Method and system for motion detection in positron emission tomography

Abstract

Methods and systems for medical imaging systems are provided. In one embodiment, a method for a medical imaging system includes: acquiring emission data during a positron emission tomography (PET) scan of a patient; reconstructing a series of live PET images while the emission data is being acquired; and tracking motion of the patient during the acquisition based on the series of live PET images. In this way, patient motion during a scan can be identified and compensated for via scan acquisition and / or data processing adjustments, resulting in diagnostic PET images that are reduced in motion artifacts and improved in diagnostic quality.

Description

TECHNICAL FIELD

[0001] Embodiments of the subject disclosed herein relate to non-invasive diagnostic imaging, and more particularly to positron emission tomography (PET). BACKGROUND

[0002] Positron emission tomography (PET) generates images representing the distribution of positron emitting radioactive tracers within a patient's body, which can be used to observe metabolic processes within the body and diagnose diseases. During operation of a PET imaging system, a radioactive tracer is initially injected into the patient, which emits positrons as it decays. Each emitted positron can travel a relatively short distance before encountering an electron, at which point annihilation occurs. When a positron interacts with an electron through annihilation, the entire mass of the positron-electron pair is converted into two 511 keV gamma photons (also referred to as 511 keV events). The photons are emitted in opposite directions along a line of response (LOR). The annihilation photons are detected by detectors placed on both sides of the LOR in a configuration such as a detector ring, as coincident events. Thus, during data acquisition, the detectors detect coincident events that reflect the distribution of the radioactive tracer within the patient's body. Accordingly, images reconstructed from the acquired image data include annihilation photon detection information. Typically, the images are reconstructed upon completion of data acquisition, and until the images are reconstructed, it can be unknown whether the acquired data is sufficient to produce high quality images. SUMMARY

[0003] In one embodiment, a method for a medical imaging system includes acquiring emission data during a positron emission tomography (PET) scan of a patient, reconstructing a series of live PET images while the emission data is being acquired, and tracking motion of the patient during the acquisition based on the series of live PET images. In this way, patient motion during the PET scan can be accurately detected and compensated for, thereby reducing motion artifacts and improving the diagnostic quality of the resulting PET images.

[0004] It should be appreciated that the above Brief Summary is provided merely for purposes of summarizing some embodiments of the subject matter disclosed herein so as to provide a basic understanding of some aspects of the subject matter. This Brief Summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to limit the scope of the claimed subject matter in any way. Furthermore, the claimed subject matter is not limited to implementing any embodiments highlighted in the above Summary or in any portion of this disclosure. BRIEF DESCRIPTION OF DRAWINGS

[0005] The present disclosure will be better understood by reading the following description of non-limiting embodiments, with reference to the appended drawings, in which:

[0006] Figure 1A pictorial view of an exemplary multi-modality imaging system is shown in accordance with the embodiments of the present disclosure.

[0007] Figure 2 A block schematic diagram of an exemplary positron emission tomography (PET) imaging system is shown in accordance with the embodiments of the disclosure.

[0008] Figure 3 A high level flowchart of an exemplary method for performing PET computed tomography with real-time PET image reconstruction to detect and respond to patient motion is shown in accordance with the embodiments of the present disclosure.

[0009] Figure 4 A flowchart of an exemplary method for real-time reconstruction of a PET image to detect patient motion is shown in accordance with the embodiments of the present disclosure.

[0010] Figure 5 How a bin is selected on a response line for real-time PET image reconstruction is shown schematically in accordance with the embodiments of the present disclosure.

[0011] Figure 6 How a projection weight for real-time PET image reconstruction can be determined is shown schematically in accordance with the embodiments of the present disclosure.

[0012] Figure 7 An example of detecting patient motion during PET using a real-time reconstructed PET image is shown. DETAILED DESCRIPTION

[0013] The following description relates to various embodiments of medical imaging systems. In particular, methods and systems for real-time reconstruction of positron emission tomography (PET) images to detect patient motion are provided. In Figure 1 Examples of imaging systems that can be used to acquire images processed in accordance with the techniques of the present technology are provided in the detailed description. In this document, the imaging system can be a multi-modality system. In one embodiment, the multi-modality imaging system can be a positron emission tomography / computed tomography (PET / CT) imaging system, where the first modality is a CT imaging system and the second modality is a PET imaging system (e.g., as shown in Figure 1 and Figure 2 ).

[0014] When a patient is scanned using a PET imaging system, events captured within the imaging system's field of view (FOV) can be used to reconstruct a functional image of the patient. However, uncompensated patient motion during scanning can degrade the quality of the resulting images. For example, due to patient motion during scanning, PET images may exhibit image artifacts, blurring, and increased noise, which can reduce the diagnostic value of the images. If the technician operating the PET imaging system observes patient movement, the technician can instruct the patient to remain still and extend the duration of the scan. Alternatively, the technician can repeat the scan. However, because the technician may not be able to clearly see the patient within the imaging system, they may not be aware of the patient's movement, resulting in degradation of the PET image or other mixed-modality image. In some cases, an inaccurate diagnosis may be made from the degraded image. In other cases, the degraded image may not be used, and a rescan may be requested. A rescan may involve reinjecting the patient with radiotracer and repeating the entire scan, which increases costs, causes discomfort to the patient, and reduces the usability of the imaging system.

[0015] therefore, Figure 3 An exemplary method for detecting and tracking patient motion in a PET imaging system is shown in Figure 4 1. Real-time PET images reconstructed using an exemplary fast reconstruction method shown in FIG. In this paper, list mode reconstruction is performed on image space data acquired during a very short time frame, thereby allowing a manageable data size that can be processed in real time. For example, time-of-flight (TOF) PET information can be utilized in the fast reconstruction method to generate image projections such as Figure 5 and Figure 6 Image registration between real-time PET images from different time frames can be used for patient motion analysis, an example of which is shown in Figure 7 When patient motion exceeds a threshold that will result in motion-related degradation in the final PET image used for diagnosis (which is different from the real-time PET image used for motion detection), various motion detection responses can be employed. For example, the scan time can be selectively extended to capture additional motion-free data, data acquired during periods of patient motion can be discarded, advanced motion correction reconstruction techniques can be used during final image reconstruction for data acquired during periods of patient motion, etc. Further, in hybrid imaging modalities such as Figure 1 In a PET / CT system (shown in FIG), CT data acquired before detected patient motion can also be repeated. By accounting for patient motion in real time during a PET scan, imaging resources can be used more efficiently and scanning costs can be reduced. Furthermore, patient discomfort can be reduced. Even further, the time it takes to make an accurate diagnosis can be reduced.

[0016] Although a PET / CT imaging system is described by way of example, it should be understood that the present technology can also be usable when applied to images acquired using other imaging modalities, such as CT, tomosynthesis, MRI, ultrasound, etc. The present discussion of the PET / CT imaging modality is provided merely as an example of one suitable imaging modality. In other examples, a PET / MRI imaging system or other imaging system including a PET imaging modality can be used.

[0017] As used herein, the phrase "reconstructed image" is not intended to exclude embodiments of the present disclosure in which data representative of an image is generated rather than a visual image. Thus, as used herein, the term "image" refers broadly to both visual images and data representative of visual images. However, many embodiments generate (or are configured to generate) at least one visual image.

[0018] Turning now to the drawings, Figure 1 and Figure 2 A multi-modality imaging system 10 is shown. The multi-modality imaging system 10 can be a suitable type of imaging system, such as, for example, a positron emission tomography (PET) imaging system, a single photon emission computed tomography (SPECT) imaging system, a PET / computed tomography (CT) imaging system, a PET / ultrasound imaging system, a PET / magnetic resonance imaging (MRI) system, or any other imaging system capable of generating tomographic images by PET. Various embodiments are not limited to multi-modality medical imaging systems, but can be used in single-modality medical imaging systems, such as, for example, a standalone PET imaging system or a standalone SPECT imaging system. Moreover, various embodiments are not limited to medical imaging systems for imaging human subjects, but can include veterinary systems or non-medical systems for imaging non-human subjects.

[0019] Referring first to Figure 1 , the multi-modality imaging system 10 includes a first modality unit 11 and a second modality unit 12. These two modality units enable the multi-modality imaging system 10 to scan a subject or patient using the first modality unit 11 in a first modality and using the second modality unit 12 in a second modality. The multi-modality imaging system 10 allows multiple scans in different modalities to facilitate improved diagnostic capabilities of a single modality system. In Figure 1In the illustrated embodiment, the multi-modality imaging system 10 is a positron emission tomography / computed tomography (PET / CT) imaging system 10. In this example, the first modality unit 11 is a CT imaging system 11 and the second modality unit 12 is a PET imaging system 12. The PET / CT system 10 is shown to include a gantry 13 (or first gantry portion) included in the CT imaging system 11 and a gantry 14 (or second gantry portion) included in the PET imaging system 12. For example, the CT imaging system 11 can generate anatomical images of a patient while the PET imaging system 12 can generate functional images corresponding to the distribution of a radioactive tracer that is a marker of a physiological process, such as metabolism. As noted above, modalities other than CT and PET can be employed with the multi-modality imaging system 10.

