Medical image processing apparatus and medical image processing method

Abstract

To improve image quality.SOLUTION: A medical image processing apparatus according to an embodiment comprises an acquisition unit, a first creation unit, a reconstruction unit, and a second creation unit. The acquisition unit acquires a plurality of dynamic sinograms, the dynamic sinograms each representing detection events of gamma rays in a plurality of detector elements. The first creation unit sums up the plurality of dynamic sinograms to create an activity map based on the radioactivity level of the gamma rays. The reconstruction unit reconstructs, by using the plurality of dynamic sinograms, a plurality of dynamic images, the plurality of dynamic images each corresponding to one of the plurality of dynamic sinograms. The second creation unit creates at least one parametric image by using the plurality of dynamic sinograms and the activity map.SELECTED DRAWING: Figure 3B

Description

[Technical field]

[0001] The embodiments disclosed in this specification and the drawings relate to a medical image processing apparatus and a medical image processing method. [Background technology]

[0002] The Patlak method is one of the methods for analyzing dynamic PET data. In the Patlak method, a parametric image is generated by calculating the slope and tangent point of the Patlak plot, which allows the extraction of information contained in the dynamic PET data.

[0003] However, in dynamic PET analysis, one of the challenges is to appropriately adjust the magnitude of the penalty function. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] US Patent Application Publication No. 2019 / 0365341 Summary of the Invention [Problem to be solved by the invention]

[0005] One of the problems to be solved by the embodiments disclosed in this specification and the drawings is to improve image quality. However, the problems to be solved by the embodiments disclosed in this specification and the drawings are not limited to the above problem. Problems corresponding to the effects of each configuration shown in the embodiments described later can also be positioned as other problems. [Means for solving the problem]

[0006] A medical image processing apparatus according to an embodiment includes an acquisition unit, a first generation unit, a reconstruction unit, and a second generation unit. The acquisition unit acquires a plurality of dynamic sinograms, each dynamic sinogram representing a detection event of a gamma ray at a plurality of detector elements. The first generation unit sums the plurality of dynamic sinograms to generate an activity map based on a radioactivity level of the gamma ray. The reconstruction unit uses the plurality of dynamic sinograms to reconstruct a plurality of dynamic images, each dynamic image corresponding to a respective one of the plurality of dynamic sinograms. The second generation unit uses the plurality of dynamic sinograms and the activity map to generate at least one parametric image. [Brief description of the drawings]

[0007] [Figure 1] FIG. 1 is a diagram illustrating the processing performed by a medical image processing apparatus according to an embodiment. [Diagram 2] FIG. 2 is a flowchart illustrating the procedure of processing performed by the medical image processing apparatus according to the embodiment. [Figure 3A] FIG. 3A is a diagram illustrating an example of the configuration of a PET device according to an embodiment. [Figure 3B] FIG. 3B is a diagram illustrating an example of the configuration of the PET device according to the embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0008] Hereinafter, embodiments of a medical image processing apparatus and a medical image processing method will be described in detail with reference to the drawings.

[0009] The following disclosure presents many different embodiments, or examples, for implementing various features of the presented subject matter. Specific examples of components and configurations are described below to simplify the disclosure. These are, of course, merely examples and are not intended to be limiting. For example, a configuration in the following description in which a first feature is above or on a second feature may include an embodiment in which the first and second features are configured in direct contact, and may also include implementations in which additional features may be configured between the first and second features such that the first and second features do not need to be in direct contact. In addition, the disclosure may repeat reference numbers and / or letters in various examples. This repetition is for simplicity and clarity, and does not by itself dictate a relationship between the various embodiments and / or configurations described. Furthermore, spatially relative terms, such as top, bottom, below, below, below, above, above, etc., may be used herein to facilitate description of the relationship of one element or feature to another element or feature as shown in the figures. Spatially relative terms are intended to encompass various orientations of the device during use or operation in addition to the orientation depicted in the figures. The device may be otherwise oriented (rotated 90 degrees or at other orientations) and the spatially relative descriptions used herein may likewise be interpreted accordingly.

[0010] The order of description of the different procedures described herein is presented for clarity. In general, these procedures can be performed in any suitable order. Also, although each of the different features, techniques, configurations, etc. of this specification may be discussed elsewhere in this disclosure, it is intended that each of the concepts can be implemented independently of each other or in combination with each other. Thus, the present invention can be embodied and viewed in many different ways.

[0011] First, the configuration of a medical image processing apparatus according to an embodiment and a PET apparatus incorporating the medical image processing apparatus will be described with reference to FIGS. 3A and 3B.

