Apparatus, system, method and computer program for providing a nuclear image of a region of interest of a patient

Abstract

The invention relates to an apparatus allowing to improve the image quality of a nuclear image, such as a PET image. The apparatus (110) comprises a providing unit (111) for providing nuclear image data of a region of interest, a providing unit (112) for providing a motion signal indicative of a motion of the region of interest, a determining unit (113) for determining different motion states of the region of interest based on the motion signal, a determining unit (114) for determining for each motion state nuclear image data corresponding to the motion state, a reconstructing unit (115) for reconstructing for each motion state an absorption map based on the corresponding nuclear image data of the respective motion state, and a reconstructing unit (116) for reconstructing one or more nuclear images of the region of interest based on the nuclear image data and the absorption maps reconstructed for each motion state.

Description

Technical Field

[0001] The present invention relates to an apparatus, system, method, and computer program for providing nuclear images of regions of interest to a patient. Background Technology

[0002] For example, nuclear image data acquisition using a PET imaging system is generally very time-consuming; that is, image data needs to be acquired over an extended period of time to detect the necessary number of events. Therefore, during this extended period, patient movement, such as respiratory movements or other involuntary movements, cannot be completely avoided. Reconstructing nuclear images based on nuclear image data already acquired during the patient's movement results in low-quality nuclear images. A common method for correcting nuclear image data for respiratory movement is to provide 4D CT image data in addition to the nuclear image data. 4D CT image data involves CT image data acquired during at least one respiratory cycle of the patient. Because CT image data can be acquired very quickly, 4D CT image data maps the patient's respiratory movement very accurately. In this method, the nuclear image data is then classified according to different respiratory states, for example, using a respiratory sensor, and registered with the 4D CT image data. This allows for correction of attenuation in the nuclear image data and also correction of motion within the nuclear image data. However, because 4D CT image data is acquired over a very short time period, while nuclear imaging data is acquired over a longer time period, irregularities in patient movement, such as in respiratory patterns or other irregular unconscious movements, will not be represented by the 4D CT image data, thus still leading to artifacts and inaccuracies in the reconstruction of nuclear images. One possible alternative to avoid the effects of movement (e.g., respiratory movement) in nuclear imaging is to use an external respiratory sensor and gate the acquisition of nuclear imaging data based on the signal from the external respiratory sensor; that is, to acquire nuclear imaging data only during one respiratory state within the patient's respiratory cycle. However, this type of acquisition results in increased acquisition time until the nuclear imaging system has detected the required number of events. In many cases, increased acquisition time is unacceptable due to patient discomfort during nuclear imaging data acquisition. Summary of the Invention

[0003] The object of this invention is to provide an apparatus, system, method, and computer program that allows for improvement of image quality of nuclear images without increasing patient discomfort.

[0004] In a first aspect of the invention, an apparatus is provided for providing nuclear images of a region of interest (ROI) of a patient, wherein the apparatus comprises: a) a nuclear image data providing unit for providing nuclear image data of the ROI of a patient acquired using a nuclear imaging device; b) a motion signal providing unit for providing motion signals indicating motion of the ROI of the patient during the acquisition of the nuclear image data; c) a motion state determining unit for determining different motion states of the ROI based on the motion signals, wherein each of the different motion states indicates a different state of the ROI; d) a corresponding image data determining unit for determining nuclear image data corresponding to each motion state, wherein if nuclear image data has been acquired during the state of the ROI corresponding to the motion state, the nuclear image data corresponds to the motion state; e) an absorption map reconstruction unit for reconstructing an absorption map for each motion state based on the corresponding nuclear image data of the corresponding motion state, wherein the absorption map indicates the absorption of nuclear radiation in the ROI; and f) a nuclear image reconstruction unit for reconstructing one or more nuclear images of the ROI based on the nuclear image data and the absorption map reconstructed for each motion state.

[0005] Since the corresponding image data determination unit determines nuclear image data corresponding to each motion state, and the absorption map reconstruction unit reconstructs an absorption map for each motion state based on the corresponding nuclear image data for that motion state, the reconstructed absorption map is determined based on the same nuclear image data that will later be used to reconstruct the nuclear image, and can be considered to correspond to the same time period in which nuclear images were acquired. Therefore, the absorption map very accurately reflects the patient's motion during the acquisition of the nuclear image data. Since the nuclear image reconstruction unit subsequently reconstructs one or more nuclear images of the region of interest based on the nuclear image data, and the absorption map very accurately reflects the patient's motion during the acquisition time period, motion in the nuclear image can be corrected very accurately. Furthermore, due to the high consistency between the nuclear image data and the absorption map reconstructed based on the nuclear image data, attenuation correction is also improved. Therefore, the image quality of the reconstructed nuclear image can be improved without extending the nuclear image data acquisition and without causing further discomfort to the patient.

[0006] The nuclear image data providing unit is adapted to provide nuclear image data of a patient's region of interest acquired using a nuclear imaging device. The nuclear image data providing unit can be a storage unit in which nuclear image data has been stored and can be retrieved. Furthermore, the nuclear image data providing unit can be a retrieval unit for retrieving nuclear image data from, for example, a nuclear imaging system used for acquiring nuclear image data, wherein the nuclear image data providing unit is subsequently adapted to provide the acquired nuclear image data.

[0007] The provided nuclear image data can encompass any type of nuclear image data acquired using a nuclear imaging system. This nuclear imaging system relates to an imaging system that uses a detector to detect radiation emitted from at least one region of interest within the patient's body. Specifically, the nuclear imaging system does not involve a system for acquiring radiation transmitted through the patient's body, for example, by providing a radiation source outside the patient's body, as is done in X-ray CT imaging. The nuclear imaging system used to acquire nuclear image data can, for example, involve PET imaging systems, SPECT imaging systems, etc. The nuclear image data can therefore encompass PET image data, SPECT image data, etc.

[0008] The motion signal providing unit is adapted to provide motion signals indicating the movement of a patient's region of interest during the acquisition of nuclear image data. The motion signal providing unit can be a storage unit for storing and retrieving motion signals. Furthermore, the motion signal providing unit can be a retrieval unit for retrieving motion signals, for example, from a motion sensor that has already acquired motion signals during the acquisition of nuclear image data. The motion signal can be any signal indicating movement in the region of interest during the acquisition of nuclear image data. For example, the motion signal can be a 1D signal recording the position or acceleration of a point in the region of interest over time. However, the motion signal can also be a 2D or 3D signal recording the position or acceleration of more than one location in the region of interest over time. For example, the motion signal can also involve motion graphs, such as vector graphics, indicating the movement of each portion of the region of interest (e.g., each pixel or voxel) over time.

[0009] In an embodiment, the motion signal providing unit is adapted to provide a signal from a sensor configured to detect motion in a patient's region of interest as a motion signal. For example, the motion signal can be acquired using a motion sensor attached to the patient's region of interest and measuring the motion in that region during the acquisition of nuclear image data. Alternatively, the motion signal providing unit can be adapted to extract a motion signal indicative of motion in the patient's region of interest from a monitoring camera monitoring the patient during the acquisition of nuclear image data. In this embodiment, known motion extraction and tracking methods for monitoring images can be used to extract the motion signal from the monitoring image.

[0010] In a preferred embodiment, the motion signal providing unit is adapted to determine the motion signal based on the nuclear imaging data. For example, the motion signal providing unit is adapted to determine the motion signal directly from the nuclear imaging data using time-of-flight information provided in the nuclear imaging data. A. Salomon et al.'s "Robust real-time extraction of respiratory signals from PET list-mode data" (Physics in Medicine & Biology, 2018, Vol. 63, No. 11) provides an outline of an exemplary method for determining motion based on nuclear imaging data. However, other methods can also be used to extract motion signals from the nuclear imaging data itself, i.e., signals indicating motion in the patient's region of interest. The advantage of extracting motion signals from the nuclear imaging data itself is that additional motion sensing units, such as cameras or dedicated motion sensors, are not necessarily required during nuclear imaging. Furthermore, nuclear imaging data that has been fully acquired without additional detection of motion signals, or nuclear imaging data where the provided motion signal has been corrupted, can thus be processed.

[0011] The motion signal can refer to a specific type of motion, such as a motion signal indicating only one type of motion, like respiratory motion, or a general motion signal indicating all motions of the patient in the region of interest, including irregular motions that are not part of the patient's overall circulatory motion, such as breathing or cardiac motion. In another preferred embodiment, the motion signal indicates the entire spatial motion pattern of the region of interest. This can be achieved, for example, if the motion signal providing unit is adapted to determine motion signals from the nuclear image data itself. In other embodiments, the motion signal can also indicate cardiac motion, bowel motion, or any other motion of the patient during the acquisition of the nuclear image data. In another preferred embodiment, the motion signal indicates regular body motion of the patient in the region of interest. Specifically, the motion signal can indicate circulatory motion of the region of interest. In a preferred embodiment, the motion signal indicates the patient's respiratory motion.

[0012] The motion state determination unit is adapted to determine different motion states of a region of interest (ROI) based on motion signals. Each of the different motion states indicates a different state of the ROI. For example, different states of the ROI relate to different locations or forms of anatomical structures within the ROI. For example, possible motion states of the ROI can be predetermined based on user input or prior knowledge of the ROI. The motion state determination unit is then adapted to determine, based on the motion signals, which of the possible motion states can be found during the acquisition of nuclear image data. Furthermore, the motion state determination unit is adapted to automatically determine motion states, for example, by searching for specific features in the motion signals (such as local maxima, local minima, or substantially constant time periods) and defining the state of the motion signal of the ROI during which these motion signal features have been acquired as motion states.

[0013] Furthermore, if at least two motion states have already been determined based on the motion signal, the motion state determination unit can also determine the motion state between these two determined motion states by interpolating between them, thereby acquiring a third motion state between the two motion states. For example, the interpolation can involve motion states defined between two motion states. In this case, for example, only a few motion states need to be determined directly from the motion signal, while other motion states can be determined based on the already determined motion states.

[0014] Preferably, different motion states involve different substantially stationary states of the region of interest. Specifically, the motion state determination unit is adapted to distinguish time periods of motion signals acquired during the acquisition of motion signals in which the region of interest is substantially stationary in the patient's region of interest. In this context, a substantially stationary state of the region of interest can be defined as a state in which the region of interest or a portion thereof exhibits motion below a predetermined threshold within a certain time period. For example, the threshold can be determined based on the region of interest and the accuracy that the imaging process should achieve. For example, if the region of interest involves an area of ​​interest in which little motion occurs and / or if the tumor expected to be imaged by nuclear imaging is very small, the threshold can be determined to be smaller than if it is known that much unconscious motion occurs in the region of interest and / or if the tumor expected to be imaged is large. Overall, the threshold can be selected such that motion in the region of interest in a substantially stationary state remains acceptable to the user and the planned application of the obtained nuclear images.

[0015] The time period during which the region of interest (ROI) needs to show movement below a predetermined threshold can be predetermined or adjustable during the determination of the motion state. For example, the motion state determination unit can be adapted to identify the motion state based on a minimum time period threshold. The minimum time period threshold refers to the shortest time period for which the motion state should be determined. The minimum time period threshold can be determined based on the accuracy of the provided motion signal or based on the accuracy that should be provided by the obtained kernel image. If the patient's ROI is substantially stationary during a time period longer than the minimum time period threshold, the motion state determination unit can be adapted to determine the patient's body state as a motion state. For example, if the motion signal indicates that the patient is moving rapidly during a certain time period, i.e., the ROI shows a high rate of change in a short time period, the motion state determination unit can be adapted to determine more motion states for this time period than during a time period when the motion signal indicates that the ROI is moving slowly (i.e., showing only small changes in a long time period).

[0016] If the motion signal indicates that the motion in the region of interest is regular, periodic, or cyclical, the motion state determination unit can be adapted to determine the same motion state for each cycle or loop of the periodic or cyclical motion. For example, if the motion signal indicates the patient's respiratory motion, the motion state determination unit can be adapted to determine the state of the region of interest involving the expiratory state as the first motion state and the state of the patient involving the inspiratory state as the second motion state. Both states can also be considered as states that are substantially stationary over a certain period of time.

[0017] In embodiments, the motion state determination unit can be adapted to determine the motion state based on user input. For example, the user can indicate in the representation of the motion signal which position the motion signal indicates the motion state. Furthermore, the motion state determination unit can also be adapted to determine the motion state based on stored motion states of previous, specifically similar cases. Additionally, the motion state determination unit can also be a motion state providing unit, used to provide the motion state based on stored motion states or based on user input.

[0018] The corresponding image data determination unit is adapted to determine kernel image data corresponding to each motion state. If kernel image data has been acquired during a state of the region of interest corresponding to a motion state, then the kernel image data corresponds to the motion state. For example, if the motion state is the patient's exhalation state, i.e., the patient's lungs are in a state of maximum contraction, then all kernel image data acquired during this lung state can be determined as the corresponding image data by the corresponding image determination unit. More generally, the image data determination unit is adapted to determine the time of a portion of the acquired kernel image data and to determine in which motion state, as determined by the motion state determination unit, the patient is at the time of said portion of the acquired kernel image data. This portion of kernel image data can then be considered to correspond to the corresponding motion state. Preferably, if the kernel image data can be represented as list pattern data, the corresponding image data determination unit is adapted to classify the list pattern data according to the determined motion state. For example, all portions of the list pattern data acquired during one or more time intervals of the motion state are provided together in the list pattern data and marked as belonging to the corresponding motion state.

[0019] In a preferred embodiment, the corresponding image data determining unit is adapted to determine one or more time intervals for each motion state based on motion signals, during which the region of interest is in a state corresponding to the motion state during kernel image data acquisition, and further based on whether kernel image data has been acquired during one or more time intervals for each corresponding motion state to determine the kernel image data corresponding to the motion state. In an embodiment, the corresponding image data determining unit can be adapted to determine the time interval for each motion state by providing a representation of the motion signal to a user, enabling the user to indicate the time interval for each motion state. Furthermore, the corresponding image data determining unit can also be a receiving unit, which receives the time interval for each motion state, for example, from a storage device or from a user. Additionally, the corresponding image data determining unit can be adapted to determine the time interval for each motion state based on known characteristics of the motion signal indicating the motion state. For example, for a specific motion state, if the known motion signal is above a certain threshold, the corresponding image data determining unit can determine all time intervals of the motion signal above the threshold as belonging to that motion state. Furthermore, if it is determined that the motion state is used only for a very short period of time or even only for a moment, the corresponding image data determining unit can be adapted to determine a predetermined shortest time interval near that moment as the time interval corresponding to the motion state. The shortest time interval can be predetermined based on the expected image accuracy, the anticipated motion of the region of interest, and the resolution of the motion signal. Furthermore, the corresponding image data determination unit can be adapted to automatically adjust the shortest time interval, for example, based on the rate of change of the region of interest or the velocity of a portion of the region of interest.

[0020] If a motion state has occurred at more than one time interval during nuclear imaging data acquisition, the time intervals for the motion state can be determined, for example, based on the motion signal. For instance, if different time intervals of the motion signal show the same motion pattern involving a specific motion state, all of these time intervals can be identified as showing that specific motion state. In this case, for example, the same motion pattern might be defined by showing two time intervals with a difference below a predetermined threshold. For example, the threshold could be determined based on the expected motion scale in the region of interest, the expected precision of the motion signal, etc. Furthermore, prior knowledge of the characteristic motion signal of the motion state can be used to determine the time intervals involving the motion state. For example, if the motion state is the expiratory state of a patient's lungs, the positional signal in that state can be known, i.e., the motion signal is almost constant for a certain period of time near a local minimum of the motion signal.

[0021] Specifically, if the motion signal indicates regular periodic or cyclical motion, the motion state in each period or cycle will be repeated. Therefore, the corresponding image data determination unit can be adapted to determine the time interval corresponding to a specific repetitive motion state in each period or cycle. However, if the periodic motion state has changed during the acquisition of nuclear image data, for example, due to a large change in patient position, it is possible to identify other motion states in the periodic motion following the position change.

[0022] If the nuclear image data can be represented as list pattern data, the corresponding image data determination unit can be adapted to classify the list pattern data according to the determined motion state. For example, all parts of the list pattern data that have been acquired during one or more time intervals of the motion state are provided together in the list pattern data and marked as belonging to the respective motion states.

[0023] The absorption map reconstruction unit is then adapted to reconstruct an absorption map for each motion state based on the corresponding nuclear image data for each motion state. The absorption map indicates the absorption of nuclear radiation in the region of interest. Specifically, the absorption map provides a value indicating the absorption of radiation in that portion of the region of interest for each part of the region of interest (e.g., for each pixel or voxel of the region of interest). The radiation absorption in a portion of the region of interest also indicates the density of the material in that portion. The absorption map reconstruction unit is adapted to reconstruct the absorption map for each motion state based on the corresponding nuclear image data by solving or approximating an exponential X-ray transform for the nuclear image data.

