Radiographic apparatus and radiographic image processing method

Abstract

A radiographic apparatus can suppress a decrease in quality of a generated gamma ray image even when imaging is performed under a low-quality imaging condition, such as a condition in which a radiopharmaceutical (radionuclide) having a low energy of emitted gamma rays is used or a condition in which the number of detectors that detect annihilation gamma rays is reduced. The radiographic apparatus (100) includes a first detector (1) that detects single gamma rays, a plurality of second detectors (2) that detect annihilation gamma rays, an image processing section (3) that generates a first gamma ray image (10) based on the single gamma rays detected by the first detector (1) and generates a second gamma ray image (11) based on the annihilation gamma rays detected by the plurality of second detectors (2), and the image processing section (3) high-qualityizes either one of the first gamma ray image (10) and the second gamma ray image (11) using the first gamma ray image (10) and the second gamma ray image (11).

Description

Technical Field

[0001] This invention relates to a radiographic apparatus and a radiographic image processing method, and more particularly to a radiographic apparatus and a radiographic image processing method for detecting gamma rays to generate gamma ray images. Background Technology

[0002] Previously, there were known radiographic devices for detecting gamma rays to generate gamma-ray images (see, for example, Patent Document 2 and Patent Document 1).

[0003] The ring-shaped PET device (radiography device) disclosed in Patent Document 1 above includes a single-gamma-ray radiation detector. The single-gamma-ray radiation detector has a scattering detector and an absorption detector. The scattering detector causes Compton scattering of the single-gamma rays emitted from the drug (radionucleoside) administered to the subject and incident on the scattering detector. The absorption detector photoelectrically absorbs the scattered rays generated by the Compton scattering. This defines the location of the gamma-ray generation and generates a reconstructed image.

[0004] The nuclear medicine diagnostic apparatus (radiography apparatus) described in Patent Document 2 above comprises a scatterer detector, a PET (Positron Emission Tomography) detector, and a determination unit. The scatterer detection unit detects Compton scattering points by causing Compton scattering of the annihilated gamma rays emitted when positrons emitted from a drug (radioactive nuclide) administered to a subject annihilate a nearby electron pair. Furthermore, the PET detector is configured to detect gamma rays by detecting the scintillation light emitted when the annihilated gamma rays scattered by the scatterer detector interact with a scintillator. The determination unit is configured to generate a gamma-ray image based on the gamma rays detected by the PET detector.

[0005] Existing technical documents Patent documents Patent Document 1: Japanese Patent Application Publication No. 2018-136152 Patent Document 2: Japanese Patent Application Publication No. 2022-061487 Summary of the Invention The technical problem that the invention aims to solve In the structure for detecting a single gamma ray (single gamma ray) disclosed in Patent Document 1, the radiation detector only needs to detect the single gamma ray, thus reducing the number of radiation detectors compared to the structure for detecting annihilated gamma rays disclosed in Patent Document 2. Furthermore, in the structure for detecting single gamma rays to generate a reconstructed image (gamma-ray image), based on the generation principle, a gamma-ray image with a high SNR can be generated even with a reduced number of radiation detectors. However, based on the generation principle, gamma-ray images generated by detecting single gamma rays will produce striped (linear) artifacts, thus reducing the image quality of the gamma-ray image. Moreover, in order to reduce the radiation dose to the subject, when using a reagent (radionoid) with low energy of emitted single gamma rays, the resolution of the generated gamma-ray image is reduced, resulting in a decrease in the image quality of the gamma-ray image.

[0006] Furthermore, in the structure disclosed in Patent Document 2 above, multiple pairs of PET detectors are required to detect annihilated gamma rays. This leads to a tendency for the device to become large-scaled. Therefore, when the number of PET detectors is reduced to mitigate the need for device scaling, the number of annihilated gamma ray detection events by the PET detectors decreases. In this case, statistical noise accumulates, the SNR (Signal Noise Ratio) decreases, and the quality of the generated gamma-ray image deteriorates. Therefore, it is desirable to have a radiographic device that can suppress the deterioration of the generated gamma-ray image quality even when using a reagent (radionoid) that emits low-energy gamma rays, or when reducing the number of PET detectors used to detect annihilated gamma rays, under conditions of deteriorated imaging.

[0007] The present invention was made to solve the aforementioned technical problems. One object of the present invention is to provide a radiographic device and a radiographic image processing method that can suppress the deterioration of the image quality of the generated gamma-ray image even when photographic conditions with reduced image quality are performed, such as when using a reagent (radioactive nuclide) that emits low-energy gamma rays or when the number of detectors that detect annihilated gamma rays is reduced.

[0008] Solution to the above technical problems To achieve the above objectives, the radiographic apparatus of the first aspect of the present invention comprises: a first detector that detects single gamma rays; a plurality of second detectors that detect annihilated gamma rays; and an image processing unit that generates a first gamma-ray image based on the single gamma rays detected by the first detector, and generates a second gamma-ray image based on the annihilated gamma rays detected by the plurality of second detectors. The image processing unit is configured to use the first gamma-ray image and the second gamma-ray image to enhance the quality of either the first gamma-ray image or the second gamma-ray image.

[0009] The radiation image processing method in the second aspect of the present invention includes: a step of generating a first gamma-ray image based on a single gamma ray detected by a first detector; a step of generating a second gamma-ray image based on annihilated gamma rays detected by a plurality of second detectors; and a step of using the first gamma-ray image and the second gamma-ray image to enhance the quality of either the first gamma-ray image or the second gamma-ray image.

[0010] Invention Effects In the radiographic apparatus of the first aspect and the radiographic image processing method of the second aspect described above, as described above, a first gamma-ray image generated based on a single gamma ray and a second gamma-ray image generated using annihilated gamma rays are used to enhance the image quality of either the first gamma-ray image or the second gamma-ray image. Therefore, even when using a reagent (radionoid) with low-energy emitted gamma rays, the image quality of the first gamma-ray image can be enhanced using both the first and second gamma-ray images. Furthermore, even when reducing the number of second detectors leads to a decrease in the image quality of the second gamma-ray image, the image quality of the second gamma-ray image can still be enhanced using both the first and second gamma-ray images. As a result, even when photographing under conditions of reduced image quality, such as using a reagent (radionoid) with low-energy emitted gamma rays or reducing the number of detectors detecting annihilated gamma rays, the reduction in the image quality of the generated gamma-ray image can be suppressed. Attached Figure Description

[0011] 【 Figure 1 [Illustration 1] is a block diagram showing the structure of the radiographic apparatus of the first embodiment.

[0012] 【 Figure 2 [Illustration 1] is a diagram illustrating the configuration of the first detector and the second detector in the first embodiment.

[0013] 【 Figure 3 [Image 1] is a perspective view illustrating the configuration of the first detector and the second detector in the first embodiment.

[0014] 【 Figure 4 The image shown is an example of a SPECT image.

[0015] 【 Figure 5 The image shown is an example of a PET image.

[0016] 【 Figure 6 [Illustration 1] is a diagram illustrating the structure of the image processing unit in the first embodiment that generates the first high-resolution SPECT image.

[0017] 【 Figure 7 [Illustration 1] is a diagram illustrating the structure of the image processing unit in the first embodiment that generates the second high-quality SPECT image.

[0018] 【 Figure 8 This is a flowchart illustrating the process by which the image processing unit of the first embodiment generates the first high-resolution SPECT image.

[0019] 【 Figure 9 This is a flowchart illustrating the process by which the image processing unit of the first embodiment generates the second high-quality SPECT image.

[0020] 【 Figure 10 [Illustration 1] is a block diagram showing the structure of the radiographic apparatus of the second embodiment.

[0021] 【 Figure 11 [Illustration 1] is a diagram illustrating the structure of the image processing unit in the second embodiment that generates a high-quality PET image.

[0022] 【 Figure 12 This is a flowchart illustrating the process by which the image processing unit of the second embodiment generates a high-quality PET image.

[0023] 【 Figure 13 [Illustration 1] is a block diagram showing the structure of the radiographic apparatus of the third embodiment. Detailed Implementation

[0024] Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

[0025] [First Implementation] Reference Figures 1-9 The structure of the radiographic imaging apparatus 100 of the first embodiment will be described.

[0026] (The overall structure of the radiographic imaging device) like Figure 1As shown, the radiography apparatus 100 includes a first detector 1, a plurality of second detectors 2, an image processing unit 3, a control unit 4, a storage unit 5, an input receiving unit 6, and a display unit 7. In the first embodiment, the radiography apparatus 100 includes a plurality of first detectors 1.

[0027] The radiographic device 100 administers a drug containing a radioactive isotope to the subject, based on the subject's body (the subject being imaged 40 (reference)). Figure 5 This is an imaging device that uses radiation (gamma rays) emitted by a radionuclide contained in a drug that is focused onto a tumor. The radiographic device 100 is configured to generate gamma-ray images used for the diagnosis of a patient and for confirming the effectiveness of treatment.

[0028] The first detector 1 is configured to detect single gamma rays. A single gamma ray is a single gamma ray emitted from a drug (radioactive nuclide) administered to the subject. Furthermore, the energy of the single gamma ray varies depending on the drug (radioactive nuclide) administered to the subject. In the first embodiment, the single gamma ray includes a first single gamma ray and a second single gamma ray with different energies. The first detector 1 is configured to detect both the first and second single gamma rays separately. Details regarding the first detector 1 will be described later.

[0029] Multiple second detectors 2 are configured to detect annihilated gamma rays. Details about the second detectors 2 will be described later.

[0030] Annihilation gamma rays are a pair of gamma rays emitted when a positron emitted from a drug (radioactive nuclide) administered to a subject annihilates with a nearby electron. The pair of gamma rays are emitted in opposite directions.

[0031] The image processing unit 3 is configured to generate a first gamma-ray image 10 based on a single gamma ray detected by the first detector 1. Furthermore, the image processing unit 3 is configured to generate a second gamma-ray image 11 based on annihilated gamma rays detected by a plurality of second detectors 2. The image processing unit 3 may include, for example, a GPU (Graphics Processing Unit) or a FPGA (Field-Programmable Gate Array) configured for image processing.

[0032] In the first embodiment, the first gamma-ray image 10 is a SPECT image 12 based on a single gamma ray detected in the first detector 1. Furthermore, the second gamma-ray image 11 is a PET image 13 based on annihilated gamma rays detected in a plurality of second detectors 2.

[0033] In the first embodiment, the SPECT image 12 includes a first SPECT image 12a generated based on a first single gamma ray (see reference). Figure 6 ), and the second SPECT image 12b generated based on the second single gamma ray (see reference). Figure 7 ).

[0034] The control unit 4 is configured to control various parts of the radiography apparatus 100 by executing various programs (not shown) stored in the storage unit 5. The control unit 4 includes, for example, a processor such as a CPU (Central Processing Unit) and a memory such as ROM (Read Only Memory) and RAM (Random Access Memory).

[0035] Storage unit 5 stores various programs executed by control unit 4. Furthermore, storage unit 5 stores SPECT image 12 and PET image 13 generated by image processing unit 3. Additionally, storage unit 5 stores a first high-resolution SPECT image 20 (described later) and a second high-resolution SPECT image 21 (described later). Furthermore, storage unit 5 stores a learning model 30 (described later). Furthermore, storage unit 5 stores a distribution 31 of the line attenuation coefficient (described later). Storage unit 5 includes a non-volatile storage device such as HDD (Hard Disk Drive) or SSD (Solid State Drive).

[0036] In the first embodiment, the learning model 30 is a non-learning deep learning model. The learning model 30 is, for example, a deep image prior network. The learning model 30 includes a first learning model 30a and a second learning model 30b. The first learning model 30a is a non-learning deep learning model used when generating the first high-resolution SPECT image 20 (described later). Furthermore, the second learning model 30b is a non-learning deep learning model used when generating the second high-resolution SPECT image 21 (described later).

[0037] The input reception unit 6 is configured to receive operation input from the operator. The input reception unit 6 includes, for example, input devices such as a mouse and keyboard.

[0038] Display unit 7 is configured to display SPECT image 12 and PET image 13, etc. Display unit 7 includes, for example, a liquid crystal monitor or an organic EL (Electro Luminescence) monitor and other display devices.