[0020] The gantry 13 includes an x-ray source 15 that projects a beam of x-ray radiation (or x-rays) for imaging a patient 21 positioned on a motorized table 24. In particular, the x-ray source 15 is configured to project a beam of x-ray radiation toward a detector array 18 positioned on an opposite side of the gantry 13. Although only a single x-ray source 15 is depicted, in certain embodiments, multiple x-ray sources and detectors can be employed to project multiple beams of x-ray radiation for acquiring projection data at different energy levels corresponding to the patient. Figure 1 Although only a single x-ray source 15 is depicted, in certain embodiments, multiple x-ray sources and detectors can be employed to project multiple beams of x-ray radiation for acquiring projection data at different energy levels corresponding to the patient. In some embodiments, the x-ray source 15 can enable dual energy gemstone spectral imaging (GSI) through fast peak kilovoltage (kVp) switching. In some embodiments, the x-ray detector employed is a photon counting detector that is capable of distinguishing x-ray photons of different energies. In other embodiments, two sets of x-ray sources and detectors are used to generate dual energy projections, with one set of x-ray sources and detectors set to low kVp and the other set set to high kVp. It will thus be appreciated that the methods described herein can be implemented with single energy acquisition techniques as well as dual energy acquisition techniques.

[0021] In certain embodiments, the CT imaging system 11 also includes a controller or processor 25 that is configured to reconstruct images of a target volume of the patient 21 using an iterative or analytical image reconstruction method. For example, the controller or processor 25 can reconstruct images of a target volume of the patient using an analytical image reconstruction method, such as filtered back projection (FBP). As another example, the controller or processor 25 can reconstruct images of a target volume of the patient 21 using an iterative image reconstruction method, such as advanced statistical iterative reconstruction (ASIR), conjugate gradient (CG), maximum likelihood expectation maximization (MLEM), model-based iterative reconstruction (MBIR), and the like. As described further herein, in some examples, the controller or processor 25 can use an analytical image reconstruction method, such as FBP, in addition to an iterative image reconstruction method.

[0022] In some CT imaging system configurations, the x-ray source projects a cone of x-ray radiation that is collimated to lie in the X-Y-Z plane of a Cartesian coordinate system and is commonly referred to as the "imaging plane." The x-ray radiation beam passes through an object being imaged, such as a patient or subject. The x-ray radiation beam, after being attenuated by the object, impinges on an array of detector elements. The intensity of the attenuated x-ray radiation beam received at the detector array depends on the attenuation of the radiation beam by the object. Each detector element of the array produces a separate electrical signal that is a measurement of the attenuation of the x-ray beam at the detector location. The attenuation measurements from all of the detector elements are separately acquired to produce a transmission profile.

[0023] In some CT systems, a gantry is used to rotate the x-ray source and detector array around an object to be imaged in the imaging plane so that the angle at which the radiation beam intersects the object is constantly changing. A set of x-ray radiation attenuation measurements (e.g., projection data) from the detector array at one gantry angle is referred to as a "view." A "scan" of the object includes a set of views taken at different gantry angles or viewing angles during one rotation of the x-ray source and detector. It is contemplated that the benefits of the methods described herein arise from medical imaging modalities other than CT, and thus as used herein, the term "view" is not limited to the use described above with respect to projection data from one gantry angle. The term "view" is used to refer to one data acquisition whenever there are multiple data acquisitions from different angles, whether from CT, PET, or single photon emission CT (SPECT) acquisition, and / or any other modalities, including yet to be developed modalities, and combinations of them in fusion implementations.

[0024] The projection data is processed to reconstruct an image corresponding to a two-dimensional slice taken through the object, or in some examples where the projection data includes multiple views or scans, to reconstruct an image corresponding to a three-dimensional rendering of the object. One method for reconstructing an image from a set of projection data is referred to in the art as filtered backprojection techniques. Transmission and emission tomographic reconstruction techniques also include statistical iterative methods such as maximum likelihood expectation maximization (MLEM) and ordered subsets expectation reconstruction techniques and iterative reconstruction techniques. This method converts the attenuation measurements from the scan into an integer called a "CT number" or "Hounsfield unit" that is used to control the brightness of the corresponding pixel on a display device.

[0025] To reduce the total scan time, a "helical" scan can be performed. To perform a "helical" scan, the patient is moved while data for a prescribed number of slices is acquired. Such systems produce a single helix from the cone beam helical scan. The helix mapped out by the cone beam produces projection data from which an image in each of the prescribed slices can be reconstructed.

[0026] In certain embodiments of the multi-modality imaging system 10, the controller or processor 25 can be configured to operate both the CT imaging system 11 and the PET imaging system 12. In other embodiments, the CT imaging system and the PET imaging system can each include a dedicated controller that controls the CT imaging system and the PET imaging system, respectively.

[0027] continue Figure 2 , showing Figure 1 . The PET imaging system 12 includes a detector ring assembly 40 that includes a plurality of detectors (or detector crystals) 62. For example, the detector ring assembly 40 can be positioned in the gantry 14. Further, each of the detectors 62 includes one or more crystals (e.g., scintillation crystals) and one or more light sensors. In another example, the detectors 62 can each include one or more avalanche photodiodes, photomultipliers, silicon photomultipliers, and / or another type of radiation detector. The PET imaging system 12 also includes a controller or processor 44 configured to control the normalization process and the image reconstruction process. The controller 44 is operatively coupled to an operator workstation 46. In one non-limiting example, the controller 44 can be Figure 1 As another example, the controller 44 may include Figure 1 In or communicatively connected to the controller 25 (e.g., a portion thereof), such as via wired or wireless communication. Figure 2 In the example shown, the controller 44 includes a data acquisition processor 48 and an image reconstruction processor 50 interconnected via a communication link 52. The PET imaging system 12 acquires scan data and transmits the data to the data acquisition processor 48. The scanning operation is controlled from the operator workstation 46. The image reconstruction processor 50 is used to reconstruct the data acquired by the data acquisition processor 48.

[0028] The detector ring assembly 40 includes a central opening into which a subject or patient, such as patient 21, may be positioned using, for example, a motorized table 24 (e.g., Figure 1 4 (shown) is positioned in the central opening. The motorized table 24 is aligned with the central axis of the detector ring assembly 40. The motorized table 24 moves the patient 21 into the central opening of the detector ring assembly 40 in response to one or more commands received from the operator workstation 46. A PET scanner controller 54 (also known as a PET gantry controller) is disposed (e.g., mounted) within the PET imaging system 12. The PET scanner controller 54 responds to commands received from the operator workstation 46 via a communication link 52. Thus, scanning operations can be controlled from the operator workstation 46 via the PET scanner controller 54.

[0029] A radiopharmaceutical (e.g., a radionuclide) or radioactive tracer is delivered to the patient 21 prior to the PET scan. For example, the radionuclide can be fluorine-18, carbon-11, nitrogen-13, oxygen-15, etc., and can be injected into the patient 21. For example, the radionuclide can be incorporated into molecules that are typically metabolized by the body, or into molecules that bind to receptor targets. Thus, the radionuclide accumulates within organs, blood vessels, etc. The radionuclide undergoes positron emission decay, emitting a positron that collides with an electron in the surrounding tissue. The positron encounters an electron, and when the positron collides with the electron, both the positron and the electron annihilate and transform into a pair of photons or gamma rays each having any energy of 511 keV. The two photons are directed in substantially opposite directions, and are detected when they each reach respective detectors 62 positioned opposite one another on the detector ring assembly 40. Thus, the two detectors 62 that detect coincident scintillation events are positioned substantially 180 degrees from one another. When a photon collides with a detector, it produces a scintillation event (e.g., a flash of light) on the detector crystal. Each photomultiplier tube of the respective detector 62 detects the scintillation event and produces an analog signal that is transmitted on a communication line 64. A set of acquisition circuits 66 receives the analog signals from the photomultiplier tubes via the communication lines 64. The acquisition circuits 66 produce digital signals indicative of the three-dimensional (3D) location and total energy of the event. The acquisition circuits 66 also produce an event detection pulse indicative of the time or instant at which the scintillation event occurred. These digital signals are transmitted over a communication link (e.g., a cable) to an event localizer circuit 68 in the data acquisition processor 48.

[0030] The data acquisition processor 48 includes the event localizer circuit 68, an acquisition CPU 70, and a coincidence detector 72. The data acquisition processor 48 periodically samples the signals produced by the acquisition circuits 66. The acquisition CPU 70 controls communications on the backplane bus 74 and the communication link 52. The event localizer circuit 68 processes information about each event and provides a set of digital numbers or values indicative of the detected event. For example, the information indicates when the event occurred and the location of the detector 62 that detected the event. The event data packets are communicated over the backplane bus 74 to the coincidence detector 72. The coincidence detector 72 receives the event data packets from the event localizer circuit 68 and determines whether any two of the detected events coincide. Coincidence is determined by a number of factors. First, the time stamps in each event data packet must be within a predetermined time period of one another, such as within 12.5 nanoseconds, to indicate a coincidence. Second, a line of response (LOR) 35 formed by a straight line coupling the two detectors that detected the coincident events should pass through a field of view (FOV) 22 in the PET imaging system 12. Events that cannot be paired are discarded. Coincident event pairs are localized and recorded as coincidence data packets that are communicated over a communication link 78 to a sorter / histogram generator 80 in the image reconstruction processor 50.

