diagram. Fig. 3A), C) projections are determined. Based on the imaging geometry and knowledge of the reference line, it is possible to pre-estimate the relevant part of the detector area, thus saving computation time. Instead of searching for the projected reference line λ(ℓ) across the entire detector area, the search is limited to the pre-estimated portion. For example, in preferred embodiments, the reference line is projected onto different lines as the system moves along the helical path. In step S640, phases for the profile over the time of such a displacement Δ are estimated, such as minima and maxima of the estimated chest wall positions. This is performed for each z-position, and the phases are then combined, for example, by identifying matching pairs of minima and maxima across adjacent z-positions, as described above in Fig. Figure 4 illustrates this. The phases across the z-position can then be used to trigger the S650 gating.

[0121] To achieve good results, the anatomical feature AF is preferably selected so that it always and completely remains within the FOV. However, this may not always be achievable. Additionally or instead, the proposed method relies exclusively on image information that is always within the FOV. Features outside the FOV generally do not affect the proposed processing or its results. The proposed method is therefore robust to events and information from outside the FOV. This even applies to a single reference line ℓ, where a relevant portion (such as a point) of the anatomical feature AF preferably remains completely and at all times within the FOV during the scan.Thus, all relevant information for complete and robust binning can be obtained by tracking how the projection image value (attenuation) along this single line changes / fluctuates over time due to the movement of the anatomical feature AF, such as the diaphragm. In the diaphragm example, the single reference line ℓ is observed under projection α in the anterior-posterior (AP) direction and can be used to observe the passage of the diaphragm AF through it, as previously explained.

[0122] Fig. Figure 7 illustrates the previously described device- and reconstruction-free gating system and procedure, as disclosed herein. The source XR rotates along the scan orbit. O around the anatomical feature AF, while the anatomical feature AF (such as the diaphragm, as illustrated) moves along the reference line ℓ. Single images λ1,2 are recorded along different projection directions α1,2. At least one such projection direction α1 is connected to the reference line ℓ. In particular, the (reference) projection direction α1 is preferably perpendicular to the plane of the reference line ℓ. For the diaphragm AF, the projection direction α1 for the reference line ℓ runs along the AP direction of the patient PAT. In such embodiments, the reference line ℓ runs along the cranio-caudal direction and through the diaphragm AF. The motion state bin is determined based on the "reference" single images λ1 by tracking the projection λ(ℓ) in the "lead" single image λ1, as shown in the embodiments already described.

[0123] Thus, the binning / gating / assignment step S650 and the preceding analyzer step S640 can be performed as provided herein and in Fig. Section 6 is used as a framework for a two-stage data processing pipeline, comprising the analysis / in-image recognition step S640 and, based on analysis step S640, the mapping / binning / gating step S650. Mapping / binning / gating step S660 may include an interpolation step. More specifically, analysis step S640 determines the position / state of motion of the anatomical feature based solely on information in the reference frames with respect to the reference line ℓ (e.g., AP direction), such as along or across such a line. In mapping step S650, the state of motion can then be interpolated for some projection directions α2, between the positions according to reference line ℓ, depending on the number of desired bins. This makes it possible to obtain estimates of the motion state for the other individual images λ2, which are projected along other directions α2 (see Fig. 7) were captured. In the subsequent assignment step, some or all individual images can then be assigned to the respective bin into which their respective motion state (measured or estimated) falls, regardless of the projection direction in which they were captured. The bins can be defined as intervals in the angular domain or as intervals in the positional domain, e.g., as length, or the like. For other projection directions α1, such interpolation may not be necessary.

[0124] The respective bin for individual images λ2 along the other projection directions α2 can be obtained even more precisely by spatiotemporal interpolation based on several reference images λ1a, ... λ1z, which were recorded along α1 for the reference line ℓ over time during multiple rotations. Such images λ1a, ... λ1z represent the anatomical feature AF in the same state of motion, etc. In the spatiotemporal interpolation, the time of acquisition of individual images and their distance to adjacent bins are taken into account. If, for example, and more specifically, an individual image is recorded at a time between the acquisition times of two individual images belonging to the same bin, then this individual image is also assigned to the same bin, since the motion is assumed to be continuous.Otherwise, if the single image is recorded at a time between the recording times of two single images belonging to different bins, a spatiotemporal weighting (with optional rounding up or down) is performed to find the correct bin.

[0125] As used herein, the term “motion” can include translational movements, rotational movements, but also deformations or a combination of any two or all of the above.

[0126] Furthermore, it should be noted that while the above approach is illustrated based on a reference line ℓ, it also includes other, non-linear reference paths, such as curved paths. However, linear reference paths, like the described reference line ℓ, offer advantages in computational implementation and may be preferred in this context. In particular, the projection λ(ℓ) of the reference line ℓ defines a relatively narrowly defined area in the projection images within which the pixel value distribution can be analyzed for gating purposes. Moreover, it may be sufficient to consider components of more complex movements of certain anatomical features on such a reference line ℓ.

[0127] The components of the SYS system can be implemented as one or more software modules running on one or more universal processing units (PUs), such as a workstation associated with the image sensor IA, or on a server computer associated with a group of image sensors.

[0128] Alternatively, some or all components of the SYS system can be implemented in hardware, such as a suitably programmed microcontroller or microprocessor, an FPGA (field-programmable gate array), or a wired IC chip, an application-specific integrated circuit (ASIC) integrated into the MIA imaging system. In another embodiment, the SYS system can be implemented partly in software and partly in hardware.

[0129] The various components of the SYS system can be implemented on a single processing unit (PU). Alternatively, some or more components can be implemented on different processing units, possibly located remotely in a distributed architecture and connected via a suitable communication network, such as in a cloud environment or a client-server setup, etc.

[0130] One or more of the features described herein may be configured or implemented as or with a circuit arrangement and / or combinations thereof encoded within a computer-readable medium. Circuit arrangements may include discrete and / or integrated circuit arrangements, a system-on-a-chip (SoC) and combinations thereof, a machine, a computer system, a processor and memory, and a computer program.

[0131] In a further exemplary embodiment of the present invention, a computer program or a computer program element is provided which is characterized in that it is configured to execute the process steps of the method according to one of the preceding embodiments on a corresponding system.

[0132] The computer program element could therefore be stored on a computer unit, which could also be part of an embodiment of the present invention. This data processing unit can be configured to perform or bring about the execution of the steps of the method described above. Furthermore, it can be configured to operate the components of the apparatus described above. The data processing unit can be configured to operate automatically and / or to execute user commands. A computer program can be loaded into the working memory of a data processor. The data processor can thus be equipped to execute the method of the invention.

[0133] This exemplary embodiment of the invention covers both a computer program that uses the invention from the outset and a computer program that, by means of an update, transforms an existing program into a program that uses the invention.

[0134] Furthermore, the computer program element could be able to provide all the steps necessary to perform the procedure of an exemplary embodiment of the method as described above.

[0135] According to a further exemplary embodiment of the present invention, a computer-readable medium such as a CD-ROM is presented, wherein the computer-readable medium comprises a computer program element stored thereon, the computer program element being described by the preceding section.

[0136] A computer program can be stored and / or distributed on a suitable medium (especially, but not necessarily, a non-transient medium), such as an optical storage medium or a solid-state medium supplied with or as part of other hardware, but can also be distributed in other forms, such as via the Internet or other wired or wireless telecommunications systems.

[0137] However, the computer program can also be presented via a network, such as the World Wide Web, and can be downloaded from such a network into the working memory of a data processor. According to a further exemplary embodiment of the present invention, a medium for making a computer program element available for download is provided, wherein the computer program element is arranged to carry out a method according to one of the previously described embodiments of the invention.

[0138] It should be noted that embodiments of the invention are described with reference to different subject matter. In particular, some embodiments are described with reference to method type claims, whereas other embodiments are described with reference to device type claims. However, the person skilled in the art will understand from the above and the following description that, unless otherwise noted, in addition to any combination of features belonging to one type of subject matter, any combination of features relating to different subject matter is also considered to be disclosed in this application. However, all features can be combined, thereby providing synergistic effects that are greater than the simple summation of the features.

[0139] While the invention has been illustrated and described in detail in the drawings and the preceding description, such illustration and description are to be regarded as illustrative or exemplary and not as limiting. The invention is not limited to the disclosed embodiments. Other variations of the disclosed embodiments can be understood and brought about by a person skilled in the art when implementing a claimed invention in practice by studying the drawings, the disclosure, and the dependent claims.

[0140] In the claims, the word "comprise" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single processor or other unit can perform the functions of several elements mentioned in the claims. The mere fact that certain measures are cited again in different dependent claims does not indicate that a combination of these measures cannot be used advantageously. Any reference numerals in the claims are not to be interpreted as limiting the scope of protection. Such reference numerals may consist of numbers, letters, or any alphanumeric combination. QUOTES INCLUDED IN THE DESCRIPTION

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

[0000] P. Koken and M. Grass in Phys. Med. Biol. 51 (2006), pp. 3433 to 3448

[0093]

Claims

[1] System (SYS) for processing projection data for X-ray tomographic imaging, comprising: an input interface (IN) for receiving projection images taken by means of an X-ray tomographic imaging device (IA) along different projection directions of an anatomical feature (AF) of a moving object as it passes through several states of motion; an in-image detection component (in-image DC) configured to detect a motion state of the anatomical feature (AF) with respect to at least one reference path based on image information in the projection images; a gating component (GC) configured to assign a subset of the projection images to a motion state bin based on the detected motion state, and an output interface (OUT) to provide the assigned subset of projection images for reconstruction. [2] System according to claim 1, wherein the path is selected based on an analysis of an overview image of the anatomical feature or on an analysis of the projection images. [3] System according to one of the preceding claims, wherein a plane of the at least one such reference path is perpendicular to a projection direction in which at least some of the projection images are taken. [4] System according to one of the preceding claims, wherein the anatomical feature is completely and at all times within the field of view of the tomographic imaging device during such a movement. [5] System according to any of the preceding claims, wherein the system includes a tomographic reconstructor (RECON) to reconstruct, for at least one state of motion and based on projection images associated with that bin and / or one or more bin(s) of similar states, a reconstructed cross-sectional image volume V that is representative of the object in the state of motion. [6] System according to any of the preceding claims, wherein the system includes a display device (DD) for displaying the reconstructed image and / or the state of motion. [7] System according to any of the preceding claims, wherein the tomographic imaging device (IA) is configured for axial or helical imaging. [8] System according to any of the preceding claims, wherein the anatomical feature includes at least one section of one or more of a diaphragm, a chest wall and an abdominal wall of a patient. [9] System according to any of the preceding claims, wherein the anatomical feature is the anatomy of a patient and wherein such movement is caused by: respiratory activity and cardiac activity. [10] System according to any of the preceding claims, including i) a path definer (PD) configured to calculate such a reference path based on an overview image or based on the or other projection images, or ii) a user interface (UI) for providing such a reference path. [11] System according to any of the preceding claims, wherein the reference path is a line. [12] Imaging arrangement (MIA) comprising the system according to any of the preceding claims and the imaging device (IA). [13] Methods for processing projection data for X-ray tomographic imaging, comprising: Receiving (S620) projection images taken by means of an X-ray tomographic imaging device (IA) along different projection directions of an anatomical feature (AF) of a moving object as it goes through several states of motion; In-image recognition (S640), with respect to at least one reference path, of a motion state of the anatomical feature (AF) based on image information in the projection images; Assigning (S650) a subset of the projection images to a motion state bin based on the detected motion state and Providing (S660) the subset of projection images thus assigned for reconstruction. [14] Computer program element which, when executed by at least one processing unit, is adapted to cause the processing unit to perform the method according to claim 13. [15] At least one computer-readable medium on which the program element according to claim 14 is stored.

Images