[0039] (Detector 1 and Detector 2) Next, refer to Figure 2 as well as Figure 3 The structures of the first detector 1 and the second detector 2 will be described.

[0040] like Figure 2 As shown, a first detector 1 and a plurality of second detectors 2 are disposed inside a housing 8. The housing 8 has an opening 8a. The opening 8a is an opening on one side of a through hole extending along axis 90 through the housing 8, and an opening on the other side. (See reference 100 for details.) Figure 1 When photographing the subject (photographic object 40), the subject is placed inside the through hole of the housing 8. Furthermore, axis 90 is a virtual axis and does not actually exist at the center of the opening 8a.

[0041] like Figure 2 As shown, the number of detectors 1 is less than the number of detectors 2.

[0042] In addition, such as Figure 2 As shown, multiple first detectors 1 are arranged at predetermined intervals around axis 90 on a circumferential surface. Figure 2 In the example shown, ten first detectors 1 are arranged at predetermined intervals in a circular pattern centered on axis 90. Each of the plurality of first detectors 1 has a scatterer detector 1a and an absorber detector 1b.

[0043] The scatterer detector 1a is configured to cause Compton scattering of a single gamma ray emitted from the photographed object 40. Furthermore, the scatterer detector 1a is configured to detect the location (scattering point) where Compton scattering occurs. The scatterer detector 1a is, for example, a scintillator. Additionally, a collimator (not shown) for limiting the direction of the single gamma ray is provided on the front surface of the scatterer detector 1a (on the center side of the opening 8a).

[0044] The absorber detector 1b is configured to detect scattered light produced by Compton scattering of a single gamma ray at the scatterer detector 1a. Specifically, the absorber detector 1b is configured to detect scintillation light produced by the scattered light generated by Compton scattering. The absorber detector 1b is, for example, a photomultiplier tube.

[0045] The image processing unit 3 determines the emission location of a single gamma ray based on the scattering point of Compton scattering detected in the scatterer detector 1a and the scintillation light absorbed by the absorber detector 1b, and generates a SPECT image 12 (see reference). Figure 1In the first embodiment, the image processing unit 3 generates the SPECT image 12, for example, by a filtered back-projection method that back-projects the projection data of a single gamma ray detected in each of the plurality of first detectors 1.

[0046] Furthermore, multiple second detectors 2 are arranged at predetermined intervals circumferentially around axis 90. Figure 2 In the example shown, 30 second detectors are arranged at predetermined intervals in a circular pattern, centered on axis 90.

[0047] Each of the plurality of second detectors 2 is configured to detect annihilated gamma rays emitted from the photographed object 40. Each of the plurality of second detectors 2 includes a scintillator (not shown), a light detection element (not shown), and a light guide (not shown).

[0048] The scintillator converts incident annihilated gamma rays into scintillation light. Furthermore, a photodetector detects the scintillation light converted from the annihilated gamma rays by the scintillator. The photodetector is, for example, a photomultiplier tube. Additionally, a light guide is configured to allow the scintillation light converted from the annihilated gamma rays by the scintillator to be incident on the photodetector. The light guide is, for example, made of resin or glass, and silicone oil. Alternatively, each of the second detectors 2 may not have a light guide.

[0049] The image processing unit 3 determines the emission location of the annihilated gamma rays based on the scintillation light detected simultaneously or within a specified time range by the photodetector element, and generates a PET image 13 (see reference). Figure 2 In the first embodiment, the image processing unit 3 generates a PET image 13, for example, by an iterative image reconstruction method that compares a virtual reconstructed image with projection data of annihilated gamma rays and corrects the virtual reconstructed image to make the virtual reconstructed image match the projection data of annihilated gamma rays.

[0050] In addition, such as Figure 3 As shown, multiple first detectors 1 and multiple second detectors 2 are arranged at different positions along the axis 90.

[0051] (SPECT and PET images) Next, refer to Figure 4 as well as Figure 5 The SPECT image 12 and PET image 13 are described below.

[0052] Figure 4 The SPECT image 12 shown is obtained by, for example, using a filtered back projection method on multiple first detectors 1 (see reference). Figure 2The image is generated by reconstructing the projection data of each detected annihilated gamma ray in the SPECT image. When reconstructed using filtered back projection, the resolution of SPECT image 12 decreases, and the outline of the photographed object 40 becomes blurred. Figure 4 In the example shown, the blurred outline of the photographic object 40 is illustrated by using dashed lines. Furthermore, in the SPECT image 12, striped (linear) artifacts 41 sometimes occur during reconstruction. On the other hand, statistical noise is difficult to generate in the SPECT image 12. That is, the SPECT image 12 is an image with high SNR but low resolution of the photographic object 40. Additionally, the photographic object 40 may be, for example, a tumor in the patient.

[0053] Figure 5 The PET image 13 shown is generated by reconstructing the annihilated gamma rays detected in each of the plurality of second detectors 2 using, for example, an iterative image reconstruction method. In PET image 13, even with a reduced number of second detectors 2, blurring and striped artifacts 41 are difficult to produce at the contours of the photographed object 40. However, with a reduced number of second detectors 2, statistical noise is superimposed in PET image 13. That is, PET image 13 is a high-resolution image of the photographed object 40 but with a low SNR. Furthermore, in Figure 5 In the example shown, the superposition of statistical noise is illustrated by the addition of profile lines.

[0054] Therefore, in the first embodiment, the image processing unit 3 (refer to...) Figure 1 The image processing unit 3 is configured to use the first gamma-ray image 10 and the second gamma-ray image 11 to enhance the quality of the first gamma-ray image 10. Specifically, the image processing unit 3 is configured to use the first gamma-ray image 10 and the second gamma-ray image 11, as well as the learning model 30 for enhancing the quality of the gamma-ray image, to enhance the quality of the first gamma-ray image 10.

[0055] In the first embodiment, the image processing unit 3 uses the SPECT image 12, the PET image 13 which contains more noise but has a higher resolution than the SPECT image 12, and the learning model 30 to generate a high-quality SPECT image 12 by increasing the resolution.