[0012] 3A and 3B show a non-limiting example of a PET device 300 in which methods 100 and 200 can be performed. The PET device 300 includes several Gamma-Ray Detectors (GRDs) (e.g., GRD1, GRD2 through GRDN), each configured as a rectangular detector module. According to one implementation, the detector ring includes 40 GRDs. In another implementation, there are 48 GRDs, and the larger the number of GRDs used, the larger the inner diameter dimension of the PET device 300 will be created.

[0013] Each GRD can include a two-dimensional array of individual detector crystals that absorb gamma rays and emit scintillation photons. The scintillation photons can also be detected by a two-dimensional array of photomultiplier tubes (PMTs) located in the GRD. A light guide can be positioned between the detector crystal array and the PMTs.

[0014] Alternatively, the scintillation photons can be detected by an array of Silicon Photomultipliers (SiPMs), with each individual detector crystal having a corresponding SiPM.

[0015] Each photodetector (e.g., PMT or SiPM) can generate an analog signal indicative of when a scintillation event occurs and the energy of the gamma ray that caused the detection event. Additionally, photons emitted from one detector crystal can be detected by two or more photodetectors, and the detector crystal corresponding to the detection event can be determined based on the analog signal generated at each photodetector, for example, using Anger logic and crystal decoding.

[0016] FIG. 3B shows a schematic diagram of a PET device system having a gamma ray (gamma ray) photon counting detector (GRD) arranged to detect gamma rays emitted from an object OBJ. The GRD can measure the timing, position, and energy corresponding to each gamma ray detector. In one implementation, the gamma ray detectors are arranged in a ring as shown in FIGS. 3A and 3B. The detector crystals can be scintillator crystals having individual scintillator elements arranged in a two-dimensional array, which can be any known scintillation material. The PMTs can be arranged such that light from each scintillator element is detected by multiple PMTs to enable Anger arithmetic and crystal decoding of scintillation events.

[0017] 3B shows an example of the configuration of a PET device 300, in which an object OBJ to be imaged is placed on a stage 316, and GRD modules GRD1 to GRDN are arranged in a circumferential direction around the object OBJ and the stage 316. The GRD may be fixedly coupled to an annular component 320 that is fixedly coupled to a gantry 340. The stage 340 houses many parts of the PET imaging device. The stage 340 of the PET imaging device also includes an opening through which the object OBJ and the stage 316 can pass, and the gamma rays emitted in opposite directions from the object OBJ due to annihilation events can be detected by the GRD, and the coincidence counts of the gamma ray pairs can be measured using timing information and energy information.

[0018] Also shown in FIG. 3B are electronic circuits and hardware for acquiring, storing, processing, and distributing gamma ray detection data. These circuits and hardware include a processor 370, a network controller 374, a memory 378, and a data acquisition system (DAS) 376. The PET imaging device also includes a data channel that sends detection measurements from the GRD to the DAS 376, the processor 370, the memory 378, and the network controller 374. The DAS 376 can control the acquisition, digitization, and transmission of detection data from the detector. In one implementation, the DAS 376 controls the movement of the stage 316. The processor 370 performs functions including reconstruction of an image from the detection data, pre-reconstruction processing of the detection data, and post-reconstruction processing of image data, as discussed herein.

[0019] The processor 370 can be configured to perform various steps of the methods 100 and / or 200 described herein and variations thereof. The processor 370 can include a CPU that may be implemented as discrete logic gates, an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA), or other Complex Programmable Logic Device (CPLD). The FPGA or CPLD implementation may be coded in VHDL, Verilog, or any other hardware description language, and the code may be stored directly in the electronic memory in the FPGA or CPLD or as a separate electronic memory. Furthermore, the memory may be a non-volatile memory such as a ROM, EPROM, EEPROM, or FLASH memory. The memory may also be a volatile memory such as a static or dynamic RAM, and a processor such as a microcontroller or microprocessor may be provided to manage the electronic memory and the interaction between the FPGA or CPLD and the memory.

[0020] Alternatively, the CPU in processor 370 may execute a computer program including a set of computer readable instructions implementing the various steps of method 100 and / or method 200, the program being stored in any of the non-transitory electronic memories mentioned above and / or a hard disk drive, a CD, a DVD, a FLASH drive, or any other known storage medium. Additionally, the computer readable instructions may be provided as a utility application, a background daemon, or a component of an operating system, or a combination thereof, and executed in conjunction with a processor such as an Intel Xenon processor or an AMD Opteron processor, and an operating system such as Microsoft VISTA, UNIX, Solaris, LINUX, Apple, MAC-OS, and other operating systems known to those skilled in the art. Additionally, a CPU may be implemented as multiple processors cooperating in parallel to execute instructions.

