Scatter correction method, pet imaging method, device, apparatus, and storage medium

Abstract

The application discloses a scattering correction method, a PET imaging method, a device, equipment and a storage medium. The scattering correction method comprises the following steps: obtaining TOF-PET data and an attenuation image of a target object during scanning; obtaining a back projection image of TOF information based on the TOF-PET data; obtaining a scattering estimation of TOF-PET imaging based on the back projection image and the attenuation image; and performing scattering correction on the TOF-PET data based on the scattering estimation of TOF-PET imaging. The application solves the technical problem of long time consumption of scattering correction in TOF-PET imaging in the prior art.

Description

Technical Field

[0001] This invention relates to the field of medical technology, specifically to a scattering correction method, a PET imaging method, an apparatus, a device, and a storage medium. Background Technology

[0002] Positron emission tomography (PET) is a non-invasive medical imaging technique that detects metabolic characteristics of human or animal organs. It is characterized by high sensitivity, good accuracy, and precise positioning.

[0003] Time-of-flight positron emission tomography (TOF-PET) scanners are advanced functional imaging tools in nuclear medicine, and their application prospects have attracted great attention from nuclear medicine imaging researchers and equipment manufacturers. In the latest TOF-PET scanners, while acquiring coincidence counts of all tilted response lines improves system sensitivity compared to traditional 2D and extended 2D data acquisition modes, it also introduces a large number of scattering coincidence counts, thereby reducing the system's resolution and quantitative accuracy of the imaging. Therefore, scattering correction is necessary to improve the accuracy of TOF-PET imaging.

[0004] In traditional scattering reconstruction, the reconstructed image without scattering correction is typically used as the tracer distribution image. Combined with the attenuation image, a preliminary scattering projection domain distribution is estimated, and this is used as a preliminary scattering estimate to reconstruct a new tracer distribution image. This process is repeated several times until the scattering distribution converges. The iterative process requires repeated image reconstruction and scattering estimation, which is therefore very time-consuming. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned technical deficiencies and provide a scattering correction method, PET imaging method, device, equipment and storage medium to solve the technical problem of long scattering correction time in the prior art during TOF-PET imaging.

[0006] To achieve the above-mentioned technical objectives, the present invention adopts the following technical solution:

[0007] In a first aspect, the present invention provides a scattering correction method, comprising the following steps:

[0008] Acquire TOF-PET data and attenuation images of the target object during the scan;

[0009] A back-projection image of TOF information is obtained based on the TOF-PET data;

[0010] Based on the back-projection image and the attenuation image, a scattering estimate for TOF-PET imaging is obtained;

[0011] The TOF-PET data is scattered and corrected based on the scattering estimation from the TOF-PET imaging.

[0012] In some embodiments, obtaining the scattering estimate of the TOF-PET imaging based on the back-projected image and the attenuation image includes:

[0013] The attenuation image and the back-projection image are input into the fully trained first training model to obtain the tracer distribution image;

[0014] Based on the tracer distribution image and the attenuation image, the scattering estimate of the TOF-PET imaging is obtained.

[0015] In some embodiments, the scattering correction method further includes:

[0016] A first training set is acquired and a first initial training model is constructed. The first training set includes several first target parameters and several sets of first input parameters corresponding to the first target parameters. The first input parameters are back-projected images and their corresponding attenuation images. The first target parameters are tracer distribution images.

[0017] The first initial training model is trained using the first training set to obtain a fully trained first training model.

[0018] In some embodiments, the first training model is a machine learning model or a deep learning model.

[0019] In some embodiments, after acquiring the TOF-PET data and attenuation image of the target object during scanning, and before acquiring the back-projection image of TOF information based on the TOF-PET data, the method further includes:

[0020] The TOF-PET data is pre-corrected.

[0021] Secondly, the present invention also provides a PET imaging method, comprising the following steps:

[0022] Acquire TOF-PET data and static images of the target object during scanning, wherein the static images are acquired after scattering correction using the scattering correction method described above;

[0023] A back-projection image of TOF information is obtained based on the TOF-PET data;

[0024] The static image and the back-projected image are input into the fully trained second training model to obtain the PET reconstructed image.

[0025] In some embodiments, the PET imaging method further includes:

[0026] A second training set is obtained and a second initial training model is constructed. The second training set includes several second target parameters and several sets of second input parameters corresponding to the second target parameters. The second input parameters are static images and back-projected images, and the second target parameters are PET reconstructed images.

[0027] The second initial training model is trained using the second training set to obtain a fully trained second training model.

[0028] Thirdly, the present invention also provides a scattering correction device, comprising:

[0029] The initial data acquisition module is used to acquire TOF-PET data and attenuation images of the target object during the scanning process;

[0030] The back projection image acquisition module is used to acquire a back projection image of TOF information based on the TOF-PET data;

[0031] The scattering estimation acquisition module is used to acquire the scattering estimate of the TOF-PET imaging based on the back-projection image and the attenuation image;

[0032] A scattering correction module is used to perform scattering correction on the TOF-PET data based on the scattering estimation of the TOF-PET imaging.

[0033] Fourthly, the present invention also provides an electronic device, comprising: a processor and a memory;

[0034] The memory stores a computer-readable program that can be executed by the processor;

[0035] When the processor executes the computer-readable program, it implements the steps in the scattering correction method or PET imaging method as described above.

[0036] Fifthly, the present invention also provides a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the steps in the scattering correction method or PET imaging method as described above.

[0037] Compared with existing technologies, the scattering correction method, PET imaging method, apparatus, device, and storage medium provided by this invention first acquire TOF-PET data and attenuation images of the target object during scanning. Then, backprojection processing is performed on the TOF-PET data to obtain a backprojected image of the TOF information. Next, a scattering estimate is obtained based on this backprojected image and the attenuation image. Finally, scattering correction is performed on the TOF-PET data based on the scattering estimate. This effectively reduces the number of iterations used for scattering estimation, accelerating the scattering correction speed while ensuring the accuracy of the scattering estimate. During PET imaging, a combination of scatter-corrected static images and backprojected images is used for PET reconstruction imaging. The scatter-corrected static image provides a scattering correction kernel for the dynamic TOF-PET data. Therefore, the image reconstruction speed can be accelerated during PET imaging. Furthermore, by training a second training model, the static image and backprojected image can be directly input into the second training model during reconstruction to obtain the PET reconstructed image, further accelerating the image reconstruction speed. Attached Figure Description

[0038] Figure 1 This is a flowchart of an embodiment of the scattering correction method provided by the present invention;

[0039] Figure 2 This is a flowchart of an embodiment of the PET imaging method provided by the present invention;

[0040] Figure 3 This is a schematic diagram of an embodiment of the scattering correction device provided by the present invention;

[0041] Figure 4 This is a schematic diagram of an embodiment of the PET imaging device provided by the present invention;

[0042] Figure 5 This is a schematic diagram of the operating environment of an embodiment of the computer program of the present invention. Detailed Implementation

[0043] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0044] The scattering correction method, PET imaging method, apparatus, device, or computer-readable storage medium involved in this invention can be used in medical imaging systems such as positron emission tomography-X-ray computed tomography hybrid systems (PET-CT systems) and positron emission tomography-magnetic resonance imaging systems (PET-MRI systems). The methods, apparatus, devices, or computer-readable storage media involved in this invention can be integrated with the aforementioned systems or operate relatively independently.

[0045] This embodiment provides a scattering correction method that can be executed by medical imaging equipment such as PET-CT and PET-MRI devices, specifically by one or more processors of the device. Figure 1 This is a flowchart of the scattering correction method provided in the embodiments of the present invention. Please refer to [link / reference]. Figure 1 The scattering correction method includes the following steps:

[0046] S110. Acquire TOF-PET data and attenuation images of the target object during the scanning process;

[0047] S120. Obtain a back-projection image of TOF information based on TOF-PET data;

[0048] S130. Based on the back-projection image and the attenuation image, obtain the scattering estimate of the TOF-PET imaging;

[0049] S140. Scattering correction is performed on the TOF-PET data based on scattering estimation of TOF-PET imaging.

[0050] In this embodiment, the TOF-PET data and attenuation image of the target object during the scanning period are first acquired. Then, the TOF-PET data is back-projected to obtain the back-projected image of the TOF information. Next, the scattering estimate is obtained based on the back-projected image and the attenuation image. Finally, the TOF-PET data is scattered corrected based on the scattering estimate. This can effectively reduce the number of iterations used for scattering estimation, speed up the scattering correction while ensuring the accuracy of the scattering estimate.

[0051] In some embodiments, in step S110, the TOF-PET data may include coincidence event data and TOF information. In this embodiment, coincidence event data refers to the coincidence event data corresponding to the response line, which can be a matrix diagram encoded according to the angle and radial distance of the response line. Coincidence event data can also be called PET projection data. Coincidence event data includes true coincidence data, scattered coincidence event data, and random coincidence event data. Coincidence event data can be used to reconstruct the spatial distribution of contrast agent in the subject's body and acquire scan images. Preferably, the data format of the TOF-PET data can be a sinogram format or a list format. The list format can record the detected event information sequentially in the form of a data stream, and the data may include the photon incident crystal bar number, photon energy, and photon time-of-flight information. The sinogram format is a data storage format that combines and stores the number of coincidence events occurring on each response line. The above-mentioned TOF information can be detected by a PET detector and can be represented by a probability density function.

[0052] In some embodiments, TOF-PET data can be obtained by processing data detected by a PET scanner using computer equipment. The specific process is as follows: the operator injects a tracer labeled with a nuclide (such as F18, C11, O15, Ga68, or Ru82) into the subject's body in advance. The tracer diffuses into various tissues or blood vessels in the subject's body. Then, the PET detector detects the radiation signal generated by the annihilation effect between the positrons of the tracer and the negative electrons in the subject's body (this signal is a pair of gamma photons with equal energy and opposite direction). The detector converts the radiation signal into an electrical signal through photoelectric conversion. The detector then transmits the detected signal to the electronic circuit system. The electronic circuit system can convert the signal collected by the detector into a digital signal, which is the TOF-PET data.

[0053] In some embodiments, the attenuation image can originate from processes such as MR scanning or CT scanning. When it originates from a CT scanning process, the specific acquisition process is as follows: a CT scan is performed before the PET scan, CT scan images are acquired during the CT scan, and then tissue attenuation information is displayed based on the CT scan images, thus obtaining the attenuation image. This attenuation image can be used to assist PET images in scattering estimation. When it originates from an MR scanning process, the specific acquisition process is as follows: it is converted into an attenuation image through a UTE sequence.

[0054] In step S120, compared to traditional TOF-PET data which is stored in table mode or sine graph format, this embodiment converts the TOF-PET data into a back-projection image mode. The back-projection image can be applied to the ML-EM (Maximum Expectation Reconstruction) algorithm for reconstruction, referred to as DIRECT reconstruction. In some embodiments, the back-projection image can be a TOF histo-image or a TOF histo-projection. A TOF histo-image is obtained by determining the physical locations of multiple annihilation points based on the time difference between the arrival of two photons of the same annihilation event at both detectors in the scan data, and then accumulating these locations by mapping them to pixels in the image domain. A TOF histo-projection is generated by directly back-projecting events within a specific range of azimuth angles based on the azimuth angles of the coincident events. The back-projection images of all angles are superimposed to obtain the equivalent back-projection images of all events. The back-projection image of the TOF information can be obtained using the direct projection method or a TOF-based filtered back-projection method. The filtered back-projection method is a spatial processing technique based on Fourier transform theory. Its key feature is that it performs convolution processing on the projections at each acquisition angle before backprojection, thereby improving the shape artifacts caused by the point spread function and resulting in better reconstructed image quality. The specific process is as follows: first, the projection data obtained from the linear array detector undergoes a one-dimensional Fourier transform, and then convolves with a filter function to obtain convolutionally filtered projection data in each direction; then, these are backprojected along each direction; and after appropriate processing, a tomographic image of the scanned object is obtained. In this embodiment of the invention, the direct point projection method is preferred for acquiring the backprojected image of TOF information. This method is fast, has high throughput, and can achieve real-time imaging.

[0055] In some embodiments, to increase the accuracy of the back-projected image of the TOF information, the following step is further included between steps S110 and S120:

[0056] Pre-calibration processing was performed on the TOF-PET data.

[0057] In this embodiment, the purpose of the pre-correction process is to subtract scattering data, thereby increasing the accuracy and speed of scattering correction. In practical implementation, during the acquisition of the back-projected image of the TOF information, image-domain correction data can be selectively added, such as image-domain random correction and energy-based preliminary scattering correction data, to further improve the accuracy of the TOF information back-projected image. Specifically, image-domain random correction refers to projecting random data into the image domain, while energy-based preliminary scattering correction refers to projecting scattering projection data corresponding to different energy spectra into the image domain.

[0058] Step S130 is used for scattering estimation in TOF-PET imaging. In the same scan, when there is no significant motion influence, the attenuation image remains unchanged. Therefore, a scattering correction kernel can be extracted for the reconstruction of dynamic multi-frame images. In some embodiments, estimation methods such as Single Scatter Simulation (SSS), Monte Carlo Simulation (MCS), Double Scatter Simulation (DSS), or Residual Estimation Approach can be used to obtain the scattering estimate. This invention does not limit this method. The obtained scattering estimate can be used for the reconstruction of TOF images or non-TOF images. The generated scattering estimate can be in the form of sinogram data, list-mode data, or other relevant projection domain data formats.

[0059] In some embodiments, to improve the accuracy of scattering estimation and to perform scattering estimation quickly, step S130 includes:

[0060] The attenuation image and back-projection image are input into the fully trained first training model to obtain the tracer distribution image;

[0061] Scattering estimates from TOF-PET imaging are obtained based on tracer distribution and attenuation images.

[0062] In this embodiment, a training model is used to generate the tracer distribution image. The tracer distribution image obtained by the first training model is used for scattering correction, and since preliminary scattering correction has already been performed, the number of iterations for scattering estimation can be effectively reduced in subsequent scattering correction steps, ensuring the accuracy of scattering estimation. To ensure that the first training model can effectively obtain the tracer distribution image based on the backprojected image and attenuation image of the TOF information, the first training model needs to be trained first.

[0063] In some embodiments of the present invention, the specific process of establishing the first training model is as follows: firstly, an initial model is established through an algorithm, and then the initial model is trained based on historical tracer distribution images, historical back-projection images and their corresponding decay images. After the training is completed, the first training model can be obtained. Based on the first training model, the tracer distribution image can be output through the input back-projection image and its corresponding decay image.

[0064] Preferably, the scattering correction method further includes:

[0065] A first training set is acquired and a first initial training model is constructed. The first training set includes several first target parameters and several sets of first input parameters corresponding to the first target parameters. The first input parameters are back-projected images and their corresponding attenuation images. The first target parameters are tracer distribution images.

[0066] The first initial training model is trained using the first training set to obtain a fully trained first training model.

[0067] In this embodiment, each individual in the first training set used for training includes two first input parameters and one first target parameter. The two first input parameters are the back-projected image and its corresponding attenuation image. The back-projected image and the attenuation image can be obtained through steps S110 to S120. The first target parameter is the tracer distribution image, which can be obtained using the Ordered Subsets Expectation Maximization (OSEM) algorithm. OSEM is a method for PET image reconstruction. Its specific implementation process is as follows: all projection data are divided into multiple subsets. Each time the data of a subset is used, all pixels are updated once, which is called a sub-iteration. All subsets are used in turn to complete one iteration. Each iteration will cause the image to converge a little. After multiple iterations, a reconstructed image is obtained, which is closer to the true value. Therefore, in this embodiment of the invention, the back-projected image and its corresponding attenuation image are used as inputs to the first initial training model, and the tracer distribution image is used as the output of the first initial training model for model training. This is equivalent to the model automatically learning the scattering distribution of the projection domain and performing preliminary scattering estimation, thus enabling rapid scattering estimation. After training, the true tracer distribution image can be directly obtained from the back-projected image and its corresponding attenuation image, effectively reducing the number of iterations required for scattering estimation compared to existing technologies. Moreover, the attenuation image provides the distribution of substances that physically scatter, increasing the information content of the training data. Furthermore, since traditional scattering correction is based on attenuation maps and tracer distribution images, this embodiment of the invention adds the attenuation image as input, increasing the amount of information for the training network, facilitating network learning, and ultimately enhancing the robustness of the training network.

[0068] In some embodiments, the first training model is a machine learning model or a deep learning model. The first initial training model also corresponds to a machine learning model or a deep learning model. The machine learning model may include, but is not limited to, linear regression models, ridge regression models, support vector regression models, support vector machines, decision trees, fully connected neural networks, recurrent neural networks, etc. The deep learning model may include, but is not limited to, deep belief networks (DBNs), convolutional neural networks (CNNs), recurrent neural networks (RNNs), recurrent neural networks (RNNs), generative adversarial networks (GANs), etc. In specific embodiments of the present invention, the network structure of the first training model may be any one of 2DU-net, 3DU-net, U-net++, U-net3+, and V-net. Of course, in other embodiments, other models may be used to implement the functions of the present invention, and the present invention is not limited thereto.

[0069] It should be understood that a deep learning model can learn certain knowledge and capabilities from existing data (historical medical images) to process new data, and can be designed to perform various tasks, in the embodiments of the present invention, