Medical image processing device, medical image processing method and program

Abstract

To improve the image quality.SOLUTION: A medical image processing apparatus according to an embodiment comprises: an acquisition unit; a calculation unit; and a reconstruction processing unit. The acquisition unit acquires information about an electron density function of a subject and a nuclide administered to the subject. The calculation unit calculates positron range kernel about the subject on the basis of the electron density function and the nuclide. The reconstruction processing unit performs PET image reconstruction of the subject on the basis of the positron range kernel.SELECTED DRAWING: Figure 1

Description

[Technical Field]

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

[0002] Current PET devices count the gamma rays that are paired together when a positron emitted from a labeled drug annihilates a surrounding electron, and generate an image of the distribution of annihilation points calculated based on the counting data.

[0003] Here, the pair annihilation points and the positron emission points that represent the distribution of the labeled drug used to observe the pathology do not necessarily coincide, and the distribution image of the pair annihilation points and the distribution image of the positron emission points are essentially different. However, in the case of F-18, the labeled nucleus mainly used in PET examinations, the positron range is sufficiently short, averaging 0.44 mm in water, so there are cases where no major practical problems arise even if the distribution image of the pair annihilation points is considered to be the distribution image of the positron emission points.

[0004] However, the positron ranges of labeled nuclei other than F-18 are comparable to or larger than the pixel size. For example, the average water positron ranges of Rb-82, used in cardiac PET examinations, and Ga-68, used in immune PET, are 5 mm or more and 2.5 mm, respectively. Therefore, when conventional reconstruction methods are applied directly to images, the blurring of the positron range can reduce the visibility and quantification of lesions. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] U.S. Patent No. 6,987,270 Summary of the Invention [Problem to be solved by the invention]

[0006] 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 below can also be positioned as other problems. [Means for solving the problem]

[0007] A medical image processing apparatus according to an embodiment includes an acquisition unit, a calculation unit, and a reconstruction processing unit. The acquisition unit acquires an electron density function of a subject and information on a nuclide administered to the subject. The calculation unit calculates a positron range kernel for the subject based on the electron density function and the nuclide. The reconstruction processing unit performs PET image reconstruction of the subject based on the positron range kernel. [Brief explanation of the drawings]

[0008] [Figure 1] FIG. 1 is a diagram showing a PET device according to an embodiment. [Figure 2] FIG. 2 is a diagram illustrating the background according to the embodiment. [Figure 3] FIG. 3 is a diagram illustrating the processing performed by the medical image processing apparatus according to the embodiment. [Figure 4] FIG. 4 is a diagram illustrating the processing performed by the medical image processing apparatus according to the embodiment. [Figure 5] FIG. 5 is a flowchart illustrating the procedure of processing performed by the medical image processing apparatus according to the embodiment. [Figure 6] FIG. 6 is a flowchart illustrating an example of the procedure of the process performed in step S200 of FIG. DETAILED DESCRIPTION OF THE INVENTION

[0009] Hereinafter, embodiments of a medical image processing apparatus, a medical image processing method, and a program will be described in detail with reference to the drawings.

[0010] (Embodiment) Fig. 1 is a diagram showing the configuration of a PET device 100 as a medical image processing device according to an embodiment. As shown in Fig. 1, the PET device 100 according to the embodiment includes a gantry device 31 and a medical image processing device 32 that also functions as a console device. The gantry device 31 includes a detector 3, a front-end circuit 102, a tabletop 103, a bed 104, and a bed driver 106.

[0011] Detector 3 is a detector that detects radiation by detecting scintillation light (fluorescence), which is light re-emitted when an annihilation gamma ray emitted from a positron in subject P interacts with a light emitter (scintillator) and causes a substance that has become excited and transitions back to the ground state. Detector 3 detects radiation energy information of the annihilation gamma ray emitted from a positron in subject P. Multiple detectors 3 are arranged in a ring shape around subject P, and are made up of, for example, multiple detector blocks.

[0012] One example of a specific configuration of detector 3 is a photon-counting, Anger-type detector, which includes, for example, a scintillator, a photodetector, and a light guide. Another example of a configuration is a non-Anger-type detector in which a scintillator and a photodetector are optically coupled one-to-one. That is, each pixel included in detector 3 includes a scintillator and a photodetector that detects generated scintillation light.