[0024] In a preferred embodiment, the absorption map reconstruction unit is adapted to reconstruct the absorption map for each motion state using a machine learning algorithm. Specifically, the absorption map reconstruction unit is adapted to use a machine learning algorithm to solve for the exponential X-ray transform. Preferably, the machine learning algorithm involves a trained neural network, specifically a generative adversarial network (GAN). For example, such a machine learning algorithm can be trained as the input and desired output of a neural network under training by providing multiple nuclear image datasets for different situations and providing corresponding known absorption maps for each situation (e.g., acquired using an X-ray CT imaging system). The neural network can then be trained to determine corresponding absorption maps also used for other nuclear image datasets. The paper "MedGAN: Medical Image Translation using GANs" by K. Armanious et al. (Computerized Medical Imaging and Graphics, Vol. 79, 2019) provides an example of a method for providing a machine learning algorithm to determine an absorption map based on nuclear image data. Because the absorption map for each determined motion state is determined based on nuclear image data rather than, for example, X-ray CT image data, the absorption map reflects the position of the region of interest during each corresponding motion state at the location where nuclear image data has been acquired.

[0025] In a preferred embodiment, the absorption map corresponds to a pseudo-CT image, wherein the absorption information provided by the pseudo-CT image corresponds to the absorption information provided by the CT image acquired during the CT imaging process. Preferably, the absorption information provided by the pseudo-CT image involves providing absorption information over the same range as that provided by the CT image, i.e., using the same values. Specifically, the absorption map reconstruction unit is adaptable to correspondingly scale the reconstructed absorption map. Alternatively, if a machine learning algorithm is used to reconstruct the absorption map, the machine learning algorithm can be trained using CT images of the region of interest, such that the trained machine learning algorithm will provide a corresponding pseudo-CT image as output when provided with kernel image data as input.

[0026] The nuclear image reconstruction unit is adapted to reconstruct one or more nuclear images of the region of interest based on nuclear image data and absorption maps reconstructed for each motion state from the nuclear image data. For example, the nuclear image reconstruction unit can be adapted to reconstruct one or more nuclear images of the region of interest using known reconstruction algorithms for nuclear image data that are being corrected using absorption data, such as X-ray CT image data.

[0027] In a preferred embodiment, the kernel image reconstruction unit is adapted to reconstruct an absorption-corrected kernel image for each motion state based on corresponding kernel image data and an absorption map. In this case, the absorption-corrected kernel image for each motion state can be provided to the user as a 4D kernel image set. Furthermore, in an embodiment, the kernel image reconstruction unit can be adapted to reconstruct the motion-corrected kernel image based on the kernel image data and the absorption map for each motion state. For example, the kernel image reconstruction unit can be adapted to use the absorption-corrected kernel image for each motion state and register the absorption-corrected kernel images for each motion state with each other to reconstruct the motion-corrected kernel image based on the registered absorption-corrected kernel image. Alternatively, in an embodiment, the kernel image reconstruction unit can be adapted to register the absorption maps for each motion state with each other and use a registration for registering the kernel image data to one of the motion states to reconstruct the motion-corrected kernel image. Since the absorption map is reconstructed based on the kernel image data, the registration of the absorption map directly provides the registration of the kernel image data from which the absorption map has been reconstructed. Therefore, this embodiment allows for very easy reconstruction of the motion-corrected kernel image.

[0028] Furthermore, the nuclear image reconstruction unit is also adapted to register high-resolution CT image data (e.g., preoperative high-resolution CT image data) with absorbance maps and use the registered high-resolution CT image data to reconstruct one or more nuclear images. Since absorbance maps indicate radiation absorption in the region of interest and therefore contain essentially the same information as X-ray CT images, registration between high-resolution CT image data and absorbance maps is very easy using known registration algorithms. Moreover, since absorbance maps very accurately reflect the state of the region of interest during each corresponding motion state, and the registered high-resolution CT image data also well reflects the motion state of the region of interest, even though they have not yet been acquired, for example, in the same motion state as the acquisition of the nuclear image data or even within the same time period, reconstructing one or more nuclear images using the registered high-resolution CT image data allows for more accurate motion and attenuation correction on the resulting one or more nuclear images. Preferably, an elastic registration algorithm is used to register absorbance maps to each other or to register absorbance maps with high-resolution CT image data.

[0029] In another aspect of the invention, a nuclear imaging system is proposed, comprising a) a detector for detecting nuclear events in the detector's field of view and determining nuclear image data of the patient's region of interest based on the detected nuclear events, and b) means as described above for providing one or more nuclear images. Preferably, the nuclear imaging system is a PET or SPECT imaging system, wherein the detector is adapted to detect gamma radiation originating from the patient's region of interest.

[0030] In another aspect of the invention, a method for providing nuclear images of a region of interest (ROI) for a patient is provided, wherein the method includes a) providing nuclear image data of the ROI of a patient acquired using a nuclear imaging device, b) providing motion signals indicating motion of the ROI during the acquisition of the nuclear image data, c) determining different motion states of the ROI based on the motion signals, wherein each of the different motion states indicates a different state of the ROI, d) determining nuclear image data corresponding to each motion state, wherein if nuclear image data has already been acquired during a state of the ROI corresponding to the motion state, the nuclear image data corresponds to the motion state, e) reconstructing an absorption map for each motion state based on the corresponding nuclear image data of the corresponding motion, wherein the absorption map indicates absorption of nuclear radiation in the ROI, and f) reconstructing one or more nuclear images of the ROI based on the nuclear image data and the absorption map reconstructed for each motion state.

