Breathing motion compensation for preserving sharpness
By identifying and defining rigid regions in nuclear imaging and combining this with deformation vector field calculations, the problem of spinal blurring caused by respiratory motion in nuclear imaging was solved, improving image quality and signal-to-noise ratio while maintaining the clarity of rigid regions.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- KONINKLIJKE PHILIPS NV
- Filing Date
- 2021-08-01
- Publication Date
- 2026-05-22
AI Technical Summary
Existing nuclear imaging techniques can easily lead to blurring of the spinal region when performing respiratory motion compensation, affecting image quality. This is especially true in PET and SPECT imaging, where respiratory motion blur caused by long-term scanning makes it difficult to accurately locate pathological processes.
By identifying and defining rigid regions of an object, such as the spine, rib cage, and pelvis, and combining deformation vector field calculations, kernel image reconstruction is performed to maintain the clarity of these regions while compensating for the movement of organs such as the lungs and liver.
It improves the signal-to-noise ratio of nuclear images, reduces blurring in rigid regions, and maintains image clarity and overall quality, especially the clarity of rigid regions such as the spine.
Smart Images
Figure CN116034395B_ABST
Abstract
Description
Technical Field
[0001] This invention generally relates to motion-compensated nuclear imaging. In particular, but not exclusively, this invention relates to reconstructing motion-compensated nuclear images. Background Technology
[0002] In nuclear imaging, emission computed tomography (ECT) imaging, such as positron emission tomography (PET) and single-photon emission tomography (SPECT), is specifically performed to visualize and quantify a patient's metabolic status. For example, PET or SPECT images can help locate pathological processes (e.g., tumor growth or inflammation) and areas of abnormal blood perfusion in organs. Therefore, emission computed tomography is also known as functional imaging. In contrast, applied imaging techniques that visualize the anatomical structure of a subject (e.g., MR, CT, or ultrasound imaging) are generally referred to as structural imaging.
[0003] In PET, a positron-emitting substance is administered to the patient. This substance, often referred to as a radiopharmaceutical or radiotracer, is selected such that it is absorbed by cells involved in the pathological process being examined. When the radiotracer emits positrons, these positrons annihilate electron-positron pairs with nearby electrons, producing a pair of annihilated photons. Each of these annihilated photons has an energy of 511 keV, and both photons travel in substantially opposite directions. These photons are recorded by the PET detector essentially simultaneously (a phenomenon known as coincidence). Based on this coincidence, the PET system reconstructs an activity distribution, or activity map, which shows the spatial distribution of the positron-electron annihilation rate within the patient; this is also referred to herein as a PET image. The activity distribution, or PET image, essentially corresponds to the spatial distribution of the radiotracer within the subject and can therefore be evaluated for diagnostic purposes.
[0004] In SPECT imaging, radioactive tracers that emit gamma photons are used. The radioactive tracer can be selected based on the specific anatomical features that need to be visible in the image. The energy of the detected gamma rays will depend on the material used, and this energy is typically in the range of 40 keV–140 keV. Moreover, for SPECT, the distribution of activity visible in SPECT images essentially corresponds to the spatial distribution of the radioactive tracer within the subject, thus allowing for evaluation for diagnostic purposes.
[0005] To achieve a sufficiently high signal-to-noise ratio (SNR), nuclear imaging typically requires scanning the patient for several minutes at each bed position. These scan times are already too long for the subject, as they cannot hold their breath during such a long scan, especially for patients with lung conditions affecting breathing. Since PET and SPECT scans are usually performed at multiple bed positions, this typically increases the total scan time by 10-20 minutes, which only exacerbates the problem. Therefore, allowing the patient to breathe freely during imaging is crucial. Optionally, breathing can be supported via feedback mechanisms.
[0006] However, respiratory motion during the acquisition of emission computed tomography (ET) images can lead to significant blurring and errors in image-based assessment of metabolic parameters and precise localization of pathological processes. Therefore, respiratory motion compensation is necessary. This motion compensation is typically performed on gated ET imaging. Here, the acquired ET data is divided into two or more motion states (often referred to as gates or bins), where each gate contains ET data acquired during a phase of respiratory motion. To further evaluate the ET images, a specific motion phase can be defined or selected as a reference gate, and other ET images can be mapped onto this ET image using image registration to align them with the reference gate.
[0007] Current motion compensation techniques focus on aligning organs such as the lungs, liver, and kidneys, which follow the movement of the diaphragm and primarily move in a head-to-tail direction. After alignment, the surface area ratio (SNR) typically increases, and the visibility of organs or lesions is improved or even achieved. However, for current registration methods, motion compensation for organs also leads to blurring of adjacent rigid, previously aligned areas within the anatomical structure. This blurring of the image structure degrades image quality. Summary of the Invention
[0008] The present invention seeks to provide a method for reconstructing nuclear images, wherein motion of organs such as the lungs and liver, as well as other soft tissues, is compensated for while maintaining the clarity of rigid regions such as the spine.
[0009] One insight of this invention is that currently known organ motion compensation methods also lead to blurring of the spinal region because, for the entire image volume, a uniform elasticity at the underlying layer is assumed, and motion can occur in three directions. In contrast, human and animal physiology has rigid regions, such as the spine, rib cages, and pelvis. The pleura and peritoneum allow the lungs and abdominal organs to slide against these skeletal structures. Therefore, this invention further seeks to provide a method for compensating for motion in nuclear image reconstruction that better reflects the anatomical characteristics of the object.
[0010] To this end, a method and system for reconstructing motion-compensated nuclear images of an object are provided, along with an arrangement for acquiring nuclear images of the object and a computer program for performing the method.
[0011] The method for reconstructing a motion-compensated kernel image of an object includes: receiving kernel image data for multiple motion states of the object; reconstructing a kernel image from the kernel image data for each motion state; and calculating a deformation vector field for each motion state to map the reconstructed kernel image of that motion state onto a reference motion state. Calculating the deformation vector field herein includes the following steps: providing an initial deformation vector field for each motion state; defining at least one rigid region of the object; incorporating the defined rigid region into the initial deformation vector field; and calculating the deformation vector field having the incorporated rigid region. Examples of these rigid regions are at least portions of the spine and rib cage. Another example of a rigid region is the pelvis. The method further includes: mapping the reconstructed kernel image of each motion state onto the reference motion state using the deformation vector field; and combining the mapped kernel images of the multiple motion states into a motion-compensated kernel image. The method is preferably implemented by a computer or by other suitable computing modules.
[0012] Preferably, the method further includes receiving structural image data of the object, and the at least one rigid region is defined using the structural image data. In a particular embodiment, the structural image data is segmented to define the rigid region. In this embodiment, the rigid region is defined as a region of interest, preferably by segmenting the structural image data into a binary mask or a region of interest contour.
[0013] According to one aspect of the method, the defined rigid region is incorporated into the deformation vector field by directly inserting the defined rigid region into the defined rigid region.
[0014] According to another aspect of the method, the defined rigid region is indirectly incorporated into the deformation vector field by inserting the defined rigid region into the elastic matrix.
[0015] In an embodiment, the method further includes adjusting the transition region between the boundary of the rigid region and the adjacent region. Preferably, adjusting the boundary region includes one or more of the following operations: applying a smoothing filter to the transition region of the deformable vector field; assigning a higher elasticity to the transition region of the elasticity matrix than the elasticity of the adjacent region.
[0016] According to an alternative aspect of the method, the defined rigid region is incorporated by constraining the step of calculating the deformation vector field such that voxels directly adjacent to the boundary of the rigid region and the neighboring region are only allowed to be displaced parallel to the boundary.
[0017] In another alternative aspect of the method, defining the at least one rigid region includes analyzing the deformation vector field, and the defined rigid region is incorporated into the analyzed vector field. Preferably, the steps of defining the at least one rigid region, incorporating the defined rigid region into the analyzed vector field, and calculating the deformation vector field are performed iteratively until a stopping criterion is met.
[0018] The system for reconstructing motion-compensated kernel images of an object includes a kernel image reconstruction unit, which includes an input unit for receiving kernel image data for multiple motion states of the object. The reconstruction unit is configured to reconstruct a kernel image from the kernel image data for each motion state. The system also includes a deformation vector field calculator, which is configured to calculate a deformation vector field for each motion state to map the reconstructed kernel image of that motion state onto a reference motion state. The deformation vector field calculator includes a rigid region detector, which includes an input unit for receiving structural image data. The rigid region detector is configured to use the structural image data to define one or more rigid regions of the object. The deformation vector field calculator also includes a deformation vector field processor, which is configured to: provide an initial deformation vector field for each motion state, incorporate the defined rigid regions into the initial deformation vector field, and calculate an updated deformation vector field with the incorporated rigid regions. The system further includes a kernel image assembly unit, which is configured to map the reconstructed kernel image of each motion state onto the reference motion state using the updated deformation vector field. The kernel image assembly unit is further configured to combine the mapped kernel images of the plurality of motion states into a motion-compensated kernel image. Preferably, the system further includes a display for displaying the motion-compensated kernel image.
[0019] An arrangement apparatus for acquiring nuclear images of an object includes a nuclear imaging device for acquiring nuclear image data of the object and the aforementioned system for reconstructing the nuclear image of the object. In an advantageous embodiment, the arrangement apparatus further includes a structural imaging device for acquiring structural image data of the object.
[0020] The computer program product includes instructions that, when the computer program is run, cause the processor to perform the methods described above.
[0021] Another advantage is that the blurring of rigid regions in the kernel image of the object is at least reduced, and in the best case, completely removed. By identifying rigid regions and including them in the deformation vector field calculation, the resulting deformation vector field does not artificially shift these regions when mapping different motion states to a reference state. This improves the sharpness of these regions in the resulting image.
[0022] Another advantage is the improved overall image quality of nuclear images. The signal-to-noise ratio improvement achieved by compensating for motion or movement of organs (e.g., lungs, liver, and other soft tissues) is maintained, while the sharpness of rigid regions (e.g., the spine) is also preserved. The resulting image has an improved signal-to-noise ratio and reduced blur. Attached Figure Description
[0023] In the following figures:
[0024] Figure 1 An arrangement apparatus for acquiring a nuclear image of an object is schematically and exemplaryly illustrated, the arrangement apparatus including a system for reconstructing a motion-compensated nuclear image of the object.
[0025] Figure 2 An example of a method for reconstructing a kernel image of an object with motion compensation is illustrated schematically.
[0026] Figure 3 Another example of a method for reconstructing a kernel image of an object is illustrated schematically.
[0027] Figure 4a and Figure 4b An example of incorporating a rigid region into an initial deformation vector field is illustrated schematically.
[0028] Figure 5 An example of a method for reconstructing a nuclear image is illustrated schematically. Detailed Implementation
[0029] In the following examples, nuclear imaging or functional imaging is described with reference to exemplary applications of PET imaging. However, it should be understood that functional imaging techniques are not limited to these examples, and alternative emission computed tomography (ECT) imaging techniques such as SPECT imaging may also be used. Additionally, in the following examples, structural imaging performed to obtain information about the anatomical structure of a subject is described with reference to exemplary applications of CT imaging. However, it should also be understood that structural imaging techniques are not limited to these examples, and alternatives such as MR imaging or ultrasound imaging may also be used.
[0030] Figure 1A system 120 for reconstructing a nuclear image of an object with motion compensation is illustrated. In this example, system 120 is shown as part of an arrangement device 100 for acquiring a nuclear image of an object.