[0021] The processor 370 (processing circuit) has an acquisition function, a first generation function, a second generation function, and a reconstruction function, which will be described later. In the embodiment, each processing function performed in the acquisition function, the first generation function, the second generation function, and the reconstruction function is stored in the memory 378 in the form of a program executable by a computer. The processor 370 is a processor that realizes the function corresponding to each program by reading the program from the memory 378 and executing it. In other words, the processor 370 in a state in which each program is read has each of those functions. Note that in FIG. 3B, the processing functions performed in the acquisition function, the first generation function, the second generation function, and the reconstruction function are described as being realized by a single processor 370, but the processor 370 may be configured by combining multiple independent processors, and each processor may realize a function by executing a program. In other words, each of the above-mentioned functions may be configured as a program, and one processor 370 may execute each program. As another example, a specific function may be implemented in a dedicated independent program execution circuit. The acquisition function, the first generation function, the second generation function, and the reconstruction function are examples of an acquisition unit, a first generation function, a second generation function, and a reconstruction function, respectively. Details of these functions will be described later.

[0022] The memory 378 may be a hard disk drive, a CD-ROM drive, a DVD drive, a FLASH drive, RAM, ROM, or other electronic storage device known in the art.

[0023] A network controller 374, such as an Intel Ethernet PRO network interface card from Intel Corporation, USA, can interface between the various components of the PET imaging device. Additionally, the network controller 374 can interface with an external network. As will be appreciated, the external network can be a public network, such as the Internet, or a private network, such as a LAN or WAN network, or a combination thereof, and can also include PSTN or ISDN sub-networks. The external network can be wired, such as an Ethernet network, or wireless, such as a cellular network, including EDGE, 3G, and 4G wireless cellular systems. The wireless network can be WiFi, Bluetooth, or any other wireless type of communication known in the art.

[0024] In the above description, specific details are described, such as the particular form of the processing system, and a description of various components and processes used in the system. However, it should be understood that the techniques herein may be implemented in other embodiments that differ from these specific details, and that such details are for illustrative purposes and not for limiting purposes. The embodiments disclosed herein have been described with reference to the accompanying drawings. Similarly, for the purpose of explanation, specific numbers, materials, and configurations are described to provide a thorough understanding. Nevertheless, the embodiments may be implemented without such specific details. Components having substantially the same functional configuration may be given similar reference numerals, and therefore repeated descriptions may be omitted.

[0025] Returning to FIG. 1, the process according to the embodiment will be described. Some methods for generating parametric images during dynamic Positron Emission Tomography (PET) include reconstructing dynamic emission images from dynamic sinogram data, and then performing Patlak analysis on the Time Activity Curve (TAC) to generate the parametric images. This can be considered an indirect method. However, the indirect method is susceptible to noise, which may significantly degrade the quality of the final parametric image.

[0026] In some methods, direct reconstruction can reconstruct Patlak parametric images from dynamic sinogram data by incorporating the Patlak model into the image reconstruction procedure. Unfortunately, direct methods may be more complex and time-consuming than indirect methods, limiting their practical use. For example, a quadratic penalty may be applied to frame-by-frame reconstruction and Patlak parametric image estimation to maximally enhance lesion detection performance. However, this method may limit the freedom to adjust the penalty strength by other rules and may not include the option to change the penalty type / parameters from one region to another.

[0027] As described herein, regularization methods are incorporated into dynamic PET reconstruction including regularization / penalty parameter selection, where a spatial deformation map can be determined for penalty type, penalty parameter and penalty strength based on an estimated activity map of the combined dynamic scan data.

[0028] The method described herein can improve PET image quality by spatially varying the penalty function during parametric fitting based on activity level mapping derived from the reconstructed images. In one embodiment, the method utilizes activity estimation from reconstructing measured emission data summed over a predetermined scan time to derive smoothing strength for dynamic reconstruction. The optimal penalty strength may depend on the activity level of a given image. That is, it may be useful to use different penalties for images with different amounts of activity. Furthermore, this idea can be carried over to different regions within an image. That is, regions with high activity levels may benefit from different penalties than regions with low activity levels. The smoothing strength based on static reconstruction can be applied to both frame-by-frame reconstruction and voxel or pixel-based Patlak model fitting. While the penalty type and parameters, such as edge preservation parameters, can be varied, the parameters of interest are the smoothing strength and how much it varies based on activity and how much it is imposed on the dynamic parametric images. Thus, the method allows scaling of the penalty term according to any desired function of activity.