[0056] (Generation of the first high-resolution SPECT image) Figure 6The example shown illustrates a case where, for instance, during the diagnosis of a subject, an agent (radionium) capable of simultaneously generating SPECT image 12 and PET image 13 is administered to the subject for imaging. The agent (radionium) capable of simultaneously generating SPECT image 12 and PET image 13 is, for example, an 18F agent containing gamma rays emitting 511 keV.

[0057] like Figure 6 As shown, the image processing unit 3 inputs the first SPECT image 12a and the PET image 13 into the first learning model 30a. The image processing unit 3 is configured to use the first learning model 30a, which is a non-learning deep learning model, to perform iterative calculations starting from a random image (not shown) to approximate the input first gamma-ray image 10 (first SPECT image 12a), thereby enhancing the resolution of the first gamma-ray image 10 to achieve high-quality processing. In the first embodiment, the image processing unit 3 is configured to use the SPECT image 12 (first SPECT image 12a) and the PET image 13 to generate a first high-quality SPECT image 20, which has been enhanced by increasing the resolution of the SPECT image 12.

[0058] like Figure 6 As shown, compared to the first SPECT image 12a, the blurring of the outline of the photographed object 40 is reduced in the first high-resolution SPECT image 20. Furthermore, compared to the first SPECT image 12a, the striped artifacts 41 are suppressed in the first high-resolution SPECT image 20. Moreover, compared to the PET image 13, noise superposition is suppressed in the first high-resolution SPECT image 20. In other words, the first high-resolution SPECT image 20 has a higher resolution than the first SPECT image 12a and a higher SNR than the PET image 13.

[0059] (Second high-resolution SPECT image) Figure 7 The example shown illustrates the structure of the image processing unit 3 high-resolution SPECT image 12 when imaging is performed to confirm the treatment effect on the patient. When confirming the treatment effect on the patient, a different agent (radioactive nuclide) is used than that used during diagnosis. For example, when confirming the treatment effect on the patient, an image containing gamma rays emitting 208 keV is used. 177 Lu's medicine.

[0060] In the first embodiment, the image processing unit 3 is configured to generate a second high-quality SPECT image 21 that has been enhanced by increasing the resolution of the second SPECT image 12b using the second SPECT image 12b and the first high-quality SPECT image 20.

[0061] In the first embodiment, the image processing unit 3 inputs the second SPECT image 12b and the first high-resolution SPECT image 20 into the second learning model 30b. Then, the image processing unit 3 performs iterative calculations starting from a random image (not shown) to make it approximate the input second SPECT image 12b, thereby generating a second high-resolution SPECT image 21 that has been improved by increasing the resolution of the second SPECT image 12b.

[0062] like Figure 7 As shown, compared to the second SPECT image 12b, the blurring of the outline of the photographic object 40 is reduced in the second high-resolution SPECT image 21. Furthermore, compared to the second SPECT image 12b, the striped artifacts 41 are suppressed in the second high-resolution SPECT image 21. That is, the second high-resolution SPECT image 21 is a higher resolution image than the first SPECT image 12a.

[0063] Furthermore, in cases where treatment is conducted over a long period, the treatment effect on the subject may sometimes be confirmed multiple times. Therefore, in the first embodiment, the image processing unit 3 generates multiple second SPECT images 12b based on second single gamma rays detected at different times. Then, when the image processing unit 3 upscales the first second SPECT image 12b among the multiple second SPECT images 12b, it uses a first upscaled SPECT image 20 to improve the resolution of the second SPECT image 12b. Furthermore, when the image processing unit 3 upscales the second and subsequent second SPECT images 12b among the multiple second SPECT images 12b, it uses a second upscaled SPECT image 21 generated from the previous second SPECT image 12b to improve the resolution of the second SPECT image 12b.

[0064] (Absorption Correction) Here, the first gamma-ray image 10 and the second gamma-ray image 11 are generated based on gamma rays emitted from the body of the subject (the photographic object 40). The gamma rays emitted from the body of the subject are absorbed within the subject's body. Therefore, if the gamma rays absorbed within the subject's body are not considered, the image quality of the generated image is reduced. Therefore, in the first embodiment, the image processing unit 3 is configured to perform a distribution 31 based on a pre-generated linear attenuation coefficient (see reference 1) on at least one of the first gamma-ray image 10 and the second gamma-ray image 11. Figure 1Absorption correction is performed on both the first gamma-ray image 10 and the second gamma-ray image 11, and the image with the absorption correction is used to achieve high image quality. Specifically, the image processing unit 3 is configured to perform absorption correction on both the first gamma-ray image 10 and the second gamma-ray image 11. The distribution 31 of the linear attenuation coefficient is prepared in advance by performing X-ray CT imaging on the photographic object 40 or a phantom that imitates the photographic object 40.

[0065] (Generation and processing of the first high-resolution SPECT image) Next, refer to Figure 8 For image processing unit 3 (refer to) Figure 1 Generate the first high-resolution SPECT image 20 (refer to...) Figure 6 The processing of ) will be explained.

[0066] In step 101, the image processing unit 3 generates a first gamma-ray image 10 based on single gamma rays detected by the plurality of first detectors 1. Specifically, the image processing unit 3 generates a first SPECT image 12a (refer to...) based on single gamma rays detected by the plurality of first detectors 1. Figure 6 ).

[0067] Next, in step 102, the image processing unit 3 generates a second gamma-ray image 11 based on the annihilated gamma rays detected by the plurality of second detectors 2. In the first embodiment, the image processing unit 3 generates a PET image 13 (see reference 102) based on the annihilated gamma rays detected by the plurality of second detectors 2. Figure 6 ).

[0068] Next, in step 103, the image processing unit 3 performs absorption correction on each of the first SPECT image 12a and the PET image 13. In the first embodiment, the image processing unit 3 performs absorption correction on each of the first SPECT image 12a and the PET image 13 based on the distribution 31 of the line attenuation coefficient, and then outputs each image again.

[0069] Next, in step 104, the image processing unit 3 uses the first gamma-ray image 10 and the second gamma-ray image 11 to enhance the quality of the first gamma-ray image 10. In the first embodiment, the image processing unit 3 generates a first high-quality SPECT image 20, which enhances the quality of the first SPECT image 12a, by inputting the absorption-corrected first SPECT image 12a and the absorption-corrected PET image 13 into the first learning model 30a. After that, the processing ends.

[0070] Furthermore, either step 101 or step 102 can be performed in any order.

[0071] (Generation and processing of the second high-resolution SPECT image) Next, refer to Figure 9 For image processing unit 3 (refer to) Figure 1 Generate a second high-resolution SPECT image 21 (refer to...) Figure 7 The processing of the first high-resolution SPECT image 20 will be explained. Additionally, for processing identical to that used in generating the first high-resolution SPECT image 20 described above, the same reference numerals will be used, and detailed descriptions will be omitted.

[0072] In step 110, the image processing unit 3 generates a second SPECT image 12b. Specifically, the image processing unit 3 generates a second high-resolution SPECT image 21 based on a second single gamma ray.

[0073] Next, in step 103, the image processing unit 3 performs absorption correction on the second SPECT image 12b.

[0074] Next, in step 111, the image processing unit 3 acquires the first high-resolution SPECT image 20. In the first embodiment, the image processing unit 3 acquires from the storage unit 5 images that have been pre-generated and stored in the storage unit 5, such as during the patient's diagnosis (see reference 1). Figure 1 The first high-resolution SPECT image 20 in the image.

[0075] Next, in step 112, the image processing unit 3 generates a second high-resolution SPECT image 21. Specifically, the image processing unit 3 inputs the second SPECT image 12b and the first high-resolution SPECT image 20 into the second learning model 30b (refer to...). Figure 7 The process involves iterative calculations to generate a second high-resolution SPECT image 21. Afterward, the processing ends.

[0076] Furthermore, the processing of steps 110 and 103, or the processing of step 111, can be performed in any order.

[0077] [Second Implementation] Next, refer to Figures 10-12 The second embodiment will now be described. In the radiography apparatus 200 of the second embodiment, the structure by which the image processing unit 201 renders the PET image 13 in high quality will be described. Furthermore, for structures identical to those in the first embodiment described above, the same reference numerals will be used, and detailed descriptions will be omitted.

[0078] (The overall structure of the radiographic imaging device) like Figure 10 As shown, the radiography apparatus 200 of the second embodiment includes a plurality of first detectors 1, a plurality of second detectors 2, a control unit 4, a storage unit 5, an input receiving unit 6, a display unit 7, and an image processing unit 201.

[0079] In the second embodiment, the storage unit 5 stores the first gamma-ray image 10 (SPECT image 12), the second gamma-ray image 11 (PET image 13), the high-resolution PET image 22 (described later), the distribution of the line attenuation coefficient 31, and the learning model 32 used when generating the high-resolution PET image 22.

[0080] In the second embodiment, the image processing unit 201 is configured to enhance the quality of the second gamma-ray image 11 using the first gamma-ray image 10 and the second gamma-ray image 11. Specifically, the image processing unit 201 is configured to enhance the quality of the second gamma-ray image 11 using the first gamma-ray image 10, the second gamma-ray image 11, and the learning model 32 for enhancing the quality of the gamma-ray image. More specifically, the image processing unit 201 uses the SPECT image 12, the PET image 13, and the learning model 32 to generate the enhanced PET image 13 by reducing noise.

[0081] like Figure 11 As shown, the image processing unit 201 of the second embodiment (refer to...) Figure 10 The image processing unit 201 is configured to use a learning model 32, which is a non-learning deep learning model, to iteratively compute from a random image to make it approximate the input second gamma-ray image 11 (PET image 13), thereby improving the quality of the second gamma-ray image 11 by reducing the noise of the second gamma-ray image 11. Specifically, the image processing unit 201 is configured to use the SPECT image 12 and the PET image 13 to generate a high-quality PET image 22 that has been improved by reducing the noise of the PET image 13.

[0082] like Figure 11 As shown, compared to PET image 13, noise superposition is suppressed in the high-quality PET image 22. Furthermore, compared to SPECT image 12, the blurring of the outline of the photographic object 40 is reduced in the high-quality PET image 22. Additionally, compared to SPECT image 12, striped artifacts 41 are suppressed in the high-quality PET image 22. In other words, the high-quality PET image 22 has a higher SNR than PET image 13 and a higher resolution than SPECT image 12.

[0083] (High-quality PET image generation and processing) Next, refer to Figure 12 For image processing unit 201 (refer to) Figure 10 Generate high-quality PET images 22 (refer to) Figure 10 The processing of the image processing unit 3 in the first embodiment described above will be explained. Additionally, the processing of the image processing unit 3 in the first embodiment described above (see [reference]) will be explained. Figure 1 Generate the first high-resolution SPECT image 20 (refer to...) Figure 6 The same processing is applied to the same parts, with the same reference numerals attached, and detailed descriptions omitted.

[0084] In steps 102 and 101, the image processing unit 201 generates SPECT image 12 and PET image 13. Furthermore, in step 103, the image processing unit 201 performs absorption correction on each of the SPECT image 12 and PET image 13. Additionally, the processing in steps 102 and 101 can be performed either first.

[0085] Next, in step 210, the image processing unit 201 generates a high-quality PET image 22. Specifically, the image processing unit 201 inputs the absorption-corrected PET image 13 and the SPECT image 12 into the learning model 32 to generate the high-quality PET image 22. After that, the processing ends.

[0086] Furthermore, the other structures of the second embodiment are the same as those of the first embodiment described above.

[0087] [Third Implementation] Next, refer to Figure 13 The third embodiment will now be described. In the radiography apparatus 300 of the third embodiment, the structure by which the image processing unit 301 uses the learning model 33 to enhance the SPECT image 12 to high quality will be described. Furthermore, for structures identical to those in the first embodiment described above, the same reference numerals will be used, and detailed descriptions will be omitted.

[0088] (The overall structure of the radiographic imaging device) like Figure 13 As shown, the radiography apparatus 300 of the third embodiment includes a plurality of first detectors 1, a plurality of second detectors 2, a control unit 4, a storage unit 5, an input receiving unit 6, a display unit 7, and an image processing unit 301.

[0089] In the third embodiment, the storage unit 5 stores the first gamma-ray image 10 (SPECT image 12), the second gamma-ray image 11 (PET image 13), the first high-resolution SPECT image 20, the second high-resolution SPECT image 21, the distribution of the line attenuation coefficient 31, and the learning model 33 used when generating the first high-resolution SPECT image 20 and the second high-resolution SPECT image 21.

[0090] Learning model 33 is a learning-oriented deep learning model. Learning model 33 may include, for example, Convolutional Neural Networks (CNNs) or U-nets.

[0091] The learning model 33 was generated by learning to upscale the SPECT image 12 using training data.

[0092] Learning model 33 includes learning model 33a (first learning model) and learning model 33b (second learning model).

[0093] The first learning model 33a is generated by learning to upscale the SPECT image 12 (the first SPECT image 12a) using SPECT image 12 (the first SPECT image 12a) and PET image 13. Furthermore, the second learning model 33b is generated by learning to upscale the SPECT image 12 (the second SPECT image 12b) using SPECT image 12 (the second SPECT image 12b) and the first upscaled SPECT image 20.

[0094] In the third embodiment, the image processing unit 301 is configured to use a learning model 33 that has learned to improve the resolution of the first gamma-ray image 10, thereby improving the resolution of the first gamma-ray image 10 to achieve high image quality.

[0095] Specifically, the image processing unit 301 uses the first SPECT image 12a, the PET image 13, and the first learning model 33a to generate a first high-quality SPECT image 20 that has been enhanced from the first SPECT image 12a.

[0096] Furthermore, the image processing unit 301 uses the second SPECT image 12b, the first high-resolution SPECT image 20, and the second learning model 33b to generate a second high-resolution SPECT image 21 that has been made high-resolution from the second SPECT image 12b.

[0097] Furthermore, the other structures of the third embodiment are the same as those of the first embodiment described above.

[0098] [Variation Example] Furthermore, it should be considered that the embodiments disclosed herein are illustrative in all respects and not restrictive. The scope of the invention is shown by the claims rather than by the description of the above embodiments, and includes all modifications (variations) within the scope and equivalent meaning of the claims.

[0099] For example, the image processing unit may not use a learning model, but instead upscale the first gamma-ray image (SPECT image) or the second gamma-ray image (PET image). In this case, the image processing unit can upscale the SPECT image or the PET image by performing a weighted average of the SPECT image and the PET image.

[0100] Furthermore, the number of the first detector is merely an example and may be a number other than that shown in the first embodiment. There may also be only one first detector. In this case, it can be configured such that the first detector rotates about an axis centered on the opening and simultaneously detects a single gamma ray.

[0101] Furthermore, the number of the second detector is merely an example, and there could be multiple detectors other than the number shown in the first embodiment.

[0102] Furthermore, the image processing unit only needs to generate at least the first high-resolution SPECT image, and may not need to generate the second high-resolution SPECT image.

[0103] In addition, when generating the second high-resolution SPECT image, the image processing unit can also use the second high-resolution SPECT image generated from the first two second SPECT images.

[0104] In addition, the image processing unit can also perform absorption correction during the reconstruction of the first gamma-ray image and the reconstruction of the second gamma-ray image.

[0105] In addition, the image processing unit can also perform scattering correction along with absorption correction.

[0106] Furthermore, the image processing unit may not perform absorption correction.

[0107] In addition, the image processing unit can also use a learning-type deep learning model, i.e., a learning model, that has learned to reduce the noise of the second gamma-ray image, to improve the image quality of the second gamma-ray image (PET image) by reducing the noise.

[0108] Alternatively, a structure combining the first and second embodiments is also possible. That is, the image processing unit can also be a structure that uses the first gamma-ray image and the second gamma-ray image, and is capable of upscaling the first gamma-ray image and the second gamma-ray image.

[0109] Furthermore, although the processes performed by the image processing unit have been described using a flow-driven flowchart that describes the processes sequentially along the processing flow, the present invention is not limited thereto. In the present invention, the processes performed by the image processing unit can also be performed using an event-driven processing method that executes processing on an event-by-event basis. In this case, it can be performed entirely using an event-driven method, or it can be performed by combining event-driven and flow-driven methods.

[0110] [Way] Those skilled in the art will understand that the above exemplary embodiments are specific examples of the following approaches.

[0111] (Project 1) A radiographic device comprising: a first detector that detects a single gamma ray; Multiple second detectors, which detect annihilated gamma rays; and The image processing unit generates a first gamma-ray image based on the single gamma ray detected by the first detector, and generates a second gamma-ray image based on the pair of annihilated gamma rays detected by the plurality of second detectors. The image processing unit is configured to use the first gamma-ray image and the second gamma-ray image to enhance the quality of either the first gamma-ray image or the second gamma-ray image.

[0112] The quality of either the first gamma-ray image generated from a single gamma ray or the second gamma-ray image generated from annihilated gamma rays is enhanced using both images. Therefore, even when using a low-energy reagent (radionoid) that emits gamma rays, the quality of the first gamma-ray image can be enhanced using both the first and second gamma-ray images. Furthermore, even when reducing the number of second detectors leads to a decrease in the quality of the second gamma-ray image, the quality of the second gamma-ray image can still be enhanced using both the first and second gamma-ray images. As a result, even when photography is performed under conditions of reduced image quality, such as using a low-energy reagent (radionoid) that emits gamma rays or reducing the number of detectors detecting annihilated gamma rays (the second detector), the reduction in the quality of the generated gamma-ray image can be suppressed.

[0113] (Project 2) The radiographic apparatus as described in Project 1, wherein the image processing unit is configured to use the first gamma-ray image and the second gamma-ray image, and a learning model for improving the quality of the gamma-ray image, to improve the quality of either the first gamma-ray image or the second gamma-ray image.

[0114] Because a learning model can be used to enhance the quality of either the first or second gamma-ray image, it is easier to enhance the quality of either the first or second gamma-ray image compared to a structure that enhances the quality of the first or second gamma-ray image through image processing without using a learning model. As a result, even when the quality of the first and second gamma-ray images is low, a radiographic apparatus capable of easily enhancing the quality of either the first or second gamma-ray image can be provided.

[0115] (Project 3) The radiographic apparatus as described in Project 2, wherein the first gamma-ray image is a SPECT image based on the single gamma ray detected in the first detector. The second gamma-ray image is based on the PET image of the annihilated gamma rays detected in the plurality of second detectors. The image processing unit uses the SPECT image, the PET image (which contains more noise but has higher resolution compared to the SPECT image), and the learning model to generate either a high-quality SPECT image by increasing resolution or a high-quality PET image by reducing noise.

[0116] Even when using reagents with low-energy gamma-ray emission (radioisotopes), SPECT images with improved resolution can be generated. Furthermore, even when the number of secondary detectors is reduced, PET images with reduced noise can be generated. As a result, even when taking photographs under conditions of reduced image quality, such as using reagents with low-energy gamma-ray emission (radioisotopes) or reducing the number of detectors detecting annihilated gamma rays (secondary detectors), it is possible to generate high-quality SPECT images through increased resolution or high-quality PET images through reduced noise.

[0117] (Project 4) The radiographic apparatus as described in Project 3, wherein the image processing unit is configured to generate a first high-quality SPECT image by increasing the resolution of the SPECT image using the SPECT image and the PET image.

[0118] Even when using a low-energy gamma-ray-emitting agent (radioactive nuclide), it is possible to generate a first-class high-resolution SPECT image. As a result, a radiographic device can be provided that can generate a first-class high-resolution SPECT image even when using a low-energy gamma-ray-emitting agent (radioactive nuclide). Furthermore, since a first-class high-resolution SPECT image can be generated even when using a low-energy gamma-ray-emitting agent (radioactive nuclide), it is possible to generate first-class high-resolution SPECT images that allow doctors and others to accurately determine the location and size of tumors. As a result, the burden on the patient can be reduced.

[0119] (Project 5) The radiographic apparatus described in Project 4, wherein the single gamma rays comprise a first single gamma ray and a second single gamma ray with different energies. The first detector is configured to detect the first single gamma ray and the second single gamma ray, respectively. The SPECT image includes a first SPECT image generated based on the first single gamma ray and a second SPECT image generated based on the second single gamma ray. The image processing unit is configured to generate a second high-quality SPECT image by using the second SPECT image and the first high-quality SPECT image, which is high-qualityd by increasing the resolution of the second SPECT image.

[0120] Here, the radionuclide used for imaging during diagnosis and the radionuclide used for imaging to confirm the treatment effect are sometimes different. In this case, the energy of the first single gamma ray and the energy of the second single gamma ray are different. For example, during diagnosis, when imaging is performed using a radionuclide that emits a single gamma ray and annihilates the gamma ray at an energy detectable by both the first and second detectors, a PET image can be used to upscale the first SPECT image generated using the first single gamma ray. On the other hand, when confirming the treatment effect, sometimes a radionuclide that emits gamma rays with an energy capable of generating a SPECT image but not a PET image is used. In this case, a PET image cannot be used to upscale the second SPECT image generated based on the second single gamma ray. Therefore, by configuring the second SPECT image as described above, by using a first high-resolution SPECT image that is a high-resolution version of the first SPECT image and the PET image generated during diagnosis, a second high-resolution SPECT image can be generated even when the energy of the gamma rays emitted by the agent (radioactive nuclide) used to confirm the treatment effect on the subject is not enough to generate a PET image.

[0121] (Project 6) As described in Project 5, in the radiographic apparatus, the image processing unit generates multiple second SPECT images based on the second single gamma rays detected at different times. When upscaling the first of the plurality of second SPECT images, the resolution of the second SPECT image is improved using the first upscaled SPECT image. When upgrading the second and subsequent second SPECT images among the plurality of second SPECT images, the resolution of the second SPECT image is improved by using the second upgraded SPECT image generated from the previous second SPECT image.

[0122] In cases where repeated radiographic imaging is performed to confirm the treatment effect on the subject, the subject may sometimes change between the second high-resolution SPECT image (which is not the most recent, and the first two or more images) and the most recent second SPECT image. In such cases, even if an attempt is made to improve the resolution of the most recent second SPECT image using a second high-resolution SPECT image that is not the most recent, the resolution of the second SPECT image may not improve due to the difference in the subject. Therefore, as described above, by using the most recent (first one) second high-resolution SPECT image to improve the resolution of the most recent second SPECT image when high-resolution imaging of the second and subsequent second SPECT images, it is possible to suppress the use of second high-resolution SPECT images (which are not the most recent, and the first two or more images) to improve the resolution of the most recent second SPECT image. As a result, the resolution of the most recent second SPECT image can be improved with greater precision.

[0123] (Project 7) The radiographic apparatus as described in Project 3, wherein the image processing unit is configured to generate a high-quality PET image by reducing noise in the PET image using the SPECT image and the PET image.

[0124] Even with a reduced number of second detectors, high-quality PET images with reduced noise can still be generated. As a result, a radiographic apparatus is provided that can generate high-quality PET images with reduced noise even with a reduced number of second detectors for detecting annihilated gamma rays.

[0125] (Project 8) The radiographic apparatus as described in Project 2 or 3, wherein the learning model is a non-learning deep learning model. The image processing unit is configured to use the learning model, which is a non-learning deep learning model, to perform iterative calculations starting from a random image to make it approximate the input first gamma-ray image or the second gamma-ray image, thereby improving the resolution of the first gamma-ray image to enhance its quality, or improving the noise of the second gamma-ray image to enhance its quality.

[0126] In the case of a non-learning deep learning model, since iterative computation is performed starting from a random image to approximate the first or second gamma-ray image, it is possible to enhance the quality of the first or second gamma-ray image regardless of the part of the photographic subject within the first or second gamma-ray image. As a result, since the first or second gamma-ray image can be enhanced using a highly generalized learning model, the user burden required to generate the learning model is reduced compared to a structure that generates a dedicated learning model for each part of the photographic subject to enhance the first or second gamma-ray image.

[0127] (Project 9) As described in Project 2 or 3, the radiographic device wherein the learning model is a learning-based deep learning model. The image processing unit is configured to use the learning model that learns to improve the resolution of the first gamma-ray image or reduce the noise of the second gamma-ray image, thereby improving the resolution of the first gamma-ray image to a higher quality, or improving the noise of the second gamma-ray image to a higher quality.

[0128] When the learning model is a learning-based deep learning model, since the learning model is generated by learning from parts of the photographic object captured in the first or second gamma-ray image, it is possible to use a learning model that specializes in parts of the photographic object to enhance the image quality of the first or second gamma-ray image. As a result, it is possible to further enhance the image quality of the first or second gamma-ray image.

[0129] (Project 10) The radiographic apparatus as described in any one of items 1 to 9, wherein the image processing unit is configured to perform absorption correction on at least one of the first gamma-ray image and the second gamma-ray image based on a pre-generated distribution of linear attenuation coefficients, and to perform high-quality enhancement using the absorption-corrected image.

[0130] Here, gamma rays emitted from the subject's body (the subject of the photograph) are absorbed within the subject's body. The degree of gamma ray absorption is uneven depending on the distance from the emission point to the body surface. Therefore, by correcting the absorption of at least one of the first or second gamma ray images using the distribution of the linear attenuation coefficient, the unevenness of gamma ray absorption can be eliminated, and the distribution of gamma rays emitted from the subject's body can be obtained with high accuracy. As a result, since the first or second gamma ray image before high-quality enhancement can be generated with high accuracy, the first or second gamma ray image can be further enhanced to higher quality.

[0131] (Project 11) A radiation image processing method, comprising: The step of generating a first gamma-ray image based on a single gamma ray detected by the first detector; The steps of generating a second gamma-ray image based on the annihilated gamma rays detected by multiple second detectors; and The step of using the first gamma-ray image and the second gamma-ray image to enhance either the first gamma-ray image or the second gamma-ray image to a higher quality.

[0132] Similar to radiographic devices, a radiographic image processing method can be provided that can suppress the degradation of image quality of the generated gamma-ray image even when photographic conditions with degraded image quality are performed, such as when a reagent (radionoid) that emits low-energy gamma rays is used, or when the number of detectors that detect annihilated gamma rays (second detector) is reduced.

[0133] Explanation of reference numerals in the attached figures 1. Detector No. 1 2. Second detector Image Processing Departments 3, 201, and 301 10. First Gamma-ray Image 11. Second gamma-ray image 12 SPECT images 12a First SPECT image 12b Second SPECT image 13 PET images 20 High-resolution SPECT images 21. Second high-resolution SPECT image 22 High-resolution PET images Learning models 30, 32, and 33 Distribution of line attenuation coefficient 31 100, 200, 300 X-ray imaging devices.

Claims

1. A radiographic imaging device, comprising: The first detector detects a single gamma ray; Multiple second detectors, which detect annihilated gamma rays; and The image processing unit generates a first gamma-ray image based on the single gamma ray detected by the first detector, and generates a second gamma-ray image based on the pair of annihilated gamma rays detected by the plurality of second detectors. The image processing unit is configured to use the first gamma-ray image and the second gamma-ray image to enhance the quality of either the first gamma-ray image or the second gamma-ray image.

2. The radiographic apparatus as claimed in claim 1, wherein, The image processing unit is configured to use the first gamma-ray image and the second gamma-ray image, as well as a learning model that enhances the quality of the gamma-ray image, to enhance either the first gamma-ray image or the second gamma-ray image.

3. The radiographic apparatus as claimed in claim 2, wherein, The first gamma-ray image is a SPECT image based on the single gamma ray detected in the first detector. The second gamma-ray image is based on the PET image of the annihilated gamma rays detected in the plurality of second detectors. The image processing unit uses the SPECT image, the PET image (which contains more noise but has higher resolution compared to the SPECT image), and the learning model to generate either a high-quality SPECT image by increasing resolution or a high-quality PET image by reducing noise.

4. The radiographic apparatus as claimed in claim 3, wherein, The image processing unit is configured to generate a first high-quality SPECT image by using the SPECT image and the PET image, thereby increasing the resolution of the SPECT image.

5. The radiographic apparatus as claimed in claim 4, wherein, The single gamma ray includes a first single gamma ray and a second single gamma ray with different energies. The first detector is configured to detect the first single gamma ray and the second single gamma ray, respectively. The SPECT image includes a first SPECT image generated based on the first single gamma ray and a second SPECT image generated based on the second single gamma ray. The image processing unit is configured to generate a second high-quality SPECT image by using the second SPECT image and the first high-quality SPECT image, which is high-qualityd by increasing the resolution of the second SPECT image.

6. The radiographic apparatus as claimed in claim 5, wherein, The image processing unit generates multiple second SPECT images based on the second single gamma rays detected at different times. When upscaling the first of the plurality of second SPECT images, the resolution of the second SPECT image is improved using the first upscaled SPECT image. When upgrading the second and subsequent second SPECT images among the plurality of second SPECT images, the resolution of the second SPECT image is improved by using the second upgraded SPECT image generated from the previous second SPECT image.

7. The radiographic apparatus as claimed in claim 3, wherein, The image processing unit is configured to use the SPECT image and the PET image to generate a high-quality PET image that has been improved by reducing the noise in the PET image.

8. The radiographic apparatus as claimed in claim 3, wherein, The learning model is a non-learning deep learning model. The image processing unit is configured to use the learning model, which is a non-learning deep learning model, to perform iterative calculations starting from a random image to make it approximate the input first gamma-ray image or the second gamma-ray image, thereby improving the resolution of the first gamma-ray image to enhance its quality, or improving the noise of the second gamma-ray image to enhance its quality.

9. The radiographic apparatus as claimed in claim 3, wherein, The learning model is a learning-oriented deep learning model. The image processing unit is configured to use the learning model that learns to improve the resolution of the first gamma-ray image or reduce the noise of the second gamma-ray image, thereby improving the resolution of the first gamma-ray image to a higher quality, or improving the noise of the second gamma-ray image to a higher quality.

10. The radiographic apparatus of claim 1, wherein, The image processing unit is configured to perform absorption correction on at least one of the first gamma-ray image and the second gamma-ray image based on the distribution of a pre-generated linear attenuation coefficient, and to use the absorption-corrected image for high-quality enhancement.

11. A method for processing radiation images, comprising: The step of generating a first gamma-ray image based on a single gamma ray detected by the first detector; The steps of generating a second gamma-ray image based on the annihilated gamma rays detected by multiple second detectors; and The step of using the first gamma-ray image and the second gamma-ray image to enhance either the first gamma-ray image or the second gamma-ray image to a higher quality.

Images