[0031] The arrangement device 100 includes an imaging device 110 for acquiring image data of the object. Figure 1 The diagram illustrates a combined PET / CT scanning device 111. Nuclear image data 112 is collected by the PET section of the device, while structural image data 113 is collected by the CT section. In this example, separate structural imaging data is acquired. Alternatively, structural data can be derived from the nuclear image data. System 120 receives and reconstructs the nuclear image data 112 and structural image data 113 to provide motion-compensated nuclear images of the object. Preferably, the arrangement device 100 for nuclear image acquisition also includes a display 130. This can be a computer display, which can be part of or a separate display of the arrangement device. The advantage of having such a display is that the operator or clinician can access and view the acquired and reconstructed images. Additionally or alternatively, the reconstructed images can also be stored in a database or archiving system for later access and viewing.
[0032] The system 120 for reconstructing motion-compensated kernel images of objects includes a kernel image reconstruction unit 121, a deformation vector field (DVF) calculator 122, and a kernel image assembly unit 129.
[0033] The nuclear image reconstruction unit 121 in this example is a unit for PET image reconstruction. This unit has an input section for receiving nuclear image data 112 in the form of PET data for multiple motion states of an object. The PET data can be any format suitable for use by the reconstruction unit 121, such as list-pattern data or sine wave data. The nuclear image data is categorized or binned into multiple motion states of the object. States can be defined based on the patient's respiratory cycle. Alternatively, when the heart is the organ of interest, states can be defined based on the patient's cardiac cycle. The number of states can be two (e.g., the object's maximum inhalation or maximum exhalation), but preferably more.
[0034] Reconstruction unit 121 is configured to reconstruct a kernel image from kernel image data 112 for each motion state. For this purpose, the unit may include a dedicated reconstruction algorithm. By additionally using structural image data 113 for calculating attenuation and scattering correction as part of the image reconstruction for each motion state, the quality of the reconstructed image can be improved.
[0035] The DVF calculator 122 is configured to calculate the DVF for each motion state to map the reconstructed kernel image of that motion state onto a reference motion state. One of the motion states in the kernel image data can be selected as the reference motion state. Alternatively, when acquiring structural data 113 individually, the reference motion state preferably corresponds to the motion state at the time of acquiring the structural image data 113. In practice, structural image data can often be acquired quickly in a single motion state. This is particularly suitable for CT image acquisition. Another option is to define a separate reference state that does not correspond to the motion state acquired in the image data acquisition. For example, the reference state can be defined such that the images all deform toward the center of the motion pattern, and it can be said that these images meet in the middle.
[0036] To calculate each DVF, the DVF calculator 122 includes a rigid region detector 123 and a DVF processor 124.
[0037] The rigid region detector 123 includes an input unit for receiving structural image data and is configured to use the structural image data to define one or more rigid regions of an object. Rigid regions can be detected, for example, by using a segmentation device. Such a device can automatically or semi-automatically identify rigid anatomical regions and / or can have a user interface to allow an operator to manually delineate the regions. The output of such a segmentation device can be a binary mask of the rigid region or a region of interest contour.
[0038] The DVF processor 124 is configured to: provide an initial DVF 125 for each motion state, incorporate a defined rigid region 126 into each initial DVF, and compute an updated DVF 127 with the incorporated rigid region. The reconstructed kernel image for each motion state and the corresponding DVF mapping it to the reference state are inputs to the kernel image assembly unit 129.
[0039] The kernel image assembly unit 129 is configured to map the reconstructed kernel image of each motion state onto a reference motion state using an updated deformation vector field. The assembly unit 129 is also configured to combine the mapped kernel images of multiple motion states into a motion-compensated kernel image. For example, this combination can be accomplished by adding the individual mapped images to create a summed image as the motion-compensated kernel image. Alternatively, the images can be combined by calculating the mean of the mapped images to create an average image as the motion-compensated kernel image.
[0040] Figure 2 The steps of a method 200 for reconstructing a kernel image for motion compensation of an object are schematically illustrated.