[0013] The scintillator converts incident annihilation gamma rays emitted from positrons in the subject P into scintillation photons (optical photons) and outputs the light. The scintillator is formed of scintillator crystals such as LYSO (Lutetium Yttrium Oxyorthosilicate), LSO (Lutetium Oxyorthosilicate), LGSO (Lutetium Gadolinium Oxyorthosilicate), BGO, etc., and is arranged, for example, two-dimensionally.

[0014] Examples of photodetector elements that can be used include SiPMs (Silicon photomultipliers) and photomultiplier tubes. Photomultiplier tubes have a photocathode that receives scintillation light and generates photoelectrons, multiple dynodes that provide an electric field to accelerate the generated photoelectrons, and an anode through which the electrons flow. They multiply photoelectrons generated by scintillation light output from the scintillator and convert them into an electrical signal.

[0015] Furthermore, the gantry device 31 generates counting information from the output signal of the detector 3 using the front-end circuit 102, and stores the generated counting information in the memory unit 130 of the medical image processing device 32. The detector 3 is divided into multiple blocks and includes the front-end circuit 102.

[0016] The front-end circuit 102 converts the output signal of the detector 3 into digital data and generates counting information. This counting information includes the detection position, energy value, and detection time of the annihilation gamma ray. For example, the front-end circuit 102 identifies multiple photodetector elements that simultaneously converted scintillation light into electrical signals. The front-end circuit 102 then identifies a scintillator number (P) indicating the position of the scintillator onto which the annihilation gamma ray was incident. The position of the scintillator onto which the annihilation gamma ray was incident may be identified by performing a center of gravity calculation based on the position of each photodetector element and the intensity of the electrical signal. Furthermore, when the element sizes of the scintillators and the photodetector elements correspond to each other, for example, the scintillator corresponding to the photodetector element that produced the maximum output may be assumed to be the scintillator position onto which the annihilation gamma ray was incident, and the final identification may be performed taking into account inter-scintillator scattering.

[0017] The front-end circuit 102 also performs an integral calculation of the intensity of the electrical signal output from each photodetector element, or measures the time (time over threshold) at which the electrical signal intensity exceeds a threshold, and identifies the energy value (E) of the annihilation gamma ray incident on the detector 3. The front-end circuit 102 also identifies the detection time (T) at which the detector 3 detects scintillation light due to the annihilation gamma ray. The detection time (T) may be an absolute time or the elapsed time from the start of imaging. In this way, the front-end circuit 102 generates counting information including the scintillator number (P), the energy value (E), and the detection time (T).

[0018] The front-end circuit 102 is realized by a circuit such as a CPU (Central Processing Unit), a GPU (Graphical Processing Unit), an Application Specific Integrated Circuit (ASIC), or a programmable logic device (e.g., a Simple Programmable Logic Device (SPLD), a Complex Programmable Logic Device (CPLD), and a Field Programmable Gate Array (FPGA)). The front-end circuit 102 is an example of a front-end unit.

[0019] The top board 103 is a bed on which the subject P is placed, and is placed on the bed 104. The bed driving unit 106 moves the top board 103 under the control of the control function 105f of the processing circuit 150. For example, the bed driving unit 106 moves the top board 103 to move the subject P into the imaging opening of the gantry device 31.

[0020] The medical image processing device 32 accepts operations of the PET device 100 by an operator, controls the capturing of PET images, and reconstructs the PET images using the counting information collected by the gantry device 31. As shown in FIG. 1, the medical image processing device 32 includes a processing circuit 150, an input device 110, a display 120, and a storage unit 130. The various units included in the medical image processing device 32 are connected via a bus. Details of the processing circuit 150 will be described later.

[0021] The input device 110 is a mouse, keyboard, or the like used by the operator of the PET device 100 to input various instructions and settings, and transfers the input instructions and settings to the processing circuitry 150. For example, the input device 110 is used to input an instruction to start imaging.

[0022] The display 120 is a monitor or the like that is viewed by the operator, and under the control of the processing circuit 150, displays the subject's respiratory waveform and PET images, and displays a GUI (Graphical User Interface) for receiving various instructions and settings from the operator.

[0023] The storage unit 130 stores various data used in the PET device 100. The storage unit 130 is configured with, for example, a memory, and is realized by, for example, a semiconductor memory element such as a random access memory (RAM) or a flash memory, a hard disk, an optical disk, etc. The storage unit 130 stores counting information, which is information in which a scintillator number (P), an energy value (E), and a detection time (T) are associated with each other, coincidence information in which a coincidence number, which is a serial number of the coincidence information, is associated with a set of counting information, projection data obtained by aggregating the coincidence information, reconstructed PET images, etc.