[0029] In one embodiment, activity-dependent smoothing parameter scaling can also be used in combination with another smoothing parameter scaling, such as attenuation-dependent scaling. In particular, image-derived or population-based blood input functions can be used with this method.

[0030] FIG. 1 illustrates the reconstruction of a dynamic emission image 100b (also referred to herein as "dynamic image 100b") and parametric fitting of the emission image 100b to generate a parametric image 100c according to an embodiment of the present disclosure. As described above, the dynamic image 100b can be reconstructed frame by frame from the dynamic sinogram data 100a. The parametric image 100c can be generated by performing a Patlak analysis on the TAC on a pixel by pixel (voxel by voxel) basis to estimate the parametric image 100c. This indirect method is susceptible to noise, which reduces the image quality of the parametric image 100c. This is because the spatial information embedded in the dynamic PET data is ignored in the Patlak analysis.

[0031] In some embodiments, there is a set of measured dynamic sinogram data 100a that includes T frames, where y=[y1′, y2′, . . . , y T '], then the reconstructed image x n (hat) is generated by solving the optimization problem shown in the following equation (1).

[0032]

number

[0033] In the formula, y n is the measured sinogram of frame n, and x n is the unknown radioactive tracer / activity distribution, and L(y n |x n ) is the log-likelihood function, and φ(x n ) is a penalty function (also called a regularization function or regularizer), and β n is the penalty parameter (also called the normalized strength) that controls the degree of the penalty. In the above objective function, L(y n |x n ) is the data faithfulness term. The penalty parameter β nprovides the relative weight between the data-faithful regularization and the terms. For example, the penalty parameter β n Increasing the value of increases the degree of smoothing, reducing both the noise level and the resolution.

[0034] The reconstructed image for each frame is x(hat) = [x1'(hat), x2'(hat), …, x n '(hat)], the Patlak parametric image can be estimated pixel by pixel (or voxel by voxel) using a least-squares estimation with regularization as shown in equation (2) below.

[0035]

number

[0036] where κ and b are the parametric images of the Patlak slope and Patlak intercept, respectively. A is constructed by the blood input function and its integral from frame 1 to T, the multiplication sign on the right hand side of equation (2) represents the Kronecker product, and I T is the identity matrix, and β κ and β b is a parameter that controls the degree of penalty in a parametric image. Also, φ1 and φ2 are penalty functions, and the function system can be expressed in the form of equation (3).

[0037] In some embodiments, the penalty function may be a patch-based edge-preserving penalty function that uses nearby patches instead of individual pixels / voxels to calculate the penalty, as expressed by the following equation (3):

[0038]

number

[0039] where the l-2 norm or Euclidean distance is given by the following equation (4):

[0040]

number

[0041] This l-2 norm or Euclidean distance may be a measure of the distance between the patch of pixels / voxels surrounding pixel / voxel j and pixel / voxel l, where N j ={x j1、 x j2、 …, x jk} represents the neighborhood of pixel / voxel j and γ jl is the weight related to the distance between pixel / voxel j and pixel / voxel l, and ψ(·) is the potential function derived by the following equation (5):

[0042]

number

[0043] where δ is a position-dependent penalty parameter that controls the shape of the potential function and the extent of edge preservation.

[0044] The above-mentioned penalty function is easy to generalize and can be applied to both pairwise and patch-based penalties, as will be appreciated by those skilled in the art. Other penalty functions can be used without departing from the spirit and scope of the method described herein, as will be appreciated by those skilled in the art. For example, a Total Variation (TV) minimization penalty function penalizes solutions with high total variation. TV normalization is one non-limiting example of many different types of normalization penalties that can be used without departing from the spirit and scope of the method described herein. Other non-limiting examples of normalization penalties include, for example, Huber penalty, broken parabola penalty, quadratic penalty, and non-quadratic penalty.

[0045] Penalty parameter β n By increasing the magnitude of , noise in the reconstructed image is suppressed along with reducing the resolution. Thus, in regions of lower activity, a higher degree of smoothing may be useful, as the counting statistics of these regions make them prone to low signal-to-noise ratios (SNRs) (i.e., relatively high noise levels) and fine resolution within these regions may not be necessary. For example, higher tracer concentrations and activity levels are generally found near the region of interest (e.g., due to higher metabolic rates of lesions). Thus, lower activity may be associated with a larger penalty parameter β. n It may be useful to have an image reconstruction corresponding to

[0046] Furthermore, in areas of high activity, the signal is already large compared to the noise, so a low penalty parameter β n This allows us to prevent the resolution from being reduced without sacrificing image quality. Therefore, in these regions of high activity, the penalty parameter β n The SNR may already be high enough without increasing the smoothing by making σ large.

[0047] FIG. 2 illustrates a non-limiting example of a flowchart for implementing the method described herein, according to an embodiment of the present disclosure. In one embodiment, the penalty parameters / type may depend on the spatial variation of activity levels, which may be derived from the reconstructed image itself, allowing, for example, more smoothing in areas of low activity and less smoothing and edge enhancement in areas of high activity. In some embodiments, the penalty type, one or more penalty parameters (e.g., parameters controlling edge enhancement penalty / shape), and penalty strength may each be based on an initially estimated activity map (e.g., a pre-reconstructed PET image). The optimal degree of penalty may depend on the activity level of a given image. That is, different images with different amounts of activity may benefit from different degrees of penalty. Furthermore, this idea can be carried over to different regions within an image. That is, regions with high activity levels may benefit from different degrees of penalty than regions with low activity levels. Since different organs have different background activity levels, the penalty parameters may be location-dependent (e.g., organ activity-dependent) in addition to PET scan-dependent (e.g., image and / or acquisition time-dependent). That is, in the second generating function, processor 370 may segment the activity map based on the types of organs represented in the activity map and determine the penalty term based on the segmented activity map.

[0048] In FIG. 2, method 200 begins at step 210 by acquiring emission data 205 for tomographic reconstruction of a PET image. For example, incident radiation (e.g., gamma rays) incident on an array of scintillator crystals is converted into scintillation photons, which are converted into photoelectrons by a photodetector. Furthermore, in some embodiments, the photoelectron signals from the photodetectors may be amplified, pulse-shaped, filtered, and otherwise preconditioned and processed before being digitized using a Data Acquisition System (DAS) as time, energy, and position information of individual counts and / or coincidence pairs, and then stored in a non-transitory computer-readable memory. The PET data 205 may then be retrieved from the memory in preparation for tomographic reconstruction.

[0049] In some embodiments, the PET data 205 may be counts pre-processed, calibrated, and / or adjusted into a sinogram format (e.g., a histogram of counts). That is, the processor 370 acquires a plurality of dynamic sinograms, each dynamic sinogram representing gamma ray detection events at a plurality of detector elements, via an acquisition function. The dynamic sinograms are acquired over a predetermined time period, e.g., at predetermined time intervals. In some embodiments, the pre-processing may include position correction, energy correction, timing correction, etc. Additionally, the emission data 205 may be corrected for sensitivity using various calibration and geometric factors. Additionally, the pre-processing may include correction for detector offset and gain, variations in quantum efficiency in the detector, etc. Additionally, these corrections may be based on calibration data, empirical and known parameters.

[0050] In step 220, an initial PET image is reconstructed. For example, the PET data 205 is used to reconstruct an image of radioactivity level (e.g., tracer density) as a function of voxel location. Image reconstruction may be performed using backprojection, filtered backprojection, Fourier transform-based image reconstruction, iterative image reconstruction, matrix inversion reconstruction, statistical image reconstruction, list mode, or other reconstruction techniques or combinations thereof, as will be appreciated by those skilled in the art. That is, the processor 370 uses the multiple dynamic sinograms through a reconstruction function to reconstruct multiple dynamic images, each dynamic image corresponding to a respective one of the multiple dynamic sinograms.

[0051] In step 220, an activity map is determined from the first PET image, which is not necessarily the same as the first PET image. In one embodiment, given measured dynamic sinogram data 100a [y1', y2', ..., y T '], the dynamic sinogram data 100a is expressed as y=y1+y2+...y T Then, sequential ordered subset expectation maximization (OSEM) iterations are performed on the summed dynamic sinogram data 100a to obtain the activity map x OSEM That is, the first generation function may cause the processor 370 to sum a plurality of dynamic sinograms to generate an activity map based on the gamma radiation activity levels. As an example, the first generation function may cause the processor 370 to generate the activity map using an OSEM method that is iteratively repeated until a predetermined stopping criterion is reached.

[0052] In some embodiments, fine resolution is not required for the spatial variation of the penalty. For example, the activity map may be based on a smoothed or coarsened version of the initial PET image, where a low spatial frequency filter is applied to the initial PET image. Other denoising methods may also be applied to reduce the effect of noise on the activity map and the spatial variation of the penalty.

