Gating-based pet image attenuation correction method, system, device and storage medium

Abstract

The application discloses a kind of based on the PET image attenuation correction method, system, device and storage medium of gate, is related to medical imaging technical field.The first PET image data pair of the present application is obtained by acquiring multiple, PET image data pair includes the first PET image based on CT attenuation correction and the second PET image without attenuation correction, the first PET image and the second PET image in first PET image data pair are registered respectively based on electrocardiogram gating technology to obtain second PET image data pair, attenuation correction model is trained based on the training data set formed by multiple second PET image data pair, the attenuation correction model of training is applied to the correction processing of the correction of the PET image to be corrected gate, and the attenuation-corrected gated PET image is obtained.The application constructs attenuation correction model using deep learning algorithm, in attenuation correction process, does not depend on CT image or other scanning image, reduces the influence of artifact due to motion, improves the attenuation correction effect of PET image.

Description

Technical Field

[0001] This invention relates to the field of medical imaging technology, and in particular to a gated PET image attenuation correction method, system, device, and storage medium. Background Technology

[0002] The essence of PET imaging technology is the capture and imaging of two gamma photons with the same energy of 511 keV but opposite directions, generated from the annihilation radiation of β+ particles emitted during the β decay of a radioactive tracer injected into the subject. These gamma photons are produced by the annihilation radiation of the β+ particles colliding with nearby free electrons within a short period. The accuracy of quantitative or semi-quantitative parameters in gated cardiac PET images, such as myocardial blood flow (MBF), myocardial flow reserve (MFR), and standard uptake value (SUV), is affected by many physical factors, including photon attenuation, scattering, random events, and delay. Photon attenuation has a significant impact on the accuracy of quantitative or semi-quantitative parameters in PET. As gamma photons often need to pass through different human tissues to reach the PET detector, the different tissue densities and other physical properties of these tissues result in different energy attenuation effects on the gamma photons. Therefore, during PET image reconstruction, it is often necessary to perform attenuation correction on the PET source data based on the subject's tissue structure information to obtain a more accurate PET activity distribution image and improve the accuracy of quantitative or semi-quantitative parameters in PET.

[0003] With the development and widespread adoption of PET / CT, CT-based attenuation correction methods have been widely applied to attenuation correction of gated cardiac PET images, demonstrating advantages such as low noise levels and high resolution in short scan times. However, CT-based attenuation correction methods for gated cardiac PET images are limited by CT artifact propagation and potential mismatches between CT and PET data. For gated cardiac PET images, gating technology only reduces the impact of cardiac motion on PET reconstruction; it does not address the mismatch between PET and CT images caused by respiratory motion and sudden patient movement, which leads to artifacts. Furthermore, most CT images are ungated, making precise matching with gated PET images difficult, exacerbating artifact propagation and resulting in poor PET image quality and inaccurate image estimation. Summary of the Invention

[0004] This invention aims to at least solve one of the technical problems existing in the prior art. To this end, this invention proposes a gated PET image attenuation correction method, system, device, and storage medium, which can directly perform attenuation correction on PET images and improve the PET image processing effect.

[0005] On one hand, embodiments of the present invention provide a gated PET image attenuation correction method, comprising the following steps:

[0006] Acquire the gated PET image to be corrected;

[0007] The gated PET image to be corrected is input into a neural network-based attenuation correction model to obtain an attenuation-corrected gated PET image;

[0008] The attenuation correction model is obtained through the following steps:

[0009] Acquire multiple first PET image data pairs, wherein the first PET image data pair includes a first PET image based on CT attenuation correction and a second PET image without attenuation correction;

[0010] Based on ECG gating technology, the first PET image and the second PET image in the first PET image data pair are registered to obtain the second PET image data pair, and a training dataset is formed based on multiple second PET image data pairs;

[0011] The training dataset is input into the initialized attenuation correction model for training, resulting in a trained attenuation correction model.