[0024] The processing circuit 150 has an acquisition function 150a, a calculation function 150b, a reconstruction function 150c, a control function 150d, a reception function 150e, an image generation function 150f, a display control function 150g, and a learning function 150i.

[0025] In the embodiment, each processing function performed by the acquisition function 150a, calculation function 150b, reconstruction function 150c, control function 150d, reception function 150e, image generation function 150f, display control function 150g, and learning function 150i is stored in the storage unit 130 in the form of a program executable by a computer. The processing circuitry 150 is a processor that realizes the function corresponding to each program by reading and executing the program from the storage unit 130. In other words, the processing circuitry 150 in a state in which each program has been read has each function shown in the processing circuitry 150 of FIG. 1.

[0026] 1, the processing functions performed by the acquisition function 150a, calculation function 150b, reconstruction function 150c, control function 150d, reception function 150e, image generation function 150f, display control function 150g, and learning function 150i are described as being realized by a single processing circuit 150. However, the processing circuit 150 may be configured by combining multiple independent processors, and each processor may execute a program to realize the function. In other words, each of the above functions may be configured as a program, and one processing circuit 150 may execute each program. As another example, a specific function may be implemented in a dedicated, independent program execution circuit.

[0027] The term "processor" used in the above description refers to circuits such as a CPU (Central Processing Unit), a GPU (Graphical Processing Unit), an Application Specific Integrated Circuit (ASIC), a programmable logic device (e.g., a Simple Programmable Logic Device (SPLD), a Complex Programmable Logic Device (CPLD), and a Field Programmable Gate Array (FPGA)). The processor realizes its functions by reading and executing programs stored in the memory unit 130.

[0028] In FIG. 1, the acquisition function 150a, the calculation function 150b, the reconstruction function 150c, the control function 150d, the reception function 150e, the image generation function 150f, the display control function 150g, and the learning function 150i are examples of an acquisition unit, a calculation unit, a reconstruction processing unit, a control unit, a reception unit, an image generation unit, a display control unit, and a learning unit, respectively.

[0029] The processing circuitry 150 controls the gantry device 31 and the medical image processing device 32 using a control function 150d, thereby performing overall control of the PET device 100.

[0030] The processing circuitry 105 also controls the bed driver 106 using a control function 150d. The processing circuitry 150 also receives various instructions from a user via the input device 110 or the display 120 using a reception function 150e. The processing circuitry 150 also generates various images using an image generation function 150f based on information acquired from the front-end circuitry 102. The processing circuitry 150 also displays, on the display 120, images reconstructed by the reconstruction function 150c and images generated by the image generation function 150f using a display control function 150g.

[0031] The acquisition function 150a, the calculation function 150b, the reconstruction function 150c, and the learning function 150i will be described in detail later.

[0032] Next, the background of the embodiment will be briefly described.

[0033] Current PET devices count the gamma rays that are paired together when a positron emitted from a labeling agent annihilates a positron with a surrounding electron, and reconstruct the distribution of annihilation points calculated based on the counting information into an image. For example, as shown in Figure 2, a positron emitted at positron emission point 1 annihilates with a surrounding electron at annihilation point 2, roughly within the positron range 6, and emits a pair of gamma rays. The pair of gamma rays is counted in coincidence by detector 3. The LOR5 of the gamma rays is estimated from the coincidence information in detector 3, and the distribution of annihilation points 2 is reconstructed based on this.

[0034] Here, the annihilation point 2 does not necessarily coincide with the positron emission point 1, which represents the distribution of the labeled drug used to observe the pathology, and the distribution image of the annihilation point 2 is essentially different from the distribution image of the positron emission point 1. However, since the positron range6 of F-18, the labeled nucleus mainly used in PET examinations, is an average of 0.44 mm in water, which is sufficiently short compared to the typical spatial resolution of PET devices, there are cases in which there is no major practical problem in considering the distribution image of the annihilation point 2 as the distribution image of the positron emission point 1.

[0035] However, the positron ranges of labeled nuclei other than F-18 are comparable to or larger than the pixel size. For example, the average water positron ranges of Rb-82, used in cardiac PET examinations, and Ga-68, used in immune PET, are 5 mm or more and 2.5 mm, respectively. Therefore, when conventional reconstruction methods are applied directly to images, the blurring of the positron range can reduce the visibility and quantification of lesions.