[0031] In another aspect of the invention, a computer program for providing a kernel image of a region of interest is provided, wherein the computer program includes program code units for causing the apparatus to perform the steps described in the method when the apparatus executes the computer program.

[0032] It should be understood that the preferred embodiments of the present invention can also be any combination of the dependent claims or the above embodiments with the respective independent claims.

[0033] These and other aspects of the invention will become apparent from the embodiments described below and will be elucidated with reference to the embodiments described below. Attached Figure Description

[0034] In the following figures:

[0035] Figure 1 An embodiment of the nuclear imaging system according to the present invention is illustrated schematically and exemplary.

[0036] Figure 2 The workflow for reconstructing nuclear images based on the fundamental principles of this invention is illustrated schematically and exemplary.

[0037] Figure 3 A flowchart illustrating an exemplary embodiment of a method for providing a kernel image of a region of interest according to the present invention is shown. Detailed Implementation

[0038] Figure 1An embodiment of a nuclear imaging system according to the invention, including means for providing nuclear images of a patient's region of interest, is illustrated schematically and exemplary. In the following embodiment, the nuclear imaging system 100 includes a detector 120 for detecting nuclear events in the field of view of a detector 120. The nuclear imaging system 100 can be a PET imaging system, and the detector 120 can be a gamma radiation detector as used in a PET imaging system. Specifically, the field of view of the detector 120 includes the region of interest of a patient 122 lying on a patient table 121. The patient 122 has been injected with a radioactive material in the form of a radiopharmaceutical, including, for example, fluorine-18. In the case of a PET or SPECT imaging procedure, the radiopharmaceutical is selected to emit positrons, wherein the positrons, upon annihilation, provide two gamma photons moving in opposite directions. The detector 120 can be adapted to detect one or both of the gamma photons produced by annihilation (i.e., nuclear events) in the region of interest. If the nuclear imaging system 100 refers to a PET detector, then the detector 120 is adapted to detect two photons and provide the detection of photons in the form of list pattern data as nuclear image data. The nuclear imaging system 100 also includes means 110 for providing nuclear images of the region of interest to the patient 120.

[0039] The device 110 includes a nuclear image data providing unit 111, a motion signal providing unit 112, a motion state determining unit 113, a corresponding image data determining unit 114, an absorption map reconstruction unit 115, and a nuclear image reconstruction unit 116.

[0040] In this embodiment, the nuclear image data providing unit 111 is a receiving unit used to receive nuclear image data from the detector 120 and to provide the received nuclear image data. In this example, nuclear image data refers to PET image data acquired by the PET imaging system 100. However, in other embodiments, nuclear image data can refer to, for example, SPECT data or any other type of nuclear image data acquired using a nuclear imaging system. Nuclear imaging data has already been acquired by the detector 120 during a predetermined time period, for example, during a time period between several minutes and one hour based on the size of the region of interest.

[0041] In this embodiment, the motion signal providing unit 112 is adapted to provide a motion signal using a signal acquired by a motion sensor 123 attached to the patient's chest. In this case, the motion signal indicates respiratory motion in the patient's chest region during the acquisition of nuclear image data. The motion signal can be provided in the form of an accelerometer signal indicating the acceleration of the sensor 123 during the patient's breathing, or it can be provided as a position signal indicating the position of the sensor 123 during the patient's breathing. Generally, the motion signal refers to a sequence of measurements over time, such as a sequence of the position of the motion sensor 123 over time, or a sequence of acceleration values ​​measured by the motion sensor 123 over time.

[0042] Based on the motion signal provided by the motion signal providing unit 112, the motion state determination unit 113 is adapted to determine different motion states of the region of interest based on the motion signal. Since the provided motion signal in this case indicates the respiratory movements of the patient 122, the motion state determination unit 113 is adapted to determine different respiratory states of the patient 122's respiratory cycle based on the motion signal. For example, the motion state determination unit is adapted to determine first and second substantially stationary states as first and second motion states based on the motion signal, wherein the first and second motion states refer to the expiratory and inspiratory states of the patient 122's lungs, respectively. Since the patient is substantially stationary during the maximum inspiratory and expiratory states over short periods, the motion state determination unit can be adapted to search for portions of the motion signal indicating such substantially stationary states of the patient 122's region of interest. For example, if the motion signal provided by the motion sensor 123 refers to the acceleration of the motion sensor 123, the motion state determination unit can be adapted to search for portions of the motion signal in which the acceleration is substantially zero, and determine such portions as the motion state in which the patient is substantially stationary.

[0043] In another example, if the motion signal provided by the motion signal providing unit refers to the position of sensor 123, the motion state determination unit can be adapted to search for a portion of the motion signal in which the rate of change of the sensor's position is substantially zero, i.e., in which the sensor's position is substantially constant. In this context, the term "substantially" always refers to a deviation below a predetermined threshold, which can be defined based on the application, such as the expected quality of the signal, the patient's expected movement, and still acceptable deviation. Based on the provided motion signal, filters, such as averaging filters, can also be provided to avoid measurement inaccuracies in the motion signal when determining different motion states. Furthermore, the motion state determination unit 113 can also be adapted to determine the motion state located between the first and second motion states, i.e., the motion state between the expiratory and inspiratory states of the patient 122's lungs, as the motion state.

[0044] After the motion state determination unit has determined the motion state in the motion signal, for example, after determining the patient 122's maximum expiratory and inspiratory states in the motion signal as the motion state, the motion state determination unit can be adapted to present this determination result to the user on a display. The user can then check the determination result, for example, check whether the motion state determination unit has correctly determined the respiratory motion state, and can then use an input unit such as a keyboard or computer mouse to confirm, modify, or reject the determination result. In other embodiments, the motion state determination unit 113 can be adapted to determine different motion states during interaction with the user, for example, by providing the motion signal to the user using a display, and by receiving the determined motion state from the user using an input unit such as a keyboard or mouse.

[0045] Since respiratory movements are regular and specifically cyclical, the patient 122's body will repeat the same state of motion for each respiratory cycle, where each state of motion refers to a specific state of location of, for example, the patient 122's region of interest.

[0046] Then, the corresponding image data determination unit 114 determines nuclear image data corresponding to each motion state. Specifically, the corresponding image data determination unit 114 is adapted to determine whether nuclear image data was acquired during the time period when the patient was in each corresponding motion state during the acquisition of nuclear image data. For example, the corresponding image data determination unit 114 is adapted to determine all time intervals during the acquisition of nuclear image data when the patient was in a specific motion state based on motion signals. In an example, the corresponding image data determination unit 114 is adapted to use features of a pre-known motion signal relating to a specific motion state. In this example, the corresponding image data determination unit 114 is adapted to determine all time intervals when the patient was in a first motion state, i.e., an exhalation state, and all time intervals when the patient was in a second motion state, i.e., an inhalation state. The corresponding image data determination unit 114 is then adapted to determine, for example, based on a timestamp provided for each detected event in the nuclear image data, which nuclear image data was acquired during the time intervals relating to the first and second motion states.

[0047] If the nuclear image data refers to list pattern data, the corresponding image data determination unit 114 can be adapted to classify the list pattern data according to the determined motion state. The classified list pattern data, i.e., the corresponding nuclear image data, can then be provided, for example, in the form of a table or list linking portions of the nuclear image data to their corresponding motion states. If the motion signal indicates regular, specific, periodic movements of the patient 122, such as respiratory movements, the same motion state can be identified in each cycle of the periodic movement, and the corresponding nuclear image data can then be linked to the motion state, for example, in tabular form.

[0048] Then, the absorption map reconstruction unit 115 reconstructs an absorption map for each motion state based on the corresponding kernel image data for each motion state. Specifically, if the motion is periodic, such as breathing, kernel image data from all cycles of the periodic motion corresponding to a specific motion state that repeats during the periodic motion can be used to reconstruct the absorption map for that motion state of the periodic motion. However, in other embodiments, if the motion is not periodic or includes irregular motion, some motion states may not repeat. These motion states refer to unique states of the region of interest, and therefore kernel image data acquired only during the time period of such unique states can be used to reconstruct the absorption map.

[0049] In this example, the absorption map reconstruction unit 115 uses a trained neural network, specifically a generative adversarial network (GAN), to reconstruct an absorption map based on corresponding nuclear image data as a motion state. The trained neural network can be trained, for example, by providing multiple nuclear image datasets and the desired outputs of these datasets—specifically, X-ray CT images defining the attenuation of the region of interest imaged using the nuclear image datasets—before being applied to the nuclear image data. In this case, after the training phase, the trained neural network will provide a corresponding pseudo-CT image as an absorption map based on the nuclear image data, representing the absorption information of the region of interest imaged by the nuclear imaging system in the same manner as a standard X-ray CT image would provide.

[0050] Then, the nuclear image reconstruction unit 116 reconstructs an attenuation-corrected nuclear image based on the nuclear image data and the determined absorbance map, for example, for each motion state of the respiratory cycle. The attenuation correction and reconstruction of the nuclear image data can be based on known reconstruction methods using, for example, normally generated X-ray CT image data. Additionally or alternatively, the nuclear image reconstruction unit 116 can also reconstruct a motion-corrected nuclear image by registering the absorbance maps to each other and by using this registration to also register the nuclear image data to each other, i.e., correcting the nuclear image data from any motion. Therefore, the nuclear image reconstructed from the registered nuclear image data can be considered a motion-corrected nuclear image. During the reconstruction of the motion-corrected nuclear image, the nuclear image reconstruction unit 116 can also be adapted to use the registered absorbance map also for the attenuation-corrected reconstructed nuclear image.

[0051] If, for example, high-resolution CT image data of patient 122 has been acquired prior to the acquisition of nuclear image data, the absorption map can also be used to register the pre-procedural high-resolution CT image data with each absorption map, that is, to adapt the pre-procedural high-resolution CT image data to each motion state. The nuclear image reconstruction unit 116 can then also use the pre-procedural high-resolution CT image data, which has been registered to the absorption map, for attenuation correction of the nuclear image data during nuclear image reconstruction, or as an overlay map for presenting the nuclear image data in cases where anatomical structures are visible in the pre-procedural high-resolution CT image data.

[0052] Figure 2 The process of reconstructing one or more kernel images according to the principles of the present invention is illustrated. In this example, such as... Figure 2 As indicated by the illustrated PET imaging system 120, PET imaging data is first acquired. In this embodiment, the patient is monitored via a camera 123 that provides motion signals 211, for example, by analyzing monitoring images of changes in the region of interest. The PET imaging data is then classified according to the identified motion states in the motion signals 211, such as... Figure 2 As indicated in Table 212, based on this categorized PET image data, as shown in Table 212, for each motion state, corresponding attenuation maps 214, indicated by AM1, AM2, AM3, AM4, etc., are determined based on the PET image data. Using the motion signal 211 provided at 215, the attenuation maps 214 can be combined 216 into a 4D attenuation map 217. The 4D attenuation map 217 can then be used directly to reconstruct one or more nuclear images, or it can be averaged in step 218 into a 3D attenuation map 219, which can also be used to reconstruct nuclear images.