[0012] According to some embodiments of the present invention, the method of registering the first PET image and the second PET image in the PET image data pair based on ECG gating technology to obtain the second PET image data pair includes the following steps:

[0013] Register the first PET image in the first PET image data pair to the gated image at the end of the corresponding ECG cycle to obtain the first gated PET image;

[0014] The second PET image in the first PET image data pair is registered to the gated image of the end-diastolic phase of the corresponding electrocardiogram cycle to obtain the second gated PET image;

[0015] A second PET image data pair is formed based on the first gated PET image and the second gated PET image.

[0016] According to some embodiments of the present invention, the step of inputting the training dataset into the initialized attenuation correction model for training to obtain the trained attenuation correction model includes the following steps:

[0017] The first gated PET image in the training dataset is input into the initialized attenuation correction model for forward propagation to obtain the predicted image.

[0018] The model loss value is obtained based on the predicted image and the corresponding second gated PET image in the training dataset;

[0019] The parameters of the attenuation correction model are updated based on the model loss value;

[0020] By continuously updating the parameters of the attenuation correction model until the predicted image output by the attenuation correction model reaches a preset accuracy, or the number of parameter iterations of the attenuation correction model reaches a preset number, a trained attenuation correction model is obtained.

[0021] According to some embodiments of the present invention, the attenuation correction model includes a generator and a discriminator;

[0022] The generator includes an encoding layer and a decoding layer connected in sequence. The encoding layer is used to downsample the input image, and the decoding layer is used to upsample the output of the encoding layer to obtain the predicted image.

[0023] The discriminator is used to calculate the accuracy of the predicted image. When the accuracy is greater than a preset value, the predicted image is output.

[0024] According to some embodiments of the present invention, the encoder of the generator includes multiple convolutional layers, the decoder of the generator includes multiple deconvolutional layers, the encoder and the decoder are connected by residual blocks, and the convolutional layers of the encoder are skipped to the deconvolutional layers of the decoder corresponding to the image dimensions.

[0025] According to some embodiments of the present invention, the model loss value includes a generator loss value and a discriminator loss value, wherein the generator loss value is calculated using the following formula:

[0026] L G (x,y)=L adv (x)+λL1(G(x),y);

[0027] in, L1(x,y)=||yG(x)||1; x is the second-gated PET image, y is the predicted image, T real For the label of the second-gated PET image, T fake λ is the weight of the loss term, used to predict the label of the image.

[0028] According to some embodiments of the present invention, the discriminator loss value is calculated using the following formula:

[0029]

[0030] Where x is the second-gated PET image, y is the predicted image, and T real For the label of the second-gated PET image, Tfake To predict the label of the image.

[0031] On the other hand, embodiments of the present invention also provide a gated PET image attenuation correction system, comprising:

[0032] The first module is used to acquire the gated PET image to be calibrated;

[0033] The second module is used to input the gated PET image to be corrected into a neural network-based attenuation correction model to obtain an attenuation-corrected gated PET image.

[0034] The attenuation correction model is obtained through the following steps:

[0035] Acquire multiple first PET image data pairs, wherein the first PET image data pair includes a first PET image based on CT attenuation correction and a second PET image without attenuation correction;

[0036] Based on ECG gating technology, the first PET image and the second PET image in the first PET image data pair are registered to obtain the second PET image data pair, and a training dataset is formed based on multiple second PET image data pairs;

[0037] The training dataset is input into the initialized attenuation correction model for training, resulting in a trained attenuation correction model.

[0038] On the other hand, embodiments of the present invention also provide a gated PET image attenuation correction device, comprising:

[0039] At least one processor;

[0040] At least one memory for storing at least one program;

[0041] When the at least one program is executed by the at least one processor, the at least one processor implements the gated PET image attenuation correction method as described above.

[0042] On the other hand, embodiments of the present invention also provide a computer-readable storage medium storing computer-executable instructions for causing a computer to perform the gate-based PET image attenuation correction method as described above.