[0053] Exemplary denoising methods include linear smoothing filters, anisotropic diffusion, non-local means, and non-linear filters. Linear smoothing filters remove noise by convolving the original image with a mask that represents a low-pass filter or smoothing operation. For example, a Gaussian mask contains elements determined by a Gaussian function. This convolution causes each pixel / voxel's value to more closely match the values ​​of its neighboring pixels / voxels. Anisotropic diffusion removes noise while preserving sharp edges by evolving the image under a smooth partial differential equation similar to the heat equation. A median filter is an example of a non-linear filter that, if properly designed, can also preserve edges and avoid blurring. A median filter is an example of a Rank-Conditioned Rank-Selection (RCRS) filter that can be applied to remove salt-and-pepper noise from an image without introducing significant blurring artifacts. Furthermore, filters using a total variation (TV) minimizing regularization term can be applied when the imaged region corresponds to an estimate of uniformity over a large area defined by sharp boundaries between uniform areas. A TV filter is another example of a non-linear filter. Furthermore, non-local means filtering is an exemplary method that uses a weighted average over similar patches in an image to determine which pixels / voxels are denoised.

[0054] In step 230, the parametric image 100c can be estimated along with incorporating the penalty function. That is, the processor 370 generates at least one parametric image using the multiple dynamic sinograms and activity maps by the second generation function. In one embodiment, the dynamic image 100b can be analyzed to generate a TAC for each dynamic image 100b. That is, the processor 370 generates a time activity curve (TAC) for each of the multiple video images 100b by the first generation function. The final parametric image 240 can then be generated by performing a Patlak analysis on the TAC for each dynamic image 100b in a pixel-by-pixel (or voxel-by-voxel) manner to estimate the final parametric image 240. That is, the processor 370 generates at least one parametric image using the activity maps generated in step 220 and the time activity curves generated in step 230 by the second generation function. The Patlak parametric image can be estimated pixel-by-pixel (or voxel-by-voxel) from the frame-by-frame reconstructed image x(hat), for example, using least-squares estimation.

[0055] The Patlak parametric image may be expressed in terms of κ (the parametric image of the Patlak slope) and b (the parametric image of the Patlak intercept). The penalty is a parameter that controls the degree of penalty for the parametric image, β κ and β b That is, the processor 370 generates at least one parametric image using a second generation function, an objective function including a penalty term determined from the activity map and the dynamic sinograms, the penalty term being determined based on a first parameter β that controls the degree of penalty based on a Patlak slope κ of the time-activity curve. κand a second parameter that controls the degree of the penalty based on the Patlak intercept of the time-activity curve. In some embodiments, the spatial variation of the penalty comprises a position-dependent β n , β κ , and β b The penalty term β n , β κ , and β b The position dependence of may refer to the β-map, and j Here, the subscript j represents the number of reconstructed PET images x OSEM is the index of the j-th pixel / voxel of . The β-map is the function of β for each pixel / voxel j. j =f(x jOSEM ), where f(·) may be any user-selected function. For example, the user-selected function f(·) may include smoothing, low frequency filtering, and / or noise removal as described above. That is, the processor 370 may generate at least one parametric image using voxel-wise parametric fitting for each of the plurality of dynamic sinograms via the second generation function.

[0056] The evolving β-map is β for each frame n. nj =f n (x jOSEM ) normalization of the frame-by-frame reconstruction, and β κj =f κ (x jOSEM ), β bj =f b (x jOSEM ) can be applied to the Patlak model fitting of δ j =g(x jOSEM) may be applied to a penalty parameter δ that controls the shape and extent of edge preservation of the potential function, and may be applied to frame-by-frame reconstruction and Patlak fitting. That is, the processor 370 generates at least one parametric image using an objective function that includes a penalty term, with a second generation function, the penalty term including an edge-preserving potential function having a shape based on the activity map. In particular, an image-derived blood input function or a population-based blood input function may be used for the Patlak fitting. It will be appreciated that any globally convergent numerical optimization algorithm may be used to find an optimization system for the final cost function (with a convex penalty).

[0057] Advantageously, the method 200 provides different penalty parameters in dynamic PET reconstruction for the generation of the parametric image 100c, both for frame-by-frame reconstruction of the dynamic image 100b and for Patlak fitting. In particular, the method 200 uses spatial variation penalty parameter adjustments guided by static reconstruction of the data itself. That is, the processor 370 generates at least one parametric image using an objective function that includes a penalty term that includes one or more spatial variation smoothing parameters determined from the activity map by a second generation function. Thus, the method 200 allows the user full flexibility in selecting the smoothing and edge preservation strength according to any desired function of activity estimation, thereby allowing the reconstruction method to be optimized for a patient in a specific task in routine clinical practice.