[0036] For example, when comparing the annihilation point image when F-18, a nuclide with a relatively small positron range, is used as the target nucleus with the annihilation point image when Ga-68, a nuclide with a relatively large positron range, is used as the target nucleus, the positron range of F-18 is relatively short, so the blurring of the image due to the positron range is not so noticeable, but in the case of Ga-68, the positron range of Ga-68 is relatively large, so the annihilation point image is blurred due to the positron range. Therefore, it is desirable to reconstruct a positron emission point image instead of the annihilation point image.

[0037] The trajectory of a positron emitted from a labeled nucleus in a material and the shape of the statistically formed vanishing point distribution depend on the kinetic energy of the emitted positron and the shape of the electron density distribution around the emission point. The electron density distribution shape in each part of the living body of a subject is not necessarily uniform, and may form a surface that changes rapidly, such as a tissue boundary. Therefore, the vanishing point distribution for each positron emission point is generally locally dependent and anisotropic. For example, in lung tissue, the shape of the electron density distribution changes significantly between the lung tissue and the air.

[0038] In such cases, it is known that positron range correction using a simplified approach, such as applying uniform isotropic deconvolution to all points, can result in artifacts. Therefore, it has been thought that a treatment that takes into account the local dependence and anisotropy of the electron density distribution is desirable, but modeling this has been considered difficult due to its complexity.

[0039] In view of this background, the medical image processing apparatus 32 according to the embodiment includes a processing circuitry 150. The processing circuitry 150 acquires information about an electron density function of a subject and a nuclide administered to the subject using an acquisition function 150a. The processing circuitry 150 calculates a positron range kernel for the subject based on the electron density function and the nuclide using a calculation function 150b. The processing circuitry 150 performs PET image reconstruction of the subject based on the positron range kernel using a reconstruction function 150c.

[0040] Specifically, the processing circuitry 150 uses a positron range kernel and performs formulation using a system matrix given by the product of the positron range kernel and a detection probability matrix, thereby correcting the effect of the positron range on the PET image and improving the image quality of the output image.

[0041] In particular, this formulation allows the processing circuitry 150 to calculate a positron range kernel for each of multiple positions using the calculation function 150b, thereby enabling proper modeling of the effect of positron range on PET image reconstruction even in real-world situations where the electron density distribution has local dependency and anisotropy.

[0042] First, the method according to the embodiment will be outlined with reference to FIG. 3 and equations (1) to (4) described later.

[0043] First, to explain the relationship between the positron production point 1 and the annihilation point 2, let us define the number of annihilation gamma-ray pairs produced in the m-th voxel as ρ m Let the number of positrons emitted from the i-th voxel be λ. i Then, the following equation (1) holds.

[0044]

number

[0045] Here, the matrix T, which is the expansion coefficient of the right side of equation (1), mi is called the positron range kernel. Positron range kernel T mi As can be seen from equation (1), the number of positrons generated in the i-th voxel, λ i and the number of annihilation gamma-ray pairs generated in the m-th voxel, ρ m In other words, the number of positrons generated in the i-th voxel, λ i , positron range kernel T miMultiplying by a factor of 10 and summing over i gives the number of annihilation gamma rays produced in the mth voxel, ρ m is obtained.

[0046] Next, the number of annihilation gamma-ray pairs detected in the j-th LOR (Line Of Response) is denoted by g j Then, the following equation (2) holds.

[0047]

number

[0048] Here, the matrix H, which is the expansion coefficient of the right side of equation (2), jm is called the detection probability matrix. The detection probability matrix H jm As can be seen from equation (2), the number of annihilation gamma rays generated in the m-th voxel, ρ m and the number of pairs of annihilation gamma-rays detected in the jth LOR, g j In other words, the number of annihilation gamma rays generated in the m-th voxel, ρ m Then, the detection probability matrix H jm Multiplying by a factor of 11 and summing over m gives the number of annihilation gamma-ray pairs detected in the jth LOR, g j is obtained.

[0049] Here, when equation (1) is substituted into equation (2), the following equation (3) is obtained.

[0050]

number

[0051] Here, the system matrix H PR is given by the following equation (4).

[0052]

number

[0053] That is, as shown in equation (4), the system matrix H PR is the positron range kernel T mi and the detection probability matrix H jm This is the product of

[0054] In other words, the number of positrons generated in the i-th voxel, λ i , positron range kernel T mi 10 and the detection probability matrix H jm 11, the number of annihilation gamma-ray pairs detected in the jth LOR, g j This is the number of positrons generated in the i-th voxel, λ i , the system matrix H PR ji It can also be thought of as having acted on the system matrix H PR ji 12 is the positron range kernel T mi 10 and the detection probability matrix H jm This is the product of 11.

[0055] Figure 4 shows the system matrix PR ji The physical meaning of 12 is illustrated. That is, the system matrix H PR The (j, i) component of represents the probability that a pair annihilation gamma ray originating from a positron emitted from the i-th positron emission point 1 is detected in the j-th LOR 5. Here, the number of pair annihilation gamma ray pairs detected in each LOR g j and the system matrix H PR If the value of is known, the number of positrons emitted at each voxel can be calculated using equation (3).

[0056] Next, a specific flow of processing performed by the medical image processing apparatus 32 according to the embodiment will be described with reference to FIG.

[0057] 5, the processing of steps S300 and S400 is performed after the processing of steps S100 and S200, but the order in which steps S100 and S200 and steps S300 and S400 are performed is not limited, and for example, the processing of steps S300 and S400 may be performed after the processing of steps S100 and S200. In other words, it is sufficient that the processing of steps S200 and S400 has been completed at least before the processing of step S500 is started.

[0058] First, in step S100, the processing circuitry 150 acquires an electron density function of the subject and information about the nuclide administered to the subject using the acquisition function 150a. As an example, the processing circuitry 150 acquires an X-ray CT image of the subject using the acquisition function 150a, and acquires the electron density function of the subject based on the X-ray CT image. For example, the processing circuitry 150 acquires the electron density function of the subject using the acquisition function 150a based on the CT value of the X-ray CT image. Note that the embodiment is not limited to this. For example, the processing circuitry 150 may acquire an MRI image of the subject using the acquisition function 150a, and acquire the electron density function of the subject based on the acquired MRI image. Note that the embodiment has been described in terms of a case where the PET device 100 acquires the electron density function of the subject from an external CT device or MRI device, but the embodiment can also be applied to a PET-CT device or a PET-MRI device. In such a case, PET imaging can be performed based on CT images or MRI images taken or captured by the CT device or MRI device in the PET-CT device or PET-MRI device.

[0059] Subsequently, in step S200, the processing circuitry 150 calculates, by the calculation function 150b, a positron range kernel T mi Calculate the positron range kernel T miis a matrix that represents the probability that a positron generated in one voxel, voxel i, will annihilate with another voxel, voxel m.

[0060] The processing circuitry 150 calculates a positron range kernel for each of a plurality of positions (voxels, pixels) using the calculation function 150b, thereby making it possible to appropriately model the locality and anisotropy of the electron density distribution.

[0061] As a first method for calculating a positron range kernel for a subject, the processing circuitry 150 uses the calculation function 150b to calculate the probability that a positron emitted at voxel i will annihilate at voxel m based on a physical calculation using the electron density function and nuclide acquired in step S100, and then calculates the positron range kernel for the subject based on the calculated probability. For example, since the kinetic energy of a positron at the time of emission is known for each nucleus of a drug administered to the subject, the processing circuitry 150 uses the calculation function 150b to calculate the probability that a positron will annihilate at each position, assuming that the positron is emitted in each direction with an initial momentum estimated from the kinetic energy of the positron, based on the electron density function acquired in step S100. For example, the processing circuitry 150 uses the calculation function 150b to calculate the scattering cross section of the annihilation between the positron and an electron. As another example, the processing circuit 150 calculates the positron range kernel by formulating a transport equation for radiation using the calculation function 150b. The first method has an advantage, for example, that it is a simple method and requires a small amount of calculation compared to the second and third methods.

[0062] As a second method for calculating the positron range kernel for the object, the processing circuitry 150 may use the calculation function 150b to calculate the probability that a positron emitted at voxel i will annihilate at voxel m based on the electron density function and nuclide acquired in step S100, and then calculate the positron range kernel for the object based on the calculated probability. Compared to the first method, the second method has the advantage that calculations can be performed with a certain degree of accuracy even in cases involving multiple scatterings, for example.

[0063] As a third example of calculating a positron range kernel for a subject, the positron range kernel for the subject may be calculated by performing deep learning. As one example, the processing circuitry 150 may use the learning function 150i to learn the relationship between the electron density function and the probability distribution of the annihilation positions of the generated positrons, and may use the calculation function 150b to calculate the positron range kernel based on the learned model.

[0064] An example of such processing is shown in Fig. 6. Fig. 6 is a flowchart illustrating the processing steps of the third method for calculating the positron range kernel.

[0065] First, in step S210, the processing circuitry 150 extracts and acquires a plurality of sample images obtained by cutting out a portion of data from one CT image or a plurality of CT images using the acquisition function 150a, thereby acquiring a plurality of sample data of electron density distribution using the acquisition function 150a.

[0066] Next, in step S220, the processing circuit 150 uses the calculation function 150b to calculate, for each sample extracted in step S210, a probability distribution of positrons isotropically emitted from voxel i annihilating at voxel m. As an example, the processing circuit 150 uses the calculation function 150b to calculate, by physical calculation or Monte Carlo simulation, a probability distribution of positrons isotropically emitted from voxel i annihilating at voxel m. This creates a plurality of training data sets in which the electron density distribution is associated with the probability distribution of the position at which positrons annihilate.

[0067] Next, in step S230, the processing circuit 150 uses the learning function 150h to input the multiple pieces of training data created in step S220, in which the electron density distribution and the probability distribution of the positions where positrons annihilate, into a neural network such as a CNN (Convolutional Neural Network), thereby learning the relationship between the electron density distribution and the probability distribution of the positions where positrons annihilate through deep learning. As a result, the processing circuit 150 uses the learning function 150b to generate a trained model that has been trained on the relationship between the electron density function and the generated probability distribution of the positions where positrons annihilate.

[0068] In addition, in the embodiment, the processing of steps S210 to S230 does not need to be performed for each clinical imaging, and the learning processing of step 230 may be performed only once, for example, before clinical imaging, and the generated trained model may be executed commonly for, for example, multiple clinical imaging.

[0069] Next, in step S240, the processing circuitry 150 acquires the electron density distribution of the subject using the acquisition function 150a. As an example, the processing circuitry 150 acquires the electron density distribution of the subject using the acquisition function 150a from a CT image obtained during clinical imaging. Next, in step S250, the processing circuitry 150 inputs the electron density distribution acquired in step S240 into the trained model generated in step S230 using the calculation function 150b, acquires a probability distribution of annihilation positions obtained as an output result, and calculates a positron kernel based on the acquired probability distribution.

[0070] As described above, according to the third method, the processing circuitry 150 uses deep learning to calculate a positron range kernel for the subject using the calculation function 150b. As described above, the positron range kernel is anisotropic and locally dependent, and its value varies depending on the location. Even in such a case, by extracting a large amount of sample data from, for example, a CT image and using it for learning, the processing circuitry 150 can use the calculation function 150b to calculate an accurate positron range kernel even for an actual electron density distribution that is anisotropic and locally dependent.

[0071] Returning to FIG. 5, in step S200, the processing circuitry 150 uses the calculation function 150b to calculate a positron range kernel for the subject based on the electron density function and information on the nuclide, for example, by the first to third methods.

[0072] On the other hand, in step S300, the processing circuit 150 acquires detector geometric information, which is information indicating the position of each detector, using the acquisition function 150a. Subsequently, in step S400, the processing circuit 150 calculates, using the calculation function 150b, the detection probability matrix H jm 11. Calculate the detection probability matrix H jm 11 is a matrix representing the probability that a gamma ray pair 2 generated in association with a positron annihilated in one voxel m is detected by a detector pair associated with the jth LOR, which is one LOR 5.

[0073] Subsequently, when the processing of step S200 and the processing of step S400 are completed, in step S500, the processing circuit 150, using the calculation function 150b, calculates the positron range kernel T10 calculated in step S200 and the detection probability matrix H11 calculated in step S300, and calculates the left side of equation (4) and the system matrix H shown in FIG. PR Specifically, as shown in equation (4), the processing circuit 150 uses the calculation function 150b to calculate the product of the positron range kernel T10 calculated in step S200 and the detection probability matrix H11 calculated in step S300, thereby calculating the system matrix H PR 12. Calculate the system matrix H PR ji is a matrix representing the probability that a gamma ray generated by the annihilation of a positron generated in one voxel, voxel i, is detected by a detector pair associated with one LOR, j-th LOR.

[0074] Subsequently, in step S600, the processing circuit 150 uses the reconstruction function 150c to reconstruct the system matrix H calculated in step S500. PR 12, the processing circuitry 150 performs PET image reconstruction of the object based on the system matrix H calculated in step S500 by the reconstruction function 150c. PR Based on 12, the coefficient information g for each LOR shown on the left side of Equation (2) j Based on this, the number of positrons emitted for each voxel, λ i and reconstruct the positron emission point image.

[0075] In this embodiment, the coefficient information g for each LOR shown on the left side of equation (2) is j Based on the detection probability matrix H11, the number of annihilation gamma-ray pairs produced for each voxel, ρ, is calculated as shown on the right side of equation (2). m Therefore, it is possible to generate an image from which the influence of blurring due to the positron range is removed, thereby improving the image quality.

[0076] As a specific image reconstruction method, the processing circuitry 150 performs reconstruction using the reconstruction function 150c, for example, using the following equation (5).

[0077]

number

[0078] where λ i k is the number of positrons generated in the i-th voxel, λ i At the beginning, the processing circuit 150 uses the reconstruction function 150c to estimate the initial value λ1 of the number of positrons generated in the i-th voxel. 0 As an example, the processing circuit 150 uses the reconstruction function 150c to calculate the number of gamma ray annihilations at the i-th voxel by assuming that the positron emission point and the gamma ray annihilation point are the same, and calculates the number of gamma ray annihilations at the i-th voxel by the initial value λ1 of the number of positrons generated at the i-th voxel. 0 Then, the processing circuit 150, by using the reconstruction function 150c, calculates an estimate λ of the number of positrons generated in the i-th voxel at the k-th iteration step, for example, by evaluating the right-hand side of equation (5) and substituting it into the left-hand side. i k Based on this, the estimated value λ of the number of positrons generated in the i-th voxel at the k+1-th iteration step i k+1 The processing circuit 150 terminates the process when the estimated value has sufficiently converged using the reconstruction function 150c, calculates the estimated value at that time as the number of positrons generated in the i-th voxel, and generates a positron emission point image.

[0079] Note that the above-mentioned formula (5) shows an example of the reconstruction processing performed by the processing circuitry 150 in step S600, and there are various variations other than the above-mentioned reconstruction method that can be considered as the method of the reconstruction processing performed by the processing circuitry 150. As an example, the processing circuitry 150 may use a formula that takes into account scattered radiation correction in formula (5), for example.

[0080] In this way, the medical image processing apparatus according to the embodiment can correct the effect of the positron range in a PET image by performing formulation using a positron range kernel, thereby improving the image quality of the output image.

[0081] As an example, in this formulation, processing circuitry 150 can calculate a positron range kernel for each of multiple positions using calculation function 150b, thereby enabling proper modeling of the effect of positron range in PET image reconstruction, even in situations that correspond to real clinical images where the electron density distribution has local dependency and anisotropy.

[0082] According to at least one of the embodiments described above, image quality can be improved.

[0083] With respect to the above embodiment, the following supplementary notes are disclosed as one aspect and optional features of the invention.

[0084] (Appendix 1) One aspect of the present invention provides a medical image processing apparatus comprising an acquisition unit, a calculation unit, and a reconstruction processing unit. The acquisition unit acquires an electron density function of a subject and information on a nuclide administered to the subject. The calculation unit calculates a positron range kernel for the subject based on the electron density function and the nuclide. The reconstruction processing unit performs PET image reconstruction of the subject based on the positron range kernel.

[0085] (Appendix 2) The calculation section may calculate the positron range kernel for each of a plurality of positions.

[0086] (Appendix 3) The positron range kernel may be anisotropic.

[0087] (Appendix 4) the calculation unit calculates a detection probability matrix based on detector geometry information; calculating a system matrix based on the detection probability matrix and the positron range kernel; The reconstruction processing unit performs PET image reconstruction of the subject based on the system matrix. It is also possible.

[0088] (Appendix 5) The positron range kernel is a matrix representing the probability that a positron generated in one voxel will annihilate in another voxel, the detection probability matrix is a matrix representing the probability that a gamma ray generated in association with a positron annihilated in one voxel is detected by a detector pair associated with one LOR; The system matrix may be a matrix representing the probability that a gamma ray generated by annihilation of the positrons generated in the one voxel is detected by a detector pair associated with the one LOR.

[0089] (Appendix 6) The calculation section may calculate the positron range kernel based on physical calculations.

[0090] (Appendix 7) The calculation unit may calculate the positron range kernel based on a Monte Carlo simulation.

[0091] (Appendix 8) The calculation unit may calculate the positron range kernel based on a trained model that has been trained on a relationship between an electron density function and a probability distribution of annihilation positions of generated positrons.

[0092] (Appendix 9) The acquisition unit may acquire the electron density function based on an X-ray CT image or an MRI image of the subject.

[0093] (Appendix 10) A medical image processing method provided in one aspect of the present invention comprises: obtaining information about the electron density function of the subject and the nuclide administered to the subject; calculating a positron range kernel for the subject based on the electron density function and the nuclide; A PET image of the subject is reconstructed based on the positron range kernel.

[0094] (Appendix 11) A program provided in one aspect of the present invention includes a process of acquiring information about an electron density function of a subject and a nuclide administered to the subject, calculating a positron range kernel for the subject based on the electron density function and the nuclide, and reconstructing a PET image of the subject based on the positron range kernel. Let the computer run it.

[0095] Although several 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 within the scope and spirit of the invention, as well as within the scope of the invention and its equivalents as defined in the claims. [Explanation of symbols]

[0096] 150 Processing Circuit 150a Acquisition Function 150b calculation function 150c reconfiguration function 150d Control Functions 150e Reception Function 150f Image Generation Function 150g Display Control Function 150h learning function

Claims

1. an acquisition unit that acquires information about an electron density function of a subject and a nuclide administered to the subject; a calculation unit that calculates a positron range kernel for the subject, which is a matrix of expansion coefficients when a quantity representing the number of annihilation gamma-ray pairs generated in a voxel is expanded with respect to a quantity representing the number of positrons emitted from the voxel, based on the electron density function and the nuclide; a reconstruction processing unit that performs PET image reconstruction of the subject based on the positron range kernel; Equipped with the calculation unit calculates a detection probability matrix based on detector geometry information; calculating a system matrix by multiplying the detection probability matrix by the positron range kernel; The reconstruction processing unit reconstructs a PET image of the subject based on the system matrix.

2. The medical image processing apparatus according to claim 1 , wherein the calculation unit calculates the positron range kernel for each of a plurality of positions.

3. The positron range kernel is a matrix representing the probability that a positron generated in one voxel will annihilate in another voxel, the detection probability matrix is a matrix representing the probability that a gamma ray pair generated in association with a positron annihilation in one voxel is detected by a detector pair associated with one LOR; 2. The medical image processing apparatus according to claim 1, wherein the system matrix is a matrix representing a probability that a gamma ray generated by annihilation of the positron generated in the one voxel is detected by a detector pair associated with the one LOR.

4. The medical image processing apparatus according to claim 1 , wherein the calculation unit calculates the positron range kernel based on physical calculations.

5. The medical image processing apparatus according to claim 1 , wherein the calculation unit calculates the positron range kernel based on a Monte Carlo simulation.

6. The medical image processing apparatus according to claim 1 , wherein the calculation unit calculates the positron range kernel based on a trained model that has been trained on a relationship between an electron density function and a probability distribution of annihilation positions of generated positrons.

7. The medical image processing apparatus according to claim 1 , wherein the acquisition unit acquires the electron density function based on an X-ray CT image of the subject.

8. obtaining information about the electron density function of the subject and the nuclide administered to the subject; calculating a positron range kernel for the subject, which is a matrix of expansion coefficients when a quantity representing the number of annihilation gamma-ray pairs generated in a voxel is expanded with respect to a quantity representing the number of positrons emitted from the voxel, based on the electron density function and the nuclide; calculating a detection probability matrix based on detector geometric information, calculating a system matrix by multiplying the detection probability matrix by the positron range kernel, and reconstructing a PET image of the subject based on the system matrix; Medical image processing methods.

9. obtaining information about the electron density function of the subject and the nuclide administered to the subject; calculating a positron range kernel for the subject, which is a matrix of expansion coefficients when a quantity representing the number of annihilation gamma-ray pairs generated in a voxel is expanded with respect to a quantity representing the number of positrons emitted from the voxel, based on the electron density function and the nuclide; A process of calculating a detection probability matrix based on detector geometric information, calculating a system matrix by multiplying the detection probability matrix by the positron range kernel, and reconstructing a PET image of the subject based on the system matrix. A program that causes a computer to execute the following.

Images