[0031] Image reconstruction processor 50 includes a sorter / histogram generator 80. During operation, sorter / histogram generator 80 generates a data structure known as a histogram. The histogram includes a large number of bins, where each bin corresponds to a unique pair of detector crystals in PET imaging system 12. Because a PET imaging system typically includes thousands of detector crystals, the histogram typically includes millions of bins. Each bin of the histogram also stores a count value representing the number of coincidence events detected by the pair of detector crystals of that bin during a scan. At the end of a scan, the data in the histogram is used to reconstruct an image of patient 21. The complete histogram containing all of the data from a scan is often referred to as the "resulting histogram." The term "histogram generator" generally refers to the components (e.g., processors and memory) of controller 44 that perform the function of creating the histogram.

[0032] Image reconstruction processor 50 also includes a memory module 82, an image CPU 84, an array processor 86, and a communication bus 88. During operation, sorter / histogram generator 80 organizes the data from the coincidence events into a histogram. The histogram is then communicated to image CPU 84 via communication bus 88. Image CPU 84 controls the communication of the histogram to array processor 86 via communication bus 88. Array processor 86 receives the histogram as input and reconstructs an image in the form of an image array 92. The resulting image array 92 is then stored in memory module 82. Counts are made for all events occurring along each parallel LOR and a projection is formed. For example, a line that is integrated along all parallel LORs at an angle and a distance s from the center of FOV 22 forms a projection p The projections for all angles are further organized into a data array 90. Data array 90 can be a sinogram, which is a function of s and A single projection fills a row in the sinogram, and the sinogram includes a superposition of all projections weighted by the average counts at each point. Data array 90 is stored in memory module 82. Communication bus 88 links image CPU 84 to communication link 52. Image CPU 84 controls the communication via communication bus 88. Array processor 86 is also connected to communication bus 88. Array processor 86 receives data array 90 as input and reconstructs an image in the form of an image array 92. The resulting image array 92 is then stored in memory module 82.

[0033] Images stored in the image array 92 are transferred by the image CPU 84 to the operator workstation 46. The operator workstation 46 includes a CPU 94, a display 96, and an input device 98. The CPU 94 is connected to the communications link 52 and receives inputs, such as user commands, from the input device 98. The input device 98 can be, for example, a keyboard, a mouse, a touch screen panel, and / or a voice recognition system. Through the input device 98 and associated control panel switches, the operator can control the operation of the PET imaging system 12 and the positioning of the patient 21 for scanning. Similarly, the operator can control the display of the resulting images on the display 96 using programs executed by the workstation CPU 94 and can perform image enhancement functions.

[0034] Further, in some examples, the timing accuracy for detecting 511 keV events can be high enough that the coincident detectors 72 are able to measure the time-of-flight (TOF) difference between the two photons. For example, when a positron annihilation event occurs closer to the first detector crystal than the second detector crystal, one annihilation photon can arrive at the first detector crystal before (e.g., nanoseconds or picoseconds before) the other annihilation photon arrives at the second detector crystal. The TOF difference can be used to constrain the location of the positron annihilation event along the LOR, which can improve the accuracy and quality of the images reconstructed by the image reconstruction processor 50. The resolution of the TOF difference or TOF kernel can be a predetermined value stored in the memory of the controller 44 or can be determined based on, for example, the count rate. For example, the same TOF kernel can be used to analyze all LORs in a data set.

[0035] It is noted that the various components and processes of the controller 44 described above are provided as one example of how the controller 44 can obtain, process, and store data generated during operation of the PET imaging system 12. In other examples, the controller 44 can include different processors and memories having similar or different functionality than those described above in similar or different arrangements. In particular, the controller 44 can employ parallel or massively parallel processing. Further, in some embodiments, the various processors of the controller 44, such as the data acquisition processor 48 and the image reconstruction processor 50, can be contained within a shared housing, while in other embodiments, the various processors of the controller 44 can be contained within separate housings located at the same or different locations. Thus, in some examples, the processors of the controller 44 can span multiple locations that are communicatively connected.

[0036] During PET, such as with a PET imaging system (e.g., the PET imaging system 12 of Figure 1 and Figure 2 ) operating a medical imaging facility to image a patient (e.g., the patient 21 of Figure 1 and Figure 2When imaging a patient 21), the patient may move. Motion can cause data blurring, increased noise, reduced quantitative accuracy, and the introduction of image artifacts. As a result, the diagnostic value of the obtained images may be reduced. The technician operating the PET imaging system may not be aware that the patient has moved and may therefore not take steps to account for patient motion. In some examples, the obtained imaging data may undergo extensive post-imaging motion correction, which uses a large amount of computing resources. However, even when post-imaging motion correction is used, the diagnostic value of the image may still be reduced, which may lead to incorrect diagnosis or rescanning of scheduled patients. Even if the patient is immediately available at the medical imaging facility, the previously used radionuclide may be unavailable due to isotope decay. Therefore, the patient may be reinjected with the radionuclide before the repeat scan. If the patient is not immediately available, the patient may have to return to the medical imaging facility for a rescan. Overall, the amount of time before a diagnosis is made, patient discomfort, and imaging costs may all increase.

[0037] therefore, Figure 3 An exemplary method 300 is provided for tracking patient motion within an imaging system during a scan based on real-time reconstructed PET images. Figures 1 to 2 Method 300 is described with reference to the imaging system 10 described above, but other PET imaging systems may also be used. Instructions for performing method 300 and the remaining methods included herein may be provided by a controller (e.g., Figure 1 The controller 25 and Figure 2 The controller 44) is based on instructions stored in the memory of the controller and in combination with the sensors of the imaging system (such as those described above with respect to Figures 1 to 2 The controller may use an actuator of the imaging system to adjust the operation of the imaging system according to the following method.

[0038] At 302, the method 300 includes receiving radiotracer information from a user (e.g., a technologist) of the imaging system. Receiving the radiotracer information from the user includes a type of radiotracer injected into a patient positioned within the imaging system. The radiotracer can be a positron-emitting radionuclide. Some non-limiting examples of radiotracers include fluorine-18 fluorodeoxyglucose (FDG), carbon-11 choline, nitrogen-13 ammonia, and oxygen-15 water. In some examples, the type of radiotracer injected can depend on the anatomical structure of interest being imaged. As described above, the radiotracer injected into the patient can accumulate within organs, blood vessels, etc., and begin to decay and emit positrons. As previously explained, the positrons annihilate, generating a pair of gamma rays. In addition to the type of injected tracer, the controller can receive additional information such as an injection time, a dose of the radiotracer injected, and a pre-injection delay. In addition to the radiotracer information, the controller can receive a weight of the subject. In one example, the user can input the weight of the patient. As another example, the controller can additionally receive a selected imaging protocol, which can be selected by the user or manually input by the user.

[0039] At 304, the method 300 includes performing a CT scan. As one example, performing the CT scan can include first performing a CT scout scan. The CT scout scan can be used as an anatomical reference for the PET / CT scan. In one example, the CT scout scan can be used to define a start position and an end position for the CT acquisition and the PET acquisition. In some examples, the CT scout scan can be a whole-body scan. Once the start position and the end position are defined, the method includes acquiring additional CT image data within a region defined by the start position and the end position. For example, the CT image data can be acquired by activating an x-ray source (e.g., the x-ray source 15 of Figure 1 at specified parameters, which can be input by the user or specified by an imaging protocol selected by the user (e.g., specify kV, mA, attenuation filter position). Further, a gantry (e.g., the gantry 13 of Figure 1 can be rotated to achieve a specified angle. Further, during the CT scan, a position of a table (e.g., the table 24 of Figure 1 of the imaging system can be moved such that the scan progresses from a start scan position to a stop scan position.

[0040] At 306, the method 300 includes performing a PET scan and acquiring emission data from within a field of view (FOV) of the imaging system. The PET scan can generate functional images corresponding to dynamic occurrences, such as metabolism. To perform the PET scan and acquire emission data, the detector crystals of the PET imaging system are activated to detect gamma rays emitted from the patient due to positron emission and annihilation, and the acquisition circuit, event localizer circuit, and coincidence detector can together record coincident events, as described above with respect to Figure 2 .

[0041] At 308, the method 300 includes producing real-time PET images during data acquisition via a fast reconstruction method. As will be described in more detail below with respect to Figures 4 to 6 , in some examples, the fast reconstruction method can not use all of the emission data obtained for each image reconstruction in order to provide real-time (e.g., without significant delay) PET images. Thus, the real-time PET images include a series of live images that represent the emission data being acquired at the time of acquisition (e.g., at the time of occurrence, with only sub-second or second-level time delay). For example, some data can be skipped and / or discarded by the fast reconstruction method. As another example, the real-time fast reconstruction method can additionally or alternatively utilize subsetting to reduce processing time. Further, the fast reconstruction method can not employ attenuation or scatter correction, as the produced images can be used to determine patient position and can not be used for, e.g., diagnosis. Further, the fast reconstruction method can not employ motion correction.

[0042] The fast reconstruction method can use TOF list mode reconstruction (rather than sinogram-based reconstruction). Thus, the controller can not organize the emission data into histograms prior to reconstruction (e.g., can not use the sorter / histogram generator 80 of Figure 2 . List mode data includes a list of all detected coincident events. Each item in the list identifies the two detector crystals involved, the detected time difference between the two detector crystals (e.g., TOF information, as described above with respect to Figure 2 . The controller can evaluate an image value corresponding to an approximate location of the coincidence based on the LOR between the two detector crystals and the time difference at which the event was detected for each item in the list. After a set (e.g., subset or iteration) of events are processed, an update can be applied to the image. The number of image updates applied can vary between one update and multiple updates. For example, increasing the number of image updates can increase the reconstruction time at the cost of providing a statistically best image.

[0043] The fast reconstruction method can reconstruct image volumes for short time frames (e.g., time periods) to produce real-time PET images. For example, each short time frame can include a predefined duration, which can range from milliseconds to seconds. As one non-limiting example, each short time frame is one second. In such an example, the predefined duration can be used to determine (or limit) the event data to include for each time frame. As another example, the duration of each time frame can vary based on the number of events captured. For example, the duration can be adjusted so that a desired number of events (e.g., 4 million) are acquired. The time frame can be extended until the desired number of new events are detected, and then a subsequent time frame can begin. In such an example, the number of events can be used to determine (or limit) the event data to include for each time frame and the duration of each time frame.

[0044] A real-time (e.g., live) PET image can be reconstructed from emission (e.g., event) data obtained during one time frame, and each real-time, live PET image can be referred to herein as an “image frame.” In some examples, the time frames can be consecutive and non-overlapping, while in other examples, the time frames can partially overlap, such that a subsequent time frame begins before a preceding time frame ends. As one example, a first time frame can begin, and upon completion of the first time frame (e.g., a predetermined duration or a desired number of detected events), data acquired during the first time frame can be reconstructed via the fast reconstruction method. While the real-time PET image is produced for the first time frame, data can be collected for a second, subsequent time frame. This sequence can be iteratively repeated throughout the scan to produce a series of real-time PET images.

[0045] As one example, the series of real-time PET images can include image frames reconstructed at predetermined time points, each time point separated by a preselected interval (e.g., a predefined duration), and each image frame reconstructed from emission data acquired during the immediately preceding interval. As such, for example, one image frame can be reconstructed per second. As another example, the series of real-time PET images can include image frames reconstructed after a predetermined amount of data is acquired. As such, for example, one image frame can be reconstructed per n events acquired.

[0046] As a further example, image frames can be reconstructed from partially overlapping data sets. For example, consecutive image frames can share a percentage of events in the range of 30% to 70%, which corresponds to the proportion of total events used to reconstruct each consecutive image frame shared by consecutive image frames. In one non-limiting example, consecutive image frames can share 50% overlap in events. For example, one image frame can be reconstructed from event 1 to event N, where N is a predetermined number, and the next image frame can be reconstructed from event N / 2+1 to event 3N / 2 to achieve 50% overlap of events between consecutive frames. However, the data sets used to reconstruct consecutive image frames can overlap in other ways.

[0047] Alternatively, if the count rate of events is very high, the number of events detected per second can be too many to use all events for real-time reconstruction. As will be detailed below with respect to Figure 4 In such examples, not all detected events can be used to reconstruct image frames in order to maintain real-time performance of the system. For example, if the count rate of events is 10 million events / second and the controller is able to reconstruct 4 million events / second, the controller can reconstruct an image frame using the first 4 million events and resume collecting events at the start of the next second without using events after the first 4 million. However, the controller can continue to store all list events for subsequent (non-real-time) processing. In this way, real-time reconstruction can be maintained regardless of the event count rate.

[0048] Because the time frame is short, the number of coincident events in the emission data obtained during the time frame is relatively small, and by using TOF data, only a small portion of the image is considered for each detected coincidence. Thus, for the same emission data, list mode reconstruction is more efficient than sinogram-based reconstruction.

[0049] Further, real-time PET images can be reconstructed using efficient random number calculations. Random numbers refer to the detection of two photons that meet coincidence criteria that are not actually from the same positron annihilation event (e.g., the coincidences are random). Random numbers are a source of image degradation in PET and can be compensated for during or prior to image reconstruction. As one example, a single map can show the count rate in each detector. For a fast reconstruction method, a simple multiplication can be used instead of calculating random numbers that extend from the single map into a full sinogram, which uses over 400 million calculations. For example, the random rate of LORs between detector crystals R and i (j) can be calculated using a single count rate (SR method) according to the following equation:

[0050] R = 2S i S j T

[0051] where S iis the single count of detector crystal i, S j is the single count of detector crystal j, and τ is the timing window used to detect coincidence. Therefore, the random number calculation can be performed using fewer computational resources. Similar principles can be applied to normalization and dead time. By including random numbers and normalization in real-time PET image reconstruction, image quality can be improved, which can facilitate motion detection as described below.

[0052] At 310, method 300 includes evaluating real-time PET images over time to track patient motion. For example, as each image frame is reconstructed, it can be registered, such as by transforming each image frame into a unified coordinate system. The unified coordinate system can provide a reference for comparing image frames to each other. In real time, each newly reconstructed and registered image frame can be compared with the immediately preceding (e.g., previously reconstructed) image frame in the series (or a predetermined number of preceding image frames) to determine whether the patient has moved between image frames relative to the unified coordinate system. Thus, method 300 can provide real-time motion calculation from image space.

[0053] Registration can be performed using one or more algorithms. In some examples, the registration algorithm can utilize edge detection to define the boundaries of the patient's anatomical structure in each image frame, and can further use change detection to determine whether the boundaries defined by the edge detection have moved over time (e.g., between image frames) relative to the uniform coordinate system, such as by comparing the image frame to the image frame. Figure 7 As detailed in . For example, edge detection can provide rigid registration and can be used when the anatomical feature being imaged is a rigid body (such as a head). However, the registration algorithm can additionally or alternatively utilize other transformations and analysis methods that can compare the patient's position between image frames. For example, the controller can perform non-rigid registration between image frames, such as by using optical flow methods. Optical flow methods can create a complete 3D motion field showing how each element of the volume has moved compared to a reference frame (e.g., an initial time point). Further, for example, the registration algorithm can determine the absolute magnitude of the motion based on the patient's displacement (or displacement of one or more boundaries) between image frames.

[0054] At 312, the method 300 includes determining whether patient motion is detected. For example, patient motion can be detected when an absolute magnitude of motion exceeds a predetermined motion threshold stored in a memory of the controller. The motion threshold can be calibrated to distinguish between smaller changes in patient position and larger changes in patient position. For example, smaller changes can be more accurately compensated using the post-imaging motion correction algorithm than larger changes. As another example, larger changes in patient position can not be accurately corrected via processing or can require more complex processing than smaller changes in patient position. For example, image quality degradation can occur when the absolute magnitude of motion exceeds the motion threshold. Conversely, patient motion having an absolute magnitude below the motion threshold can be corrected without causing image quality degradation. For example, patient motion below the motion threshold can be caused by patient respiration. The magnitude of motion can refer to a change in patient between image frames, such as a displacement of the patient, a rotation of the patient, and / or another feature indicative of a statistically significant mismatch in patient position between image frames.

[0055] As yet another example, the controller can determine a degree of similarity in overall patient position in two image frames, and the controller can indicate that patient motion is detected in response to the degree of similarity falling below a threshold degree of similarity. The threshold degree of similarity can distinguish between smaller changes in patient position and larger changes in patient position as defined above.

[0056] If patient motion is not detected, the method 300 proceeds to 330 and includes reconstructing a CT image. One or more CT images can be reconstructed using, by way of non-limiting example, an analytical reconstruction algorithm such as filtered backprojection or an iterative reconstruction algorithm.

[0057] At 332, the method 300 includes reconstructing a final PET image from the acquired imaging data. In particular, the final PET image can be completed after all emission data has been acquired (e.g., after the PET scan is complete). Thus, the final PET image is not a real-time representation of the acquired emission data (e.g., the final PET image is a non-live PET image). In some examples, the final PET image can be reconstructed using emission data spanning the entire duration of data acquisition. In other examples, such as the examples to be further pointed out herein, a portion of the data can be selectively excluded. In some examples, the final PET image can include one cross-sectional image of the patient. In other examples, the final PET image can include multiple cross-sectional images of the patient. Further, because a large number of coincidence events can be used to reconstruct the final PET image relative to the number of coincidence events used to reconstruct the real-time PET image, a sinogram-based reconstruction can be performed. For example, the sinogram-based reconstruction can more efficiently process larger data sets than the list-mode reconstruction described above with respect to the real-time PET image. Thus, the sinogram-based reconstruction will be described below. However, in other examples, the list-mode reconstruction can be used to reconstruct the final PET image.

[0058] As previously explained with reference to FIG. 1, the image reconstruction processor of the PET imaging system can include a sorter / histogram generator (such as the sorter / histogram generator 80 of FIG. 1). Figure 2 As previously explained with reference to FIG. 1, the image reconstruction processor of the PET imaging system can include a sorter / histogram generator (such as the sorter / histogram generator 80 of FIG. 1). Figure 2 The sorter / histogram generator includes a histogram of a large number of bins, where each bin corresponds to a unique pair of detector crystals. Each bin of the histogram also stores a count value representative of the number of coincidence events detected by the pair of detector crystals of that bin during the scan, which can be organized into a sinogram. After all data has been acquired and the PET scan is no longer actively performed (e.g., the detector crystals are not actively detecting events), the data in the histogram is used to reconstruct a final PET image of the patient. For example, the total number of coincidence events detected by the pair of detector crystals will be proportional to the amount of radiotracer within the LOR connecting the pair. The final PET image can be used by a clinician for diagnostic purposes, while the real-time PET image can not be used for diagnosis.

[0059] The controller can use an analytical and / or iterative image reconstruction algorithm. Analytical reconstruction algorithms can provide a direct mathematical solution for the reconstructed image, while iterative image reconstruction algorithms can use multiple mathematical iterations to obtain the reconstructed image. Further, the controller can employ attenuation and / or scatter correction in reconstructing the final PET image. For example, various scatter correction methods, such as model-based scatter simulation, can be used to estimate scatter events during PET image reconstruction. Model-based scatter simulation uses knowledge of the emission activity and attenuation coefficients, and can include both single scatter and multiple scatter estimation. The emission activity can be estimated by an initial PET image reconstruction using the PET data acquired within the FOV.

[0060] At 334, the method 300 includes displaying one or more of the final PET image and the CT image. For example, the one or more images can be displayed on a display screen, such as the display 96 of the system 10. As described above, the CT image can define the anatomy, while the PET image can show dynamic body function, such as metabolism. As also described above, the radioactive tracer injected into the subject can accumulate in organs. Thus, increased uptake of the radioactive tracer in an organ can appear as a "hot spot" in the PET image. Abnormal tissue (or tumors) can have increased uptake, and thus appear as hot spots in the PET image. However, normal tissue also takes up different levels of the radioactive tracer. For example, radioactive tracers such as FDG are primarily cleared through the renal system, so a normal bladder can have the maximum amount of FDG uptake. For example, the brain can have a higher amount of FDG uptake than fatty tissue. The radioactive tracer uptake of normal tissue can be physiologic, while the radioactive tracer uptake of abnormal tissue can be pathologic. Thus, the final PET image can provide functional information to aid in diagnosis. Figure 2

[0061] In some examples, the CT image and the final PET image can be overlaid (e.g., via co-registration) in order to put the functional information from the PET image into the anatomical context given by the CT image. These views can allow the clinician to correlate and interpret information from two different imaging modalities on one image, which can result in more precise information and more accurate diagnosis. The method 300 can then end.

[0062] Returning to 312, if patient motion is instead detected, such as in response to an absolute amount of motion exceeding a threshold motion, the method 300 proceeds to 314 and includes executing a motion detection response. For example, the controller can indicate that patient motion is detected via an internal condition that initiates the motion detection response and / or indicate an alert to the user. The controller can select one or more motion detection responses from a plurality of possible motion detection responses, and execute the selected responses simultaneously or sequentially. Further, the controller can identify that the patient motion (or change) exceeds the threshold ssin a sequential timeframe in order to apply the selected motion detection responses accordingly.​

[0063] Performing a motion detection response optionally includes alerting the user of patient motion, as indicated at 316. For example, the controller can output an audible and / or visual alert to the user via a workstation, such as the operator workstation 46 of FIG. 1. As one example, the alert can inform the user that patient motion was detected, and can further prompt the user to instruct the patient to remain still. Figure 2

[0064] Performing a motion detection response also optionally includes removing data acquired during detected patient motion from the data set to be used in the final PET image reconstruction, as indicated at 318. For example, data acquired while the patient motion is greater than a threshold value can be separated from data acquired while the patient motion is not greater than the threshold value, and only data acquired while the patient motion is not greater than the threshold value can be used to reconstruct the final PET image. In addition to or as an alternative to outputting an alert at 316, removing data acquired during detected patient motion can be performed. By separating data acquired during detected motion from the data set to be used in the final PET image reconstruction, data acquired during detected patient motion can be excluded from the final PET image. Thus, motion artifacts caused by detected patient motion will not be present in the final PET image.

[0065] Performing a motion detection response optionally includes extending the acquisition time of the scan to obtain a desired amount of motion-free data, as indicated at 320. For example, the desired amount of motion-free data can be a predetermined count number or data acquisition duration in which the patient motion remains below a threshold value. The predetermined count number or data acquisition duration can be a calibration value stored in the memory of the controller, and can represent a minimum amount of desired data for achieving a final PET image with reduced noise and image artifacts. In addition to or as an alternative to one or both of outputting an alert at 316 and removing data acquired during detected motion at 318, extending the acquisition time of the scan can be performed.

[0066] ​The performing motion detection response optionally includes prompting for use of a motion correction reconstruction technique for data acquired during the detected patient motion, as indicated at 322. The motion correction technique can not be automatically performed because it can be computationally expensive. Thus, in at least some examples, the motion correction technique can be selectively applied to data acquired while the patient motion is greater than a threshold, rather than the entire data set. As another example, the motion correction technique can be selectively applied to data acquired while patient motion is indicated and data acquired after the indicated patient motion. As an illustrative example, patient motion can cause the patient to move from a first posture to a second posture. Although the patient can not be actively moving while in the second posture (e.g., the patient motion remains less than a threshold while in the second posture), all events acquired while the patient is in the second posture can be corrected in order to compensate for the difference between the first posture and the second posture. Moreover, in some examples, the emission data can be separated into "pre-motion," "during-motion," and "post-motion" data sets, which can be processed at least partially separately from one another in order to accurately correct for patient motion while improving computational efficiency. In addition to or as an alternative to any or all of outputting an alert at 316, removing data acquired during the detected motion at 318, and extending the acquisition time of the scan at 320, the prompting for use of a motion correction reconstruction technique can be performed. For example, by performing the motion correction technique, data acquired during the detected patient motion can still be used for final PET image reconstruction without introducing blurring and motion artifacts.

[0067] The performing motion detection response optionally includes displaying a real-time motion plot, as indicated at 324. For example, the real-time motion plot can show the displacement of the patient within the FOV in three spatial directions over time, which can be displayed to a user via a display screen. Because the user can not be able to see the patient motion within the gantry, the real-time motion plot can allow the user to more easily monitor the patient's motion during the scan. In addition to or as an alternative to any or all of outputting an alert at 316, removing data acquired during the detected motion at 318, extending the acquisition time of the scan at 320, and prompting for use of a motion correction reconstruction technique at 322, the displaying a real-time motion plot can be performed.

[0068] Performing motion detection response optionally includes repeating the CT scan of the patient after motion has been detected, as shown at 326. For example, a new CT image data can be acquired by activating the x-ray source while the gantry rotates to achieve the angle specified by the imaging protocol. The new CT image data can replace the earlier CT image data, which can be discarded or temporarily stored separately from the new CT image data until the CT image is reconstructed. In addition to or as an alternative to outputting an alert at 316, removing data acquired during detected motion at 318, extending the acquisition time of the scan at 320, prompting the use of motion correction reconstruction techniques at 322, and displaying a real-time motion graph at 324, a repeat CT scan can be performed.

[0069] While motion detection during a PET scan can not be directly related to patient motion during a CT scan, repeating the CT scan can increase the likelihood of obtaining CT scan data that is motion free and properly registered. However, in other hybrid imaging modalities that simultaneously acquire data of both modalities, such as in PET / MRI, the detected motion affects both imaging modalities. Thus, the method at 326 can include selectively repeating the MR scan portion performed during the detected patient motion (and, for example, not repeating the entire MR scan).

[0070] Performing motion detection response optionally includes adjusting both PET and CT data acquisition based on the detected patient motion, as shown at 328. For example, the PET and CT data acquisition can be extended in order to capture additional motion free data. Extending the PET and CT data acquisition when patient motion is detected (rather than when patient motion is not detected) can enable more efficient use of the imaging system resources while also improving the image quality of both the final PET image and the CT image. In addition to or as an alternative to any one or all of outputting an alert at 316, removing data acquired during detected motion at 318, extending the acquisition time of the scan at 320, prompting the use of motion correction reconstruction techniques at 322, displaying a real-time motion graph at 324, and repeating the CT scan at 326, both PET and CT data acquisition can be adjusted based on the detected patient motion.

[0071] The method 300 proceeds to 330 and includes reconstructing the CT image, as described above. Thus, the CT image can be reconstructed after additional data is acquired and / or compensation for the detected patient motion is performed. Thus, the CT image can have fewer motion artifacts and can improve the image quality.

[0072] The final PET image reconstructed at 332 may also have improved image quality compared to not performing the selected motion detection response. For example, data acquired during periods of patient motion may not be used to reconstruct the final PET image (e.g., when selected at 318), thereby reducing blur and noise in the final PET image while improving the quantitative accuracy of the final PET image. As another example, motion correction reconstruction techniques may be applied to data acquired during periods of patient motion (e.g., when selected at 322), and the corrected data acquired during periods of patient motion may be used to reconstruct the final PET image without introducing blur and noise.

[0073] By using fast reconstruction methods to track patient motion in real time (or near real time), various adjustments and adjustments to the imaging protocol can be made in real time to compensate for patient motion and / or exclude data acquired during periods of patient motion without adversely affecting image quality or scanner resources. This allows for the acquisition of PET images with fewer motion artifacts, thereby improving both the quality and diagnostic accuracy of PET images. Furthermore, CT images can be reconstructed from data acquired while the patient was not moving, thereby improving both the quality and accuracy of CT images. Overall, the frequency of patient rescans can be reduced, while the quantitative accuracy of the final PET images can be improved, thereby reducing imaging costs and diagnostic time.

[0074] continue Figure 4 , shows the Figure 3 Example method 400 of the fast reconstruction method described in . For example, method 400 can be used as Figure 3 A portion of method 300 (e.g., at 308) is performed by the controller in real time during PET data acquisition. At least a portion of method 400 can be performed as one iteration of image update for a single image frame, and at least in some examples, multiple iterations can be performed in parallel for a single image frame.

[0075] At 402, method 400 includes determining events for each subset of imaging data. For each short time frame (e.g., image frame) used to reconstruct a real-time PET image, as described above at 308, subsetting can be used to more quickly reach convergence. Instead of iterating over all events acquired during the time frame for each update of the image frame, every nth event (where n is an integer) can be used, and n passes through the data can be made using a different subset of events for each pass. For example, each pass can include one update.

[0076] Accordingly, in some examples, the controller can determine a value of n for determining events for each subset. The value of n can be determined based on the data density and the matrix size of the data set, for example, so as to maintain at least a lower threshold number of events in each subset. The lower threshold number of events can be a predetermined value stored in memory that corresponds to a number of events below which there is not enough data to produce an accurate image update. The controller can input the data density and the matrix size into a lookup table or algorithm that can output a value of n for a given data set. The controller can then determine events for each subset of imaging data based on the determined value of n, such as by creating a lookup table of events for a given subset. As an illustrative example, when n is 4, a first iteration can use events 1, 5, 9, 13, etc.; a second iteration can use events 2, 6, 10, 14, etc.; a third iteration can use events 3, 7, 11, 15, etc.; and a fourth iteration can use events 4, 8, 12, 16, etc. In this way, a more converged image can be reached more quickly, with four updates to the image after processing each event only once, as will be further detailed below. However, in some examples, as the time frame decreases, the number of events is fewer, and the data set cannot be divided into subsets (e.g., n equals 1).

[0077] In some examples, when the number of events captured in each time frame is above an upper threshold number of events, the controller can select to skip events so as to reduce the processing time. The upper threshold number of events can be a pre-calibrated value stored in memory that corresponds to a number of events for which the expected processing time is greater than a predetermined threshold duration for real-time processing. In some examples, the controller can adjust (e.g., increase) the value of n determined above, and then only process a portion of the subsets. Continuing the illustrative example above, where n is 4, n can be increased to 5, and only 4 of the 5 subsets can be processed, for example, resulting in a 20% reduction in processing time. As another example, additionally or alternatively, the total number of events in the data set can be reduced prior to determining events for each subset. As an illustrative example, only the first 4 million events can be selected from 5 million total events, which can cause a slight temporal bias to the results.

[0078] In some examples, the controller can update the lower threshold number of events and / or the upper threshold number of events by measuring the actual time taken to process a given size of data set. In this way, the controller can refine the fast reconstruction method at implementation time so as to ensure that real-time PET images are generated that provide accurate and timely patient motion detection.

[0079] At 404, the method 400 includes assigning each event to a subset index. For example, events in a single subset can be divided into groups or blocks of events that can be stored in contiguous memory. The contiguous memory can undergo parallel processing, where each parallel thread is assigned to process one block of events. For example, each parallel thread can have access to only one block of a given subset of imaging data stored therein, reducing the amount of data processed by each thread. Further, at the beginning of each image update (e.g., each iteration), each thread gets its own copy of the update matrix, such that the threads do not interfere with each other. When the threads have all completed their respective processing tasks, outlined below (e.g., from 406 to 410), their matrices are joined together before being applied to a single image update associated with the subset, which can be stored in contiguous memory or non-contiguous memory.

[0080] At 406, the method 400 includes, for each event, determining a start position and a stop position of a projection of a short time-of-flight (TOF) kernel within a line-of-response (LOR). For example, using the parallel processing described above, the controller can find the start position and stop position of each coincident event in a given subset of data that travels through the imaging FOV. As described above with respect to Figure 2 The TOF kernel can be standardized across the data set, such that the same TOF kernel is used for each LOR. Thus, the TOF kernel can define the shortest effective segment to project for a given LOR. As another example, using a longer segment than the TOF kernel can result in increased processing time. Thus, the controller can use the position of the TOF kernel on the LOR to determine the start position and stop position of each projection, which can be further determined by the timing difference between each photon of the detection event.

[0081] Briefly turning to Figure 5 , the diagram 500 illustrates an illustrative example of the manner in which the start position and stop position of a projection are determined for different lines of response. Figure 5 A detector ring assembly 540 of a PET imaging system is shown, which includes a plurality of detectors 562. For example, the detector ring assembly 540 can be the detector ring assembly 40 of Figure 2 , and the plurality of detectors 562 can be the plurality of detectors 62 of Figure 2 . The diagram 500 also shows a PET imaging field of view (FOV) 522. For example, the FOV 522 can include a patient positioned within the detector ring assembly 540.

[0082] Three lines of response are shown in the diagram 500: a first line A, a second line B, and a third line C. The line A does not intersect the FOV 522. Thus, no projection of the line A is determined. The lines B and C each intersect the FOV 522, and it is determined that the detected events for each line are coincident (e.g., as described above with respect to Figure 2522). Four points of line B are calculated, including b1, b2, b3, and b4. Point b1 is the location where line B enters FOV 522, point b2 is the location where the TOF kernel of line B begins, point b3 is the location where the TOF kernel of line B ends, and point b4 is the location where line B leaves FOV 522. Point b2 is determined as the starting location, and point b3 is determined as the stopping location of the projection of line B so as to encompass the TOF kernel of line B.

[0083] Four points are also calculated for line C, including c1, c2, c3, and c4. Point c1 is the location where the TOF kernel of line C begins, point c2 is the location where line C enters FOV 522, point c3 is the location where the TOF kernel of line C ends, and point c4 is the location where line C leaves FOV 522. Point c2 is determined as the starting location, and point c3 is determined as the stopping location of the projection of line C so as to encompass the TOF kernel of line C.

[0084] return Figure 4 At 408, method 400 includes calculating, for each event, projection coefficients (weights) based on the path length and TOF kernel height of each voxel traversed by the TOF kernel. For example, the controller can calculate a fast Siddon projection based on voxel identification and weights (e.g., derived from the path length and TOF kernel height) in a sparse matrix multiplication. For example, the path length and TOF kernel height at the center of each segment can be determined for each voxel boundary intersection. For example, the controller can calculate the projection of each event in the data acquired during a given time frame.

[0085] Briefly go to Figure 6 , schematic 600 shows an illustrative example of determining the effective projector path length and TOF kernel height. Although schematic 600 shows a two-dimensional representation, it should be noted that the actual calculation can be three-dimensional. Schematic 600 shows a plurality of voxels 602, a direction vector n (which can be part of the LOR, such as a start position 606 and a stop position 608). The start position 606 and the stop position 608 correspond to the above description of Figure 4 and Figure 5 The starting and stopping positions of the TOF kernel are described (eg, determined at 406). Thus, the portion of the direction vector n between the starting position 606 and the stopping position 608 corresponds to a portion of the TOF kernel that will be used in the projection of the event.

[0086] Factor a is defined as the distance from the center of the TOF kernel along the direction vector n. Thus, the center of the TOF kernel is at a = 0. For each step in x (e.g., dx, shown from left to right on diagram 600), a changes by dx / n[0]. A "next a" lookup can be maintained to cross voxel boundaries. The smallest step in a that hits a voxel boundary is used to determine path length and TOF kernel height. For example, traveling from start position 606 to stop position 608 along direction vector n, there are six voxel boundary crossings, resulting in seven segments L1, L2, L3, L4, L5, L6, and L7 of different lengths. The length of each segment can be determined based on the change in a. Each segment has a corresponding TOF kernel height. As shown, segment L1 has TOF kernel height H1, segment L2 has TOF kernel height H2, segment L3 has TOF kernel height H3, segment L4 has TOF kernel height H4, segment L5 has TOF kernel height H5, segment L6 has TOF kernel height H6, and segment L7 has TOF kernel height H7. The length of each segment multiplied by the corresponding TOF kernel height can be stored along with the voxel ID of each voxel traversed by direction vector n between start position 606 and stop position 608. The projection 610 can then be determined from sparse matrix multiplication of the stored values.

[0087] Returning to Figure 4 At 410, method 400 includes, for each event, performing a forward projection based on the projection coefficients (e.g., determined at 408 above), applying corrections, and performing a back projection based on the projection coefficients. As one example, within each subset, each event can be forward projected, adjusted for corrections, and then back projected. The forward and back projections can be performed by accessing sparse matrix elements, e.g., with each processing thread reconstructing one image to reduce or avoid memory locking. Further, as described above with respect to 308 of Figure 3 The reconstruction algorithm can not employ attenuation or scatter corrections, and can use efficient random number calculations, as described above with respect to 308 of

[0088] At 412, method 400 includes combining the back projections of each event within a given subset. For example, the back projections from all events within a subset can be summed, and this can be used to generate an image update. Thus, the data from each subset can provide one image update.

[0089] At 414, the method 400 includes iteratively updating the image based on each subset. For example, the controller can use an iterative reconstruction algorithm that updates an image estimate until a desired solution is achieved. For example, the desired solution can be a maximum likelihood solution. As described above, each subset can provide one image update. An initial blank (e.g., uniform) image can be updated using a first image update generated from a first subset of emission data acquired during a given time frame (e.g., a first iteration). The resulting updated image can be further updated using a second image update (e.g., a second iteration) generated from a second subset of emission data. This process can be repeated with image updates from each subset of events such that the image estimate can converge (e.g., reach the desired solution). The final updated image includes a real-time PET image for one time frame. The method 400 can then return. For example, the method 400 can be repeated for each subsequent time frame such that subsequent image frames can be produced via Figure 3 the method of

[0090] Next, Figure 7 An example implementation 700 of patient motion detection during PET based on real-time reconstructed PET image frames is shown. For example, a controller (e.g., the controller 25 of Figure 1 and / or the controller 44 of Figure 2 may use real-time PET images reconstructed with TOF list mode reconstruction without attenuation and scatter correction reconstruction, such as according to the fast reconstruction method of Figure 4 the method of Figure 3 detect patient motion during PET acquisition. Note that the implementation 700 is one illustrative example of how PET image frames can be analyzed to determine patient motion, and in other examples, the controller can perform other analysis.

[0091] The implementation 700 shows a series of PET image frames reconstructed from emission data acquired over time, including a first image frame 702, a second image frame 704, and a third image frame 706. Each of the first image frame 702, the second image frame 704, and the third image frame 706 is reconstructed from data acquired over a short duration (e.g., 1 second per image frame), as described above with respect to Figure 3 and Figure 4As described, each image frame shows a side profile of the patient's skull. The first image frame 702 is the earliest frame, and the third image frame 706 is the latest time frame, as shown in the time axis 701. Specifically, the first image frame 702 is the first image frame in the series (e.g., frame 1), and the second image frame 704 is the next image frame in the series (e.g., frame 2, immediately after frame 1, without any other image frames in between). The third image frame 706 is a certain number of image frames (e.g., frame n) after the second image frame 704.

[0092] In the exemplary implementation 700 , as each image frame is reconstructed, the controller performs registration to transform the image frames into a unified coordinate system (shown as a grid) and uses edge detection to define the boundaries of the patient's anatomy in each image frame. Figure 7 A first boundary line 708 is shown at the top of the patient's skull in the first image frame 702, a second boundary line 710 is shown at the top of the patient's skull in the second image frame 704, and a third boundary line 712 is shown at the top of the patient's skull in the third image frame 706, although other boundary lines may additionally or alternatively be used to determine and track patient motion between image frames. Furthermore, for comparison, the first boundary line 708 is shown as a dashed overlay on the second image frame 704, and both the first boundary line 708 and the second boundary line 710 are shown as dashed overlays on the third image frame 706.

[0093] Once at least two image frames are acquired (e.g., Figure 7 In the example shown, the controller compares the positions of corresponding boundary lines between the two image frames to determine the displacement between them. In the example shown, the patient's skull has shifted between the first image frame 702 and the second image frame 704, as shown by the position of the second boundary line 710 relative to the first boundary line 708 on the second image frame 704. The controller can directly determine the magnitude of the displacement based on the difference in the position of the second boundary line 710 relative to the first boundary line 708, and indicate patient movement in response to the magnitude exceeding a threshold, as described above. Figure 3 As described in 312 of .

[0094] In some examples, the boundaries of non-consecutive image frames in the series can also be compared to track patient motion over time. In the illustrated example, the third boundary line 712 is compared to both the second boundary line 710 and the first boundary line 708, even though one or more image frames were acquired between the second image frame 704 and the third image frame 706. The position of the third boundary line 712 relative to the second boundary line 710 shows that the patient’s skull has shifted in the same direction between the second image frame 704 and the third image frame 706 as between the first image frame 702 and the second image frame 704 (e.g., in a downward direction relative to the page), although it will be appreciated that image frames after the second image frame 704 and before the third image frame 706 can show patient motion in other directions. For example, the patient can move in an upward direction (relative to the page) before moving downward again and reaching the position shown in the third frame 706.

[0095] The controller can determine a displacement between the third boundary line 712 and one or both of the first boundary line 708 and the second boundary line 710. The magnitude of the displacement between the third boundary line 712 and the second boundary line 710 is less than the magnitude of the displacement between the first boundary line 708 and the second boundary line 710. Further, the displacement between the first boundary line 708 and the third boundary line 712 is greatest. As one example, the series illustrated in the example implementation can indicate that the patient is consistently moving in a downward direction according to any displacement shown in the intervening image frames between the second image frame 704 and the third image frame 706. In some examples, the controller can generate a plot of the displacement (or change in boundary position) between consecutive image frames to track patient motion in magnitude and direction over time. Further, in some examples, the plot can be displayed to an operator in real-time (e.g., at 324 of FIG. 3). Figure 3

[0096] ​In this way, patient motion can be tracked in real-time so as to identify emission data acquired during movement (e.g., when the magnitude of patient motion exceeds a threshold). Accordingly, various adjustments and adaptations to the imaging protocol can be made in real-time so as to compensate for patient motion and / or exclude data acquired during patient motion without adversely affecting image quality or scanner resources. By tracking patient motion in real-time, diagnostic PET images with fewer motion artifacts can be generated, thereby improving the quality of the diagnostic PET images and the diagnostic accuracy of the diagnostic PET images. Further, by tracking patient motion with PET image frames reconstructed using a fast reconstruction method that uses list-mode TOF data and does not employ scatter correction, attenuation correction, and motion correction, the PET image frames can be reconstructed in real-time (e.g., as the data is acquired), while avoiding memory lock and reducing processing time. Overall, the incidence of patient rescans can be reduced, while quantitative accuracy of the final PET images can be improved, thereby reducing imaging costs and diagnostic time.

[0097] The technical effect of computing a series of PET image frames from image space using data acquired over a short duration is that the series of PET image frames can be reconstructed in real-time, thereby enabling real-time tracking of patient motion.

[0098] One example provides a method for a medical imaging system, the method comprising: acquiring emission data during a positron emission tomography (PET) scan of a patient; reconstructing a series of live PET images while the emission data is being acquired; and tracking motion of the patient during the acquisition based on the series of live PET images.

[0099] In one example, reconstructing the series of live PET images while the emission data is being acquired comprises performing list-mode reconstruction in real-time.

[0100] In an example, each live PET image in the series of live PET images is reconstructed from emission data acquired during a defined duration. In some examples, for each live PET image in the series of live PET images, reconstructing the series of live PET images while the emission data is being acquired includes: dividing the emission data acquired during the defined duration into subsets; reconstructing an image iteration from each subset of emission data; and combining the image iterations from each subset to produce the live PET image. In one example, reconstructing an image iteration from each subset of emission data includes assigning emission data within a given subset to a plurality of groups and processing each group in parallel. In another example, reconstructing an image iteration from each subset of emission data includes: determining a start position and a stop position of a TOF kernel within each line of response (LOR) in a given subset of emission data; calculating a projection coefficient based on a path length and a kernel height of each voxel traversed by the TOF kernel; and reconstructing an image iteration by forward projecting and back projecting using the calculated projection coefficients.

[0101] In an example, tracking motion of a patient during a PET scan based on the series of live PET images includes: performing image registration on each live PET image in the series of live PET images; determining a change in the patient between time points in the series of live PET images; and in response to the change exceeding a threshold, performing a motion response. As one example, performing the motion response includes at least one of: outputting an alert; displaying a plot of the motion of the patient; extending an acquisition time for acquiring emission data; separating emission data acquired while displaced greater than the threshold from emission data acquired while displaced less than the threshold; and prompting use of a motion correction reconstruction technique for emission data acquired while displaced greater than the threshold.

[0102] The method can further include reconstructing a non-live PET image after the emission data is acquired, wherein reconstructing the non-live PET image includes performing scatter correction and attenuation correction, and reconstructing the series of live PET images includes not performing scatter correction and attenuation correction.

[0103] The method can further include reconstructing a non-live PET image after the emission data is acquired, wherein reconstructing the non-live PET image includes performing motion correction, and reconstructing the series of live PET images includes not performing motion correction.

[0104] An example method for positron emission tomography (PET) includes: reconstructing image frames using emission data acquired in real-time while performing an emission scan of a patient; indicating patient motion in response to a displacement of the patient between image frames exceeding a threshold; and in response to the patient motion, performing one or more motion detection responses.

[0105] In one example, performing one or more motion detection responses includes discarding emission data acquired during the indicated patient motion; and not using the discarded emission data during reconstruction of a final PET image of the patient.

[0106] In one example, performing one or more motion detection responses includes performing motion correction during reconstruction of a final PET image of the patient only on emission data acquired during and after the indicated patient motion.

[0107] In one example, performing one or more motion detection responses includes extending an acquisition time for performing an emission scan of the patient.

[0108] In an example, each image frame is reconstructed from emission data acquired during a determined time period of the emission scan, and for each image frame, reconstructing the image frame using real-time acquired emission data while performing the emission scan of the patient includes: determining projections from the emission data acquired during the determined time period as soon as the determined time period is completed; and reconstructing the image frame from the determined projections.

[0109] An example system includes a detector array configured to acquire emission data during a scan of a subject; and a processor operatively coupled to the detector array that stores executable instructions in a non-transitory memory that, when executed, cause the processor to: track motion of the subject in real-time during the scan based on the acquired emission data; and adjust a parameter of the scan in response to the motion of the subject exceeding a threshold. In an example, to track the motion of the subject in real-time during the scan based on the acquired emission data, the processor includes additional executable instructions in the non-transitory memory that, when executed, cause the processor to: reconstruct images of the subject at predetermined time points, each time point separated by a defined interval, and each image reconstructed from emission data acquired during an immediately preceding interval; and compare a current image of the subject to a previous image to determine a magnitude of motion of the subject between the previous image and the current image. In one example, to reconstruct the images of the subject, the processor includes additional executable instructions in the non-transitory memory that, when executed, cause the processor to: perform parallel processing on a subset of the emission data determined from the emission data acquired during the immediately preceding interval; determine projections from the subset of the emission data; and reconstruct the current image from the determined projections.

[0110] In one example, to adjust the parameter of the scan in response to the motion of the subject exceeding the threshold, the processor includes additional executable instructions in the non-transitory memory that, when executed, cause the processor to: extend a duration of the scan; and output a motion detection alert.

[0111] In one example, to adjust parameters of a scan in response to motion of the subject exceeding a threshold, the processor includes additional executable instructions in non-volatile memory that, when executed, cause the processor to: separate emission data acquired while the subject's motion exceeds the threshold from emission data acquired while the subject's motion does not exceed the threshold; and reconstruct a diagnostic image of the subject using only the emission data acquired while the subject's motion does not exceed the threshold.

[0112] As used herein, elements or steps listed in the singular and beginning with the word "one" or "a kind of" should be understood as not excluding a plurality of said elements or steps, unless such exclusion is explicitly stated. In addition, reference to "one embodiment" of the present invention is not intended to be interpreted as excluding the existence of additional embodiments that also include the cited features. In addition, unless explicitly stated otherwise, embodiments that "comprise", "include" or "have" an element or multiple elements with a particular characteristic may include additional such elements that do not have that characteristic. The terms "comprise" and "in..." are used as the concise language equivalents of the corresponding terms "comprise" and "wherein". In addition, the terms "first", "second" and "third" etc. are only used as marks, and are not intended to impose numerical requirements or specific positional order on their objects.

[0113] This written description uses examples to disclose the invention, including the best mode, and also to enable one of ordinary skill in the relevant art to practice the invention, including making and using any devices or systems and performing any included methods. The patentable scope of the invention is defined by the claims and may include other examples that occur to one of ordinary skill in the art. Such other examples are intended to be within the scope of the claims if they have structural elements that do not differ from the literal language of the claims, or if they include equivalent structural elements with insignificant differences from the literal language of the claims.

Claims

1. A method for a medical imaging system, the method comprising: collecting emission data during a positron emission tomography scan of a patient; reconstructing a series of live PET images while acquiring the emission data, wherein reconstructing the series of live PET images includes not performing scatter correction and attenuation correction; tracking the patient's motion during the acquisition based on the series of live PET images; and A non-live PET image is reconstructed after acquiring the emission data, wherein reconstructing the non-live PET image includes performing scatter correction and attenuation correction. 2 . The method of claim 1 , wherein reconstructing the series of live PET images while acquiring the emission data comprises performing list mode reconstruction in real time. 3 . The method of claim 1 , wherein each live PET image in the series of live PET images is reconstructed from emission data acquired during a defined time duration.

4. The method of claim 3 , wherein for each live PET image in the series of live PET images, reconstructing the series of live PET images while acquiring the emission data comprises: dividing the emission data acquired during the defined time duration into subsets; reconstructing an image iteratively from each subset of the emission data; as well as The images from each subset are combined iteratively to produce the live PET image. 5 . The method of claim 4 , wherein reconstructing the image from each subset of the emission data iteratively comprises assigning the emission data within a given subset to a plurality of groups and processing each group in parallel.

6. The method of claim 4, wherein reconstructing the image from each subset of the emission data iteratively comprises: determining a start position and a stop position of a TOF kernel within each line of response for a given subset of the transmission data; calculating projection coefficients based on a path length and a kernel height for each voxel traversed by the TOF kernel; as well as The image is reconstructed iteratively by performing forward projection and back projection using the calculated projection coefficients.

7. The method of claim 1 , wherein tracking the motion of the patient during the PET scan based on the series of live PET images comprises: performing image registration on each live PET image in the series of live PET images; determining changes in the patient between time points in the series of live PET images; as well as In response to the change exceeding a threshold, a movement response is performed.

8. The method of claim 7, wherein performing the motion response comprises at least one of: outputting an alert; displaying a graph of the motion of the patient; extending an acquisition time for acquiring the emission data; separating emission data acquired while the displacement is greater than the threshold from emission data acquired while the displacement is less than the threshold; And for the emission data acquired while the displacement is greater than the threshold, a motion correction reconstruction technique is prompted to be used.

9. The method of claim 1 , further comprising reconstructing a non-live PET image after acquiring the emission data, wherein reconstructing the non-live PET image comprises performing motion correction, and wherein reconstructing the series of live PET images comprises not performing the motion correction.

10. A method for positron emission tomography, the method comprising: while performing an emission scan of the patient, reconstructing image frames using emission data acquired in real time, wherein reconstructing the image frames using emission data acquired in real time includes not performing scatter correction and attenuation correction; In response to displacement of the patient between image frames exceeding a threshold value indicating patient motion; executing one or more motion detection responses in response to the patient motion; Scatter correction and attenuation correction are performed during reconstruction of a final PET image of the patient.

11. The method of claim 10, wherein performing one or more motion detection responses comprises: separating emission data acquired during the indicated patient motion; as well as The separated emission data is not used during reconstruction of the final PET image of the patient.

12. The method of claim 10, wherein performing one or more motion detection responses comprises performing motion correction during reconstruction of a final PET image of the patient only on emission data acquired during and after the indicated patient motion.

13. The method of claim 10, wherein performing one or more motion detection responses comprises extending an acquisition time used to perform the emission scan of the patient.

14. The method of claim 10 , wherein each image frame is reconstructed from emission data acquired during a determined time period of the emission scan, and for each image frame, reconstructing the image frame using the emission data acquired in real time while performing the emission scan of the patient comprises: determining a projection from the emission data acquired during the determined time period whenever the determined time period is completed; as well as The image frame is reconstructed from the determined projections.

15. A system for positron emission tomography, comprising: a detector array configured to acquire emission data during a scan of a subject; and a processor operatively coupled to the detector array storing executable instructions in a non-transitory memory, the executable instructions, when executed, causing the processor to: tracking the motion of the subject in real time during the scan based on the acquired emission data by reconstructing images of the subject at predetermined points in time without scatter correction and without attenuation correction; adjusting a parameter of the scan in response to the motion of the subject exceeding a threshold; as well as After the scanning of the subject, a diagnostic image of the subject is reconstructed with scatter correction and attenuation correction.

16. The system of claim 15 , wherein to track the motion of the subject in real time during the scan based on the acquired emission data, the processor includes additional executable instructions in non-transitory memory that, when executed, cause the processor to: reconstructing images of the subject at predetermined time points, each time point being separated by a defined interval, and each image being reconstructed from emission data acquired during an immediately preceding interval; and A current image of the subject is compared to a previous image to determine a magnitude of the motion of the subject between the previous image and the current image.

17. The system of claim 16, wherein to reconstruct the image of the subject, the processor includes additional executable instructions in non-transitory memory that, when executed, cause the processor to: performing parallel processing on a subset of emission data determined from said emission data acquired during said immediately preceding interval; determining a projection from the subsetted emission data; and The current image is reconstructed from the determined projections.

18. The system of claim 15 , wherein to adjust the parameters of the scan in response to the motion of the subject exceeding the threshold, the processor includes additional executable instructions in non-transitory memory that, when executed, cause the processor to: extending the duration of the scan; and Output motion detection alert.

19. The system of claim 15, wherein to adjust the parameters of the scan in response to the motion of the subject exceeding the threshold, the processor includes additional executable instructions in non-transitory memory that, when executed, cause the processor to: separating emission data acquired while the motion of the subject exceeds the threshold from emission data acquired while the motion of the subject does not exceed the threshold; and A diagnostic image of the subject is reconstructed using only the emission data acquired while the motion of the subject does not exceed the threshold.

Images