[0058] The embodiments of the present disclosure can also be described as follows:

[0059] (Appendix 1) One aspect of the present invention provides a medical image processing apparatus including a processing circuit (processor) that acquires a plurality of dynamic sinograms, each dynamic sinogram representing a gamma ray detection event at a plurality of detector elements, sums the plurality of dynamic sinograms to generate an activity map based on a gamma ray activity level, reconstructs a plurality of dynamic images using the plurality of dynamic sinograms, each dynamic image corresponding to a respective one of the plurality of dynamic sinograms, and generates at least one parametric image using the plurality of dynamic sinograms and the activity map.

[0060] (Appendix 2) The circuitry may be further configured to generate a time activity curve for each of the plurality of dynamic images, and at least one parametric image may be created using the activity map and the time activity curve for each of the plurality of dynamic images.

[0061] (Appendix 3) The circuitry may be further configured to generate at least one parametric image using an objective function that includes a penalty term, the penalty term may include one or more spatial variation smoothing parameters determined from the activity map.

[0062] (Appendix 4) The penalty term may include a first parameter that controls the magnitude of the penalty based on a Patlak slope of the time-activity curve, and a second parameter that controls the magnitude of the penalty based on a Patlak intercept of the time-activity curve.

[0063] (Appendix 5) The circuitry may be further configured to generate at least one parametric image using an objective function that includes a penalty term, the penalty term may include an edge-preserving potential function having a shape based on the activity map.

[0064] (Appendix 6) The circuitry may generate at least one parametric image for each of the multiple dynamic sinograms using voxel-wise parametric fitting.

[0065] (Appendix 7) The system may further comprise a detector including a plurality of detector elements configured to detect gamma rays emitted by the tracer.

[0066] (Appendix 8) The activity map may be generated using an Ordered-Subset Expectation Maximization (OSEM) method that is iteratively repeated until a predefined stopping criterion is reached.

[0067] (Appendix 9) The dynamic sinogram may be acquired at predetermined time intervals and for a predetermined period of time.

[0068] (Appendix 10) The processing circuitry may segment the activity map based on the types of organs represented in the activity map and determine a penalty term based on the segmented activity map.

[0069] (Appendix 11) One aspect of the present invention provides a medical image processing apparatus including a processing circuit (processor) that acquires a plurality of dynamic sinograms, each dynamic sinogram representing a gamma ray detection event at a plurality of detector elements, generates an activity map based on the plurality of dynamic sinograms, reconstructs a plurality of dynamic images using the plurality of dynamic sinograms, each dynamic image corresponding to a respective one of the plurality of dynamic sinograms, and generates at least one parametric image using the plurality of dynamic sinograms and an objective function including a penalty term determined from the activity map.

[0070] (Appendix 12) One aspect of the present invention provides a medical image processing method including acquiring a plurality of dynamic sinograms, each dynamic sinogram representing a gamma ray detection event at a plurality of detector elements, summing the plurality of dynamic sinograms to generate an activity map based on the gamma ray radioactivity level, reconstructing a plurality of dynamic images using the plurality of dynamic sinograms, each dynamic image corresponding to a respective one of the plurality of dynamic sinograms, and generating at least one parametric image using the plurality of dynamic sinograms and the activity map.

[0071] (Appendix 13) The method may further include generating a time activity curve for each of the plurality of dynamic images, and at least one parametric image may be generated using the activity map and the time activity curve for each of the plurality of dynamic images.

[0072] (Appendix 14) Generating the at least one parametric image may further include using an objective function that includes a penalty term, where the penalty term may include one or more spatial variation smoothing parameters determined from the activity map.

[0073] (Appendix 15) The penalty term may include a first parameter that controls the magnitude of the penalty based on a Patlak slope of the time-activity curve, and a second parameter that controls the magnitude of the penalty based on a Patlak intercept of the time-activity curve.

[0074] (Appendix 16) Generating the at least one parametric image may further include using an objective function that includes a penalty term, where the penalty term may include an edge-preserving potential function having a shape based on the activity map.

[0075] (Appendix 17) Generating the at least one parametric image may further include using voxel-wise parametric fitting for each of the plurality of dynamic sinograms.

[0076] (Appendix 18) The parametric fitting may use a least squares estimation.