[0043] The technical solution described above has at least one of the following advantages or beneficial effects: By acquiring multiple first PET image data pairs, each pair including a first PET image based on CT attenuation correction and a second PET image without attenuation correction, and registering the first and second PET images in the first PET image data pairs using ECG gating technology to obtain second PET image data pairs, a training dataset is formed based on multiple second PET image data pairs. An attenuation correction model is trained using this training dataset, and the trained attenuation correction model is applied to the correction processing of the gated PET images to be corrected, resulting in attenuation-corrected gated PET images. This invention utilizes deep learning algorithms to construct the attenuation correction model. During the PET image attenuation correction process, it does not rely on CT images or other scan images, reducing the influence of motion artifacts, improving the attenuation correction effect of PET images, and obtaining accurate PET images. Attached Figure Description

[0044] Figure 1 This is a flowchart of the attenuation correction model acquisition method provided in an embodiment of the present invention;

[0045] Figure 2 This is a schematic diagram showing the comparison before and after registration provided in an embodiment of the present invention;

[0046] Figure 3 This is a schematic diagram of the attenuation correction model structure provided in an embodiment of the present invention;

[0047] Figure 4 This is a schematic diagram comparing images without attenuation correction and with attenuation correction using different methods, provided in an embodiment of the present invention.

[0048] Figure 5 This is a schematic diagram of a gated PET image attenuation correction device provided in an embodiment of the present invention. Detailed Implementation

[0049] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar originals or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0050] In the description of this invention, it should be understood that the orientation descriptions, such as up, down, left, right, etc., are based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0051] In the description of this invention, the use of terms such as "first," "second," etc., is merely for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of technical features indicated, or implicitly indicating the order of the technical features indicated.

[0052] To clearly illustrate the examples of this invention, the terms used in the embodiments of this invention are explained as follows:

[0053] CT (Computed Tomography) is a type of imaging technology that uses precisely collimated X-ray beams, gamma rays, ultrasound waves, etc., along with highly sensitive detectors to scan a specific part of the human body in a series of cross-sectional views. It features fast scanning time and clear images and can be used to examine a variety of diseases. Depending on the type of radiation used, it can be divided into X-ray CT (X-CT) and gamma-ray CT (gamma-CT), etc.

[0054] PET (positron emission tomography) is an imaging device that reflects the genetic, molecular, metabolic, and functional states of lesions. It uses positron-emitting radionuclides labeled with glucose and other human metabolites as imaging agents. The uptake of these agents by lesions reflects metabolic changes, thus providing clinical information on the biological metabolism of diseases. It represents a new milestone in the development of life sciences and medical imaging technology.

[0055] This invention provides a gated PET image attenuation correction method. The training process of the gated PET image attenuation correction method or attenuation correction model in this application can be applied to a terminal, a server, or software running on a terminal or server. The terminal can be a tablet computer, a laptop computer, a desktop computer, etc., but is not limited to these. The server can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0056] The gated PET image attenuation correction method of this invention includes, but is not limited to, steps S110 and S120.

[0057] Step S110: Obtain the gated PET image to be corrected;

[0058] Step S120: Input the gated PET image to be corrected into the attenuation correction model based on the neural network to obtain the attenuation-corrected gated PET image.

[0059] Among them, reference Figure 1 The attenuation correction model in step S120 is obtained through the following steps:

[0060] Step S121: Acquire multiple first PET image data pairs, wherein the first PET image data pair includes a first PET image based on CT attenuation correction and a second PET image without attenuation correction;

[0061] Step S122: Based on ECG gating technology, the first PET image and the second PET image in the first PET image data pair are registered to obtain the second PET image data pair, and a training dataset is formed based on multiple second PET image data pairs.

[0062] Step S123: Input the training dataset into the initialized attenuation correction model for training to obtain the trained attenuation correction model.