[0053] Figure 3A flowchart illustrating an exemplary embodiment of a method for providing a nuclear image of a region of interest for patient 122 is shown. Method 300 includes a first step 310 of providing nuclear image data of the region of interest for patient 122, wherein nuclear image data has been acquired using, for example, a nuclear imaging system 100. In step 311, a motion signal indicative of motion of the region of interest for patient 122 is provided, for example, by receiving a motion signal from a motion sensor 123. In method 300, steps 310 and 311 of providing nuclear image data and providing motion signals can be processed in any order or simultaneously.

[0054] In the next step 312, different motion states of the region of interest are determined based on the motion signal, wherein each of the different motion states indicates a different state of the region of interest, preferably a different substantially stationary state of the region of interest. The step of determining the motion state can also refer to a step of providing the motion state, for example, based on previously stored motion states or motion states received from user input. Based on the different motion states, in step 313, nuclear image data corresponding to each motion state is determined. If nuclear image data has already been acquired during the same state of the region of interest involved in the motion state, then the nuclear image data corresponds to the motion state. In step 314, an absorption map is reconstructed for each motion state based on the corresponding nuclear image data for each motion state, for example using a trained neural network. The absorption map indicates the absorption of nuclear radiation in the region of interest. In the final step 315, one or more nuclear images of the region of interest are reconstructed based on the nuclear image data and the absorption map reconstructed for each motion state.

[0055] Hybrid PET / CT imaging is an established clinical modality in nuclear medicine because CT images allow for correction of PET image data to image degradation effects such as photon attenuation and scattering, while also providing a temporary layover that allows for anatomical localization of features within PET images. Typical PET image acquisition times do not allow for the avoidance of regular body movements, such as breathing, while CT image acquisition times are short enough to result in spatial mismatches between CT and PET images.

[0056] To overcome this problem, embodiments of the invention propose recovering pseudo-CT images only from PET image data. Specifically, an AI-based approach allows for sufficient image quality for such pseudo-CT images to correct image artifacts and for anatomical localization. Furthermore, the invention, for example, adds a temporal domain to the pseudo-CT images, i.e., by recovering pseudo-CT images not as a single image but as a time series. When PET conformation data is acquired in list mode as kernel image data, each entry carries a detection timestamp, allowing data to be separated by time intervals. The resulting time series pseudo-CT images, or derived averages of these images, better match body motion states during PET image data acquisition and provide better-matched anatomical localization coverage for the PET images.

[0057] For the problem of exponential X-ray transformation, i.e., recovering the body's absorption map from acquired PET image data, approximate solutions are known. Furthermore, generative adversarial networks (GANs) can be used to recover absorption information, i.e., the absorption map, based on patient PET image data and corresponding CT image data. Since PET image data acquisition is time-consuming, a single acquisition typically covers multiple respiratory cycles. Therefore, deriving a solution for time-related exponential X-ray transformation is of interest for providing accurate attenuation correction for PET image data.

[0058] For example, it is known that respiratory motion correction in PET image data can be achieved via external respiratory sensors and respiratory gating, including methods such as 4D respiratory CT imaging to correct motion in PET image data and provide accurate time-related absorption correction and anatomical localization. However, respiratory signals can also be derived directly from PET list pattern data, i.e., from PET image data. Furthermore, determining the entire spatial motion pattern directly from the PET image data itself is even more accurate.

[0059] To perform time-dependent estimation of exponential X-ray transform, this invention proposes, for example, using a neural network to generate pseudo-4D CT image data from PET image data. Furthermore, it is possible to derive an estimate of motion taken during PET image data acquisition from the generated pseudo-4D CT image data. This estimate can then be used, for example, to correct motion during PET image data acquisition and to accurately register with previously acquired high-resolution CT image data.

[0060] In one embodiment, a process for implementing the invention is proposed, comprising the first step of setting up a PET scanner and acquiring PET image data for a patient's moving parts, such as the chest or abdomen. In a next step, a hardware- or software-based method is applied to detect the patient's motion or respiratory signals. In yet another step, the PET list pattern data, i.e., the PET image data, is classified based on derived triggers (e.g., respiratory signals) regarding different motion states of the patient during acquisition. In a final step, a reconstruction method, such as using a trained neural network, can be used to reconstruct an absorption map from the acquired PET image data, such as pseudo-CT image data at each time step.

[0061] The generated pseudo-4D CT image data represents motion during PET image data acquisition and can be used for time-dependent absorption correction and anatomical localization. Furthermore, it can be used, for example, to spatially register high-resolution CT image data prior to the procedure using elastic registration, thereby integrating high-quality CT image data into absorption correction and anatomical localization.

[0062] Although the nuclear imaging system is a PET imaging system in the above embodiments, it can also be a SPECT imaging system in other embodiments.

[0063] Although the motion signal in the above embodiments is provided based on motion sensor signals as acceleration or position signals, in other embodiments, the motion signal can also be provided by analyzing monitoring images from a monitoring camera that monitors the patient during nuclear imaging data acquisition. Furthermore, in another embodiment, the motion signal providing unit can determine the motion signal based on the nuclear image data itself, specifically by identifying motion in the region of interest within the nuclear image data itself.

[0064] By studying the accompanying drawings, the disclosure, and the appended claims, those skilled in the art can understand and implement other variations to the disclosed embodiments when practicing the claimed invention.

[0065] In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plural.

[0066] A single unit or device can perform the functions of several items listed in the claims. The fact that certain measures are listed in different dependent claims does not mean that a combination of these measures cannot be used advantageously.

[0067] Programs such as providing nuclear image data, providing motion signals, determining different motion states, determining corresponding nuclear image data, reconstructing absorption maps, and reconstructing one or more nuclear images, which are executed by one or more units or devices, can be executed by any other number of units or devices. For example, these programs can be executed by a single device. These programs can be implemented as computer program code devices and / or dedicated hardware.