[0041] The method for reconstructing the image begins by receiving kernel image data 210 of the object. The image data is segmented for multiple motion states of the object, meaning it is divided into groups (often referred to as "gates" or "bins"). Data for each group is acquired as the object is in its corresponding motion state. This data is also known as gated kernel image data. Next, the kernel image data is reconstructed into a kernel image 220 for each motion state. Figure 2 In the example, the method includes an additional, optional, separate step of receiving structural image data 230. The structural data is used to determine an attenuation map for attenuation and scattering correction during nuclear image reconstruction 220.
[0042] Next, DVF 240 is calculated for each motion state. DVF is used to map the reconstructed kernel image of the corresponding motion state onto a reference motion state. As explained above regarding the system, the reference state can be, for example, a state selected from gated kernel image data, a separately defined state, or a motion state when acquiring structural image data.
[0043] Calculating the DVF 240 for each motion state also involves the steps of providing an initial deformation vector field 241 for each motion state, and defining at least one rigid region 242 in the object. The defined rigid region is then incorporated into the initial deformation vector field 243, and a deformation vector field 244 with the incorporated rigid region is calculated.
[0044] Then, the final reconstructed kernel image 250 is assembled by using a deformation vector field to map the reconstructed kernel image of each motion state onto a reference motion state, and then combining the mapped kernel images of multiple motion states into a motion-compensated kernel image.
[0045] Figure 3 Another example of a method 300 for reconstructing a kernel image of an object is illustrated schematically. Similar to... Figure 2 In one embodiment, the method includes: receiving nuclear image data 310 for a plurality of motion states, and reconstructing a nuclear image 320 from the nuclear image data for each motion state. In this example, the nuclear image data is PET data, and a PET image is reconstructed for each motion state. In this example, additional structural image data 330 in the form of CT data is received. The CT data is used in the PET image reconstruction to calculate an attenuation map for correcting scattering and attenuation. A deformation vector field 340 is calculated for each motion state to map the reconstructed image of that state onto a reference state.
[0046] In this example, the motion state of the CT image is used as the reference motion state. Rigid regions 342 are detected by segmenting the CT image data. The result of the segmentation is a binary mask indicating the location of one or more rigid regions. Such a binary mask can be the direct output of the segmentation algorithm. Alternatively, the output of the segmentation algorithm can be the region of interest contours of one or more rigid regions. In this case, detecting rigid regions involves the additional step of transforming the contours into a binary mask. The binary mask indicating the location of the rigid regions is provided as input to define the rigid regions in the DVF.
[0047] When calculating the DVF for each motion state, an initial DVF is provided. In the method illustrated here, the initial DVF 341 is provided by calculating the DVF for each motion state in a known manner, where all tissues are considered to have the same elasticity, without considering any rigid regions. To define one or more rigid regions, a binary mask indicating the location of the rigid regions is directly inserted into the initial DVF 343, and the boundaries of the rigid regions and the areas between them and their neighboring regions are defined and adjusted to calculate the DVF 344. This will refer to... Figure 4a Let's explain further.
[0048] Once the DVF has been calculated for each motion state, motion-compensated kernel images 350 are assembled. The calculated DVF is used to map the reconstructed PET images 320 for each motion state to the motion state 351 of the CT image data. The mapped PET images for multiple motion states are then combined into a motion-compensated PET image 352 by summing the mapped images. In this example, the summed images are then further normalized to clinically standardized uptake value units 353 to form the final motion-compensated PET image.
[0049] Figure 4a and Figure 4b An example of incorporating a rigid region into the initial deformation vector field is illustrated schematically. Figure 4a An example of incorporating a rigid region by directly inserting it into a vector field is shown, and Figure 4b An example is shown where a rigid region is indirectly incorporated into an elastic matrix.
[0050] Figure 4aA two-dimensional representation of a DVF 410 with an inserted rigid region 430 is shown. In use, such a DVF can also be three-dimensional. The dimension will correspond to the image being reconstructed. In this example, the initial DVF computed is a full vector field with vectors indicating the displacement for each voxel to map it to a reference motion state. In this example, the rigid region is detected and defined in the form of a binary mask. The binary mask can be two-dimensional or three-dimensional, corresponding to the dimension of the DVF. The binary mask is inserted into the DVF by setting the vectors of the voxels of the rigid region 430 to zero. Additionally, a transition region 440 between the boundary of the rigid region and the adjacent soft tissue region is defined and adjusted. The thickness of this region can be one or more voxels. In this example, the transition region is adjusted by applying a smoothing filter to the vector of the transition region. Applying such a filter has the advantage that the vector is adjusted in such a way that no tissue gaps will open between the rigid region and the soft tissue after mapping, and that mapping soft tissue onto the rigid region is prevented. The remaining soft tissue displacement vector 420 was retained as originally calculated.
[0051] Figure 4b A two-dimensional mesh representation of a voxel elasticity matrix 460 with inserted rigid regions according to the present invention is shown. The elasticity matrix is a common component in deformation vector field (DVF) calculations. This matrix specifies a value representing the elasticity of the tissue at each voxel location in the image. In known methods, the elasticity of all voxels is uniformly set to a standard soft tissue value. In practice, the elasticity matrix can also be three-dimensional. The dimension of the elasticity matrix will correspond to the dimension of the calculated DVF. In the present invention, the initially calculated DVF can be a zero vector field corresponding to a identical copy, or a DVF initially calculated using an elasticity matrix that assumes all voxels correspond to soft tissue with the same standard soft tissue elasticity value S. Figure 4b In this example of the invention shown, rigid regions are detected and defined in the form of a binary mask. The binary mask can be two-dimensional or three-dimensional, corresponding to the dimensions of the elasticity matrix. The binary mask is inserted into the elasticity matrix 480 by setting the elasticity of the voxels of the rigid region to zero. The elasticity of the other soft tissue voxels is left as S 470.
[0052] exist Figure 4b In the example, the transition region between the boundary of the rigid region and the adjacent region was also adjusted. The transition region shown in this figure includes a single layer of transition voxels 490, but it could also be two or more layers of transition voxels. This region is adjusted by assigning an elasticity value E higher than the standard soft tissue elasticity S to these voxels 490. The advantage of using this higher elasticity is that the movement allowed by the voxels more closely resembles that of an anatomical sliding surface.
[0053] In an alternative approach, the defined rigid region is incorporated into the final DVF by constraining the deformation vector field calculation steps such that voxels directly adjacent to the rigid region and the boundary of the adjacent region are only allowed to shift parallel to the boundary. This can be achieved, for example, by allowing only these voxels to shift such that the displacement vector components in the normal direction of the boundary are continuous across the boundary. In this approach, the rigid region is supplied to the DVF calculation algorithm, for example, as additional input in the form of a binary mask. Voxels corresponding to the rigid region are fixed in place. Voxels at the boundary between the soft tissue and the rigid region are decoupled from the rigid region and are additionally allowed to move only in directions parallel to the boundary surface. Other voxels corresponding to other soft tissues are allowed to shift as usual. In this way, the anatomical sliding surface between the rigid region and the adjacent soft tissue is modeled more accurately.
[0054] Figure 5 Another example of a method 500 for reconstructing a nuclear image is schematically illustrated. In this example, only nuclear image data is used to reconstruct a motion-compensated nuclear image of an object, and a deformation vector field is computed iteratively for each motion state. Similar to the previous example, the method steps include receiving nuclear image data (e.g., PET or SPECT data) for multiple motion states 510, reconstructing the nuclear image data into a nuclear image for each motion state 520, and computed a deformation vector field 540 for each motion state to map the reconstructed image of that state to a reference state.
[0055] In this example, the reference motion state is a separate, predefined motion state. This has the advantage of allowing a fully automated reconstruction method, which is consistent and independent of object-specific characteristics.
[0056] The calculation of the DVF begins with providing an initial DVF 541. This can be a zero field representing no deformation as a starting point, or a reference DVF retrieved from a database. Next, rigid regions are detected by analyzing the properties of the DVF 542. This is preferably done using vector operators. For example, a shear or strain map of the DVF can be generated. Local regions with high shear are indicators of transitions between regions with different types of tissue. The shear values of the DVF thus provide an estimate of the possible location of the rigid region. A rigid region may exist in a nuclear image (e.g., the spine adjacent to the lungs), but additional regions may also exist as parts of the image (e.g., the rib cage and / or parts of the pelvis).
[0057] The method also includes a check step to determine whether a region is indeed rigid. For example, a standard model of lung behavior can be used as a reference or organ atlas. If earlier structural images of the object are available, these images can also be used.
[0058] The detected rigid regions are then incorporated into the analyzed DVF by adjusting the elasticity matrix 543. This is achieved by allowing spatial variation of the elasticity matrix, specifically by increasing the elasticity at locations where high local shear motion is detected. This locally increased elasticity, in turn, allows for a further increase in the shear component. This has the advantage that the DVF motion increasingly approximates the sliding surface according to the object's anatomy. The updated DVF 544 is then calculated using the adjusted elasticity matrix.
[0059] The steps 542 (analyzing the current DVF), 543 (adjusting the elasticity matrix), and 544 (calculating the updated DVF) are repeated until the stopping criterion 545 is met. Various stopping criteria can be anticipated in this context. For example, iteration can stop when the location of the estimated rigid region no longer changes. Alternatively, iteration can stop when the change in the updated DVF compared to the previous DVF is less than a predetermined threshold. As an alternative example, iteration can also stop when the local high shear value in the DVF reaches a predetermined maximum value. An additional or alternative option is a predefined maximum number of iterations.
[0060] Once the DVF has been calculated for each motion state, the motion-compensated images are assembled 550. The calculated DVF is used to map the reconstructed kernel image for each motion state onto a predefined reference state 551. Then, the mapped kernel images for multiple motion states are combined into a motion-compensated kernel image by calculating the mean of the mapped images 552.
[0061] Any method step disclosed herein can be recorded in the form of a computer program comprising instructions that, when executed on a processor, cause the processor to perform such method steps. The instructions can be stored on a computer program product. The computer program product can be provided by dedicated hardware and hardware capable of running software associated with suitable software. When provided by a processor, functionality can be provided by a single dedicated processor, a single shared processor, or multiple individual processors (some of which can be shared). Furthermore, embodiments of the invention can take the form of a computer program product accessible from a computer-usable or computer-readable storage medium that provides program code for use by or in conjunction with a computer or any instruction execution system. For the purposes of this description, the computer-usable or computer-readable storage medium can be any device that can include, store, communicate, propagate, or transmit programs for use by or in conjunction with an instruction execution system, device, or apparatus. The medium can be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system or device or a propagation medium. Examples of computer-readable media include semiconductor or solid-state memory, magnetic tape, removable computer disks, random access memory (RAM), read-only memory (ROM), hard disks, and optical discs. Current examples of optical discs include compressed discs-read-only memory (CD-ROM), compressed discs-read / write (CD-R / W), Blu-ray, and DVD. Examples of distribution media are the Internet or other wired or wireless telecommunications systems.
[0062] By studying the accompanying drawings, disclosure, and claims, those skilled in the art will be able to understand and practice variations of the disclosed embodiments in practicing the claimed invention. Note that various embodiments can be combined to achieve additional advantageous effects.
[0063] In the claims, the word "comprising" does not exclude other elements or steps, and the words "a" or "an" do not exclude a plurality.
[0064] A single unit or device can perform the functions of several items recited in the claims. The fact that certain measures are recited in different dependent claims does not indicate that combinations of these measures cannot be used advantageously.