[0063] In some embodiments of step S110, ECG gating is a technical means to reduce or eliminate the influence of the pulsation of the heart and major blood vessels on the image. The PET image to be corrected refers to the PET image acquired based on different gating points of the ECG cycle and spatially registered with the acquired image data. Generally, 16 gating points are set for one ECG cycle, meaning 16 image data are acquired in one ECG cycle.

[0064] In some embodiments of step S120, during the PET image attenuation correction process, the gated PET image to be corrected is input into an attenuation correction model constructed based on a deep learning algorithm. This eliminates the need to rely on CT images or other scan images, reducing artifacts caused by motion, improving the attenuation correction effect of the PET image, and obtaining an accurate PET image. Furthermore, embodiments of the present invention can also reduce radiation exposure to the subject and increase the speed of attenuation correction.

[0065] In some embodiments of steps S121 to S123, multiple first PET image data pairs are acquired. Each PET image data pair includes a first PET image based on CT attenuation correction and a second PET image without attenuation correction. The first PET image and the second PET image in the first PET image data pair are registered using ECG gating technology to obtain a second PET image data pair. A training dataset is formed based on multiple second PET image data pairs. The registered first PET image in the training dataset is input into the attenuation correction model for processing to obtain a predicted attenuation correction image. The parameters of the attenuation correction model are updated in reverse according to the degree of difference between the predicted attenuation correction image and the second PET image in the training dataset. By continuously updating the model parameters, a trained attenuation correction model is obtained.

[0066] For example, Figure 4 The images shown are PET images with non-attenuation correction (NAC), PET images with CT-based attenuation correction (CTAC), and PET images with deep learning-based attenuation correction (DLAC), which are PET images output by the attenuation correction model. The PET image display results include the myocardial horizontal long axis (HLA), vertical long axis (VLA), and short axis (SA) images. The corresponding error map below is the error map between HLA and CTAC.

[0067] It should be noted that image registration is the process of mapping one image to another from two images in a dataset by finding a spatial transformation, so that points corresponding to the same spatial location in the two images correspond one-to-one, thereby achieving the purpose of information fusion.

[0068] According to some embodiments of the present invention, in step S122, the step of registering the first PET image and the second PET image in the PET image data pair based on ECG gating technology to obtain the second PET image data pair includes, but is not limited to, the following steps:

[0069] Step S210: Register the first PET image in the first PET image data pair to the gated image of the end-diastolic phase of the corresponding ECG cycle to obtain the first gated PET image;

[0070] Step S220: Register the second PET image in the first PET image data pair to the gated image of the end-diastolic phase of the corresponding ECG cycle to obtain the second gated PET image;

[0071] Step S230: Form a second PET image data pair based on the first gated PET image and the second gated PET image.

[0072] In this embodiment, after acquiring a set of PET images for a set of ECG cycles, the PET images can be converted into SUV PET images to reduce the dynamic range of image intensity, facilitating network training. Then, the cardiac portion of the images is cropped to preserve the left ventricular region. Attenuation correction based on CT images is applied to the cropped PET images to obtain the first set of gated PET images for a set of ECG cycles, while the uncorrected set of ECG cycle images serves as the second set of gated PET images. Since the intra-gated cardiac motion is minimal in the end-diastolic gated images (e.g., the 7th or 8th gate) of the ECG cycle, the 7th / 8th gated PET image for each subject is used as a reference. The remaining 15 gated cardiac PET images from the subject's ECG cycle are registered to the 7th / 8th gated PET image using the medical image registration package ANTsPy (Advanced Normalization Tools in Python), further reducing the impact of cardiac motion on network training. The registration mode of this application embodiment adopts affine registration + deformable registration technology. Mutual Information (MI) is used as the optimization criterion to obtain the deformation field, and then the deformation field is applied to the image to be registered to obtain the registered image. For example... Figure 2 As shown, the first row is the unregistered image, the second row is the registered image, and the third row is the error diagram between the two sets of images. The white dashed line is the reference line. Figure 2 Only a portion of the gating images are shown.

[0073] According to some embodiments of the present invention, step S123, which involves inputting the training dataset into the initialized attenuation correction model for training to obtain the trained attenuation correction model, includes, but is not limited to, the following steps:

[0074] Step S310: Input the first gated PET image in the training dataset into the initialized attenuation correction model for forward propagation to obtain the predicted image;

[0075] Step S320: Obtain the model loss value based on the predicted image and the corresponding second gated PET image in the training dataset;

[0076] Step S330: Update the parameters of the attenuation correction model based on the model loss value;

[0077] Step S340: By continuously updating the parameters of the attenuation correction model until the predicted image output by the attenuation correction model reaches a preset accuracy, or the parameter iteration number of the attenuation correction model reaches a preset number, a trained attenuation correction model is obtained.

[0078] In this embodiment, the attenuation correction model consists of a generator and a discriminator. For example, the generator G is a 3DRes-Unet (3D Unet and nine ResNet blocks). The discriminator D is a CNN architecture consisting of four 3D 3×3×3 convolutional layers, one fully connected layer, and one sigmoid layer. During training, an adaptive learning rate of 0.0001 can be used to apply the Adam optimizer to both the generator G and the discriminator D.

[0079] The model loss value can include the generator loss value and the discriminator loss value. The generator loss value is calculated using the following formula:

[0080] L G (x,y)=L adv (x)+λL1(G(x),y);

[0081] Among them, L adv It is the generator's adversarial loss. L1(x,y)=||yG(x)||1; x is the second-gated PET image, y is the predicted image, T real For the label of the second-gated PET image, T real =1,T fake To predict the label of an image, T fake =0, where λ is the weight of the loss term.

[0082] The discriminator loss value is calculated using the following formula:

[0083]

[0084] Where x is the second-gated PET image, y is the predicted image, and T real For the label of the second-gated PET image, T real =1,T fake To predict the label of an image, T fake =0.

[0085] According to some embodiments of the present invention, in combination Figure 3 The generator consists of an encoding layer and a decoding layer connected in sequence. The encoding layer downsamples the input image, and the decoding layer upsamples the output of the encoding layer to obtain the predicted image. The discriminator calculates the accuracy of the predicted image; if the accuracy is greater than a preset value, the predicted image is output.

[0086] The generator's encoder includes multiple convolutional layers, and the generator's decoder includes multiple deconvolutional layers. The encoder and decoder are connected via residual blocks, and the convolutional layers of the encoder are skipped to the corresponding deconvolutional layers in the decoder.

[0087] Specifically, in combination Figure 3 In this embodiment of the invention, the generator consists of two 3D U-Nets and nine ResNet blocks. The two 3D U-Nets form the encoding and decoding layers. The encoder and decoder consist of a series of convolutional layers with 3×3×3 kernels, instance normalization (IN) layers, and rectified linear units (ReLUs). Convolutional layers with strides of 2 and 3×3×3 kernels are used for downsampling. In each downsampling step, the number of feature channels is doubled, and bilinear interpolation is used for each upsampling step where the number of feature channels is halved. Corresponding layers in the encoder and decoder use skip connections. After two downsampling steps, depth features are extracted using nine residual blocks at a dropout rate of 0.5. The discriminator is a CNN architecture consisting of four 3D 3×3×3 convolutional layers, one fully connected layer, and one sigmoid layer. The discriminator's first convolutional layer consists of 64 convolutions with a stride of 2 and a kernel size of 3×3×3, connected to Leaky Rectified Linear Units (LReLUs). The second through fourth convolutional layers are connected to a batch normalization (BN) layer and the LReLU function. The slope of the LReLU function is 0.2. The number of kernels in each subsequent convolutional layer is twice that of the preceding layer.