[0077] (Appendix 19) The activity map may be generated using an Ordered-Subset Expectation Maximization (OSEM) method that is iteratively repeated until a predefined stopping criterion is reached.

[0078] (Appendix 20) The method may further include segmenting the activity map based on a type of organ represented in the activity map and generating at least one parametric image, wherein the penalty term is based on the segmentation of the activity map.

[0079] (Appendix 21) One aspect of the present invention provides a non-transitory computer-readable storage medium including executable instructions which, when executed by a processing circuit, cause the processing circuit to perform a method including acquiring a plurality of dynamic sinograms, each dynamic sinogram representing a gamma ray detection event at a plurality of detector elements, summing the plurality of dynamic sinograms to generate an activity map based on a gamma ray radioactivity level, reconstructing a plurality of dynamic images using the plurality of dynamic sinograms, each dynamic image corresponding to a respective one of the plurality of dynamic sinograms, and generating at least one parametric image using the plurality of dynamic sinograms and the activity map.

[0080] According to at least one of the embodiments described above, image quality can be improved.

[0081] Although some embodiments have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, substitutions, modifications, and combinations of embodiments can be made without departing from the spirit of the invention. These embodiments and their modifications are included in the scope of the invention and its equivalents as described in the claims, as well as in the scope and spirit of the invention. [Explanation of symbols]

[0082] 370 Processor 378 Memory

Claims

1. an acquisition unit for acquiring a plurality of dynamic sinograms, each dynamic sinogram representing a gamma ray detection event at a plurality of detector elements; a first generator that sums the plurality of dynamic sinograms to generate an activity map based on the gamma ray activity level; a reconstruction unit that uses the plurality of dynamic sinograms to reconstruct a plurality of dynamic images, each dynamic image corresponding to a respective one of the plurality of dynamic sinograms; and a second generator configured to generate at least one parametric image using data derived from the plurality of dynamic sinograms and the activity map.

2. The first generator further generates a time activity curve for each of the plurality of dynamic images; The medical image processing apparatus according to claim 1 , wherein the second generator is configured to generate the at least one parametric image using the activity map and the time activity curve.

3. The medical image processing apparatus of claim 2 , wherein the second generator generates the at least one parametric image using an objective function that includes a penalty term that includes one or more spatial variation smoothing parameters determined from the activity map.

4. The medical image processing device according to claim 3 , wherein the penalty term includes a first parameter that controls a degree of the penalty based on a Patlak slope of the time activity curve, and a second parameter that controls a degree of the penalty based on a Patlak intercept of the time activity curve.

5. 3. The medical image processing device of claim 2, wherein the second generator generates the at least one parametric image using an objective function that includes a penalty term, the penalty term including an edge-preserving potential function having a shape based on the activity map.

6. The medical image processing apparatus according to claim 1 , wherein the second generating unit generates the at least one parametric image using voxel-by-voxel parametric fitting for each of the plurality of dynamic sinograms.

7. 2. The medical imaging apparatus of claim 1, wherein the activity map is generated using an Ordered-Subset Expectation Maximization (OSEM) method that is iteratively repeated until a predefined stopping criterion is reached.

8. The medical imaging system of claim 1 , wherein the dynamic sinogram is acquired at predetermined time intervals and for a predetermined period of time.

9. the second generating unit segments the activity map based on a type of organ represented in the activity map, and determines the penalty term based on the segmented activity map. The medical image processing device according to claim 3 .

10. an acquisition unit for acquiring a plurality of dynamic sinograms, each dynamic sinogram representing a gamma ray detection event at a plurality of detector elements; a first generator for generating an activity map based on the plurality of dynamic sinograms; a reconstruction unit that uses the plurality of dynamic sinograms to reconstruct a plurality of dynamic images, each dynamic image corresponding to a respective one of the plurality of dynamic sinograms; an objective function including a penalty term determined from the activity map and data obtained from the plurality of dynamic sinograms; and a second generator for generating at least one parametric image using the Medical imaging equipment.

11. A medical image processing method performed by a medical image processing device, comprising: obtaining a plurality of dynamic sinograms, each dynamic sinogram representing a gamma ray detection event at a plurality of detector elements; summing the plurality of dynamic sinograms to generate an activity map based on the activity levels of the gamma rays; using the plurality of dynamic sinograms to reconstruct a plurality of dynamic images, each dynamic image corresponding to a respective one of the plurality of dynamic sinograms; A method of medical image processing using data derived from the plurality of dynamic sinograms and the activity map to generate at least one parametric image.

Images