[0068] Computer programs can be stored / distributed on suitable media such as optical storage media or solid-state media, provided 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.

[0069] Any reference numerals in the claims should not be construed as limiting the scope.

[0070] This invention relates to an apparatus for improving the image quality of nuclear images, such as PET images. The apparatus includes a providing unit for providing nuclear image data of a region of interest (ROI), a providing unit for providing motion signals indicating motion of the ROI, a determining unit for determining different motion states of the ROI based on the motion signals, a determining unit for determining nuclear image data corresponding to each motion state, a reconstruction unit for reconstructing an absorption map for each motion state based on the corresponding nuclear image data for each motion state, and a reconstruction unit for reconstructing one or more nuclear images of the ROI based on the nuclear image data and the absorption maps reconstructed for each motion state.

Claims

1. An apparatus for providing nuclear images of regions of interest of a patient (122), wherein, The device (110) includes: - Nuclear image data providing unit (111), which is used to provide nuclear image data of the region of interest of a patient (122) acquired using a nuclear imaging device. - A motion signal providing unit (112) for providing motion signals indicating the motion of the region of interest of the patient (122) during the acquisition of the nuclear image data. - A motion state determination unit (113) is used to determine different motion states of the region of interest based on the motion signal, wherein each of the different motion states indicates a different state of the region of interest. - A corresponding image data determination unit (114) is used to determine kernel image data corresponding to each motion state, wherein if kernel image data has been acquired during the state corresponding to the motion state in the region of interest, then the kernel image data corresponds to the motion state. - An absorption map reconstruction unit (115) is used to reconstruct an absorption map for each motion state based on corresponding nuclear image data of the corresponding motion state, wherein the absorption map indicates the absorption of nuclear radiation in the region of interest, and - A nuclear image reconstruction unit (116) is used to register CT image data to the absorption map for each motion state, and to reconstruct one or more nuclear images of the region of interest based on the nuclear image data and the registered CT image data for each motion state.

2. The apparatus according to claim 1, wherein, The absorption map reconstruction unit (115) is adapted to use a machine learning algorithm to reconstruct the absorption map for the motion state.

3. The apparatus according to claim 2, wherein, The machine learning algorithm involves a trained neural network.

4. The apparatus according to any one of claims 1-3, wherein, The absorption map corresponds to a pseudo-CT image, wherein the absorption information provided by the pseudo-CT image corresponds to the absorption information provided by the CT image acquired during the X-ray CT imaging process.

5. The apparatus according to any one of claims 1-3, wherein, The corresponding image data determination unit (114) is adapted to determine one or more time intervals for each motion state, during which the region of interest is in a state corresponding to the motion state during the acquisition of the nuclear image data based on the motion signal, and the corresponding image data determination unit (114) is also adapted to determine the nuclear image data corresponding to the motion state based on whether the nuclear image data has been acquired during the one or more time intervals of the corresponding motion state.

6. The apparatus according to any one of claims 1-3, wherein, The motion signal providing unit (112) is adapted to determine the motion signal based on the nuclear image data.

7. The apparatus according to any one of claims 1 to 3, wherein, The motion signal providing unit (112) is adapted to provide a sensor signal as a motion signal, the sensor being configured to detect motion of the region of interest of the patient (122).

8. The apparatus according to any one of claims 1-3, wherein, The motion signal indicates regular body movements in the region of interest of the patient (122).

9. The apparatus according to any one of claims 1-3, wherein, The nuclear image reconstruction unit (116) is adapted to reconstruct an absorption-corrected nuclear image for each motion state based on the corresponding nuclear image data and the absorption map.

10. The apparatus according to any one of claims 1-3, wherein, The nuclear image reconstruction unit (116) is adapted to reconstruct a motion-corrected nuclear image based on the nuclear image data and the absorption map for each motion state.

11. The apparatus according to claim 10, wherein, The nuclear image reconstruction unit (116) is adapted to register the absorption maps of each motion state with each other and to use the registration to register the nuclear image data to one of the motion states in order to reconstruct the motion-corrected nuclear image.

12. The apparatus according to any one of claims 1-3, wherein, The nuclear image data refers to PET image data or SPECT image data.

13. A nuclear imaging system, comprising: - A detector (120) for detecting nuclear events in the field of view of the detector (120) and determining nuclear image data of the region of interest of the patient (122) based on the detected nuclear events. - The apparatus (110) according to claim 1.

14. A method for providing a nuclear image of a region of interest in a patient (122), wherein, The method includes: - Provides (310) nuclear image data of the region of interest of the patient (122) acquired using a nuclear imaging device, - Provide (311) a motion signal indicating the motion of the region of interest of the patient (122) during the acquisition of the nuclear image data, - Based on the motion signal, determine (312) different motion states of the region of interest, wherein each of the different motion states indicates a different state of the region of interest. - For each motion state, determine (313) kernel image data corresponding to the motion state, wherein the kernel image data corresponds to the motion state if the kernel image data has already been acquired during a state corresponding to the region of interest of the motion state. - An absorption map is reconstructed for each motion state (314) based on the corresponding nuclear image data of the respective motion state, wherein the absorption map indicates the absorption of nuclear radiation in the region of interest, and - Register the CT image data to the absorption map for each motion state, and reconstruct one or more nuclear images of the region of interest based on the nuclear image data and the registered CT image data for each motion state. (315) 15. A computer program product for providing a kernel image of a region of interest, wherein, The computer program product includes program code units, which, when the program code units are... When the apparatus (110) according to claim 1 is operated, the program code unit is used to cause the apparatus (110) to perform the steps of the method according to claim 14.

Images