[0088] This invention also provides a gated PET image attenuation correction system, comprising:

[0089] The first module is used to acquire the gated PET image to be calibrated;

[0090] The second module is used to input the gated PET image to be corrected into the attenuation correction model based on the neural network to obtain the attenuation-corrected gated PET image;

[0091] The attenuation correction model is obtained through the following steps:

[0092] Acquire multiple first PET image data pairs, wherein the first PET image data pair includes a first PET image based on CT attenuation correction and a second PET image without attenuation correction;

[0093] Based on ECG gating technology, the first PET image and the second PET image in the first PET image data pair are registered to obtain the second PET image data pair, and a training dataset is formed based on multiple second PET image data pairs;

[0094] The training dataset is input into the initialized attenuation correction model for training, resulting in a trained attenuation correction model.

[0095] It is understood that the content of the above-described gating-based PET image attenuation correction method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above-described gating-based PET image attenuation correction method embodiments, and the beneficial effects achieved are also the same as those achieved in the above-described gating-based PET image attenuation correction method embodiments.

[0096] Reference Figure 5 , Figure 5 This is a schematic diagram of a gated PET image attenuation correction device according to an embodiment of the present invention. The gated PET image attenuation correction device of this embodiment includes one or more control processors and a memory. Figure 5 The example consists of a control processor and a memory.

[0097] The control processor and memory can be connected via a bus or other means. Figure 5 Taking the example of a connection between China and Israel via a bus.

[0098] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the control processor, and these remote memories can be connected to the gated PET image attenuation correction device via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0099] Those skilled in the art will understand that Figure 5 The device structure shown does not constitute a limitation on gating-based PET image attenuation correction devices, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0100] The non-transient software program and instructions required to implement the gated PET image attenuation correction method applied to the gated PET image attenuation correction device in the above embodiments are stored in the memory. When executed by the controlled processor, the gated PET image attenuation correction method applied to the gated PET image attenuation correction device in the above embodiments is executed.

[0101] Furthermore, one embodiment of the present invention provides a computer-readable storage medium storing computer-executable instructions that are executed by one or more control processors, causing the one or more control processors to perform the gate-based PET image attenuation correction method in the above method embodiment.

[0102] It will be understood by those skilled in the art that all or some of the steps and systems in the methods disclosed above can be implemented as software, firmware, hardware, and suitable combinations thereof. Some or all of the physical components can be implemented as software executed by a processor, such as a central processing unit, digital signal processor, or microprocessor, or as hardware, or as an integrated circuit, such as an application-specific integrated circuit. Such software can be distributed on a computer-readable medium, which can include computer storage media (or non-transitory media) and communication media (or transient media). As is known to those skilled in the art, the term computer storage media includes volatile and non-volatile, removable and non-removable media implemented in any method or technology for storing information (such as computer-readable instructions, data structures, program modules, or other data). Computer storage media includes, but is not limited to, RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storage, magnetic cartridges, magnetic tape, disk storage or other magnetic storage devices, or any other medium that can be used to store desired information and is accessible to a computer. Furthermore, as is known to those skilled in the art, communication media typically contain computer-readable instructions, data structures, program modules, or other data in modulated data signals such as carrier waves or other transmission mechanisms, and may include any information delivery medium.

[0103] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention.

Claims

1. A gating-based PET image attenuation correction method, characterized in that, Includes the following steps: Acquire the gated PET image to be corrected; The gated PET image to be corrected is input into a neural network-based attenuation correction model to obtain an attenuation-corrected gated PET image; The attenuation correction model is obtained through the following steps: Acquire multiple first PET image data pairs, wherein the first PET image data pair includes a first PET image based on CT attenuation correction and a second PET image without attenuation correction; Based on ECG gating technology, the first PET image and the second PET image in the first PET image data pair are registered to obtain the second PET image data pair, and a training dataset is formed based on multiple second PET image data pairs; The training dataset is input into the initialized attenuation correction model for training, resulting in a trained attenuation correction model.