[0065] Any reference numerals in the claims should not be construed as limiting the scope.
Claims
1. A method (200, 300, 500) for reconstructing a motion-compensated kernel image of an object, the method comprising: Receive kernel image data (210, 310, 510) for multiple motion states of the object. Receive anatomical structure image data (230, 330) of the object; For each motion state, the kernel image data is reconstructed into a kernel image (220, 320, 520). For each motion state, a deformation vector field (240, 340, 540) is calculated to map the reconstructed kernel image of that motion state onto a reference motion state. Calculating the deformation vector field includes the following steps: An initial deformation vector field (241, 341, 541) is provided for each motion state. Segmenting the anatomical image data to define rigid regions; and The defined rigid region is incorporated into the initial deformation vector field (243, 343, 543) by setting the vector of the voxel of the rigid region in the initial deformation vector field to zero, so that the deformation vector field (244, 344, 544) with the incorporated rigid region is calculated. The method further includes: The reconstructed kernel image for each motion state is mapped onto the reference motion state (351, 551) using the deformation vector field; and The mapped kernel images of the multiple motion states are combined into motion-compensated kernel images (352, 552).
2. The method of claim 1, further comprising segmenting the anatomical image data to define the rigid region as a region of interest.
3. The method of claim 2 further includes segmenting the anatomical structure image data by dividing the anatomical structure image data into binary masks or region of interest contours to define the rigid region as a region of interest.
4. The method according to any one of claims 1-3, wherein, The defined rigid region is indirectly incorporated into the deformation vector field by inserting the defined rigid region into the elastic matrix (460).
5. The method according to any one of claims 1-3, further comprising adjusting the transition region (440, 490) between the boundary of the rigid region and the adjacent region.
6. The method according to claim 4, wherein, Adjusting the boundary region includes assigning a higher elasticity to the transformation region of the elasticity matrix than to the neighboring regions.
7. The method according to any one of claims 1-3, wherein, The defined rigid region is incorporated by constraining the calculation of the deformation vector field such that voxels directly adjacent to the boundary of the rigid region and the neighboring region are only allowed to be displaced parallel to the boundary.
8. A system (120) for reconstructing a motion-compensated kernel image of an object, wherein, The system includes: A nuclear image reconstruction unit (121) includes an input unit for receiving nuclear image data for multiple motion states of the object, and the reconstruction unit is configured to reconstruct a nuclear image from the nuclear image data for each motion state. A deformation vector field calculator (122) configured to calculate a deformation vector field for each motion state for mapping the reconstructed kernel image of that motion state onto a reference motion state, the deformation vector field calculator comprising: A rigid region detector (123) includes an input unit for receiving anatomical image data, the rigid region detector being configured to segment the anatomical image data to define one or more rigid regions of the object; A deformation vector field processor (124) is configured to provide an initial deformation vector field (125) for each motion state, and to incorporate the defined rigid region into the initial deformation vector field (126) by setting the vector of the voxel of the rigid region in the initial deformation vector field to zero to compute an updated deformation vector field (127) with the incorporated rigid region. The system also includes: A kernel image assembly unit (129) is configured to map the reconstructed kernel image of each motion state onto the reference motion state using the updated deformation vector field, and is also configured to combine the mapped kernel images of the plurality of motion states into a motion-compensated kernel image.
9. The system of claim 8 further includes a display (130) for displaying the motion-compensated nuclear image.
10. An arrangement apparatus (100) for acquiring nuclear images of an object, comprising: Nuclear imaging device (111) for acquiring nuclear image data (112) of the object. The system (120) according to any one of claims 8-9 is used to reconstruct the nuclear image of the object.
11. The arrangement device according to claim 10 further includes a structural imaging device (111) for acquiring anatomical image data (113) of the object.
12. A computer program product comprising instructions which, when the computer program is run, cause a processor to perform the method according to any one of claims 1-7.