2. The gating-based PET image attenuation correction method of claim 1, wherein, The method of registering the first PET image and the second PET image in the PET image data pair based on ECG gating technology to obtain the second PET image data pair includes the following steps: Register the first PET image in the first PET image data pair to the gated image at the end of the corresponding ECG cycle to obtain the first gated PET image; The second PET image in the first PET image data pair is registered to the gated image of the end-diastolic phase of the corresponding electrocardiogram cycle to obtain the second gated PET image; A second PET image data pair is formed based on the first gated PET image and the second gated PET image.

3. The gating-based PET image attenuation correction method of claim 2, wherein, The step of inputting the training dataset into the initialized attenuation correction model for training to obtain the trained attenuation correction model includes the following steps: The first gated PET image in the training dataset is input into the initialized attenuation correction model for forward propagation to obtain the predicted image. The model loss value is obtained based on the predicted image and the corresponding second gated PET image in the training dataset; The parameters of the attenuation correction model are updated based on the model loss value; By continuously updating the parameters of the attenuation correction model until the predicted image output by the attenuation correction model reaches a preset accuracy, or the number of parameter iterations of the attenuation correction model reaches a preset number, a trained attenuation correction model is obtained.

4. The gating-based PET image attenuation correction method of claim 3, wherein, The attenuation correction model includes a generator and a discriminator; The generator includes an encoding layer and a decoding layer connected in sequence. The encoding layer is used to downsample the input image, and the decoding layer is used to upsample the output of the encoding layer to obtain the predicted image. The discriminator is used to calculate the accuracy of the predicted image. When the accuracy is greater than a preset value, the predicted image is output.

5. The gating-based PET image attenuation correction method of claim 4, wherein, The encoder of the generator includes multiple convolutional layers, and the decoder of the generator includes multiple deconvolutional layers. The encoder and the decoder are connected by residual blocks, and the convolutional layers of the encoder are skipped to the corresponding deconvolutional layers in the decoder.

6. The gating-based PET image attenuation correction method of claim 4, wherein, The model loss value includes the generator loss value and the discriminator loss value, and the generator loss value is calculated using the following formula: L G (x,y) = L adv (x) + λL1(G(x),y); wherein, L1(x, y) = ||y - G(x)||1; x is the second gated PET image, y is the predicted image, T real is the label of the second gated PET image, T fake is the label of the predicted image, and λ is the weight of the loss term.

7. The gating-based PET image attenuation correction method of claim 6, wherein, The discriminant loss value is calculated using the following formula: where x is the second gated PET image, y is the predicted image, T real is the label for the second gated PET image, T fake is the label for the predicted image.

8. A gating-based PET image attenuation correction system, characterized by, include: The first module is used to acquire the gated PET image to be calibrated; The second module is used to input the gated PET image to be corrected into a neural network-based attenuation correction model to obtain an attenuation-corrected gated PET image. The attenuation correction model is obtained through the following steps: Acquire multiple first PET image data pairs, wherein the first PET image data pair includes a first PET image based on CT attenuation correction and a second PET image without attenuation correction; Based on ECG gating technology, the first PET image and the second PET image in the first PET image data pair are registered to obtain the second PET image data pair, and a training dataset is formed based on multiple second PET image data pairs; The training dataset is input into the initialized attenuation correction model for training, resulting in a trained attenuation correction model.

9. A gating-based PET image attenuation correction apparatus characterized by comprising: include: At least one processor; At least one memory for storing at least one program; When the at least one program is executed by the at least one processor, the at least one processor implements the gated PET image attenuation correction method as described in any one of claims 1 to 7.

10. A computer readable storage medium having stored therein a program that is executable by a processor, characterized in that, When the processor executes the program, it is used to implement the gated PET image attenuation correction method as described in any one of claims 1 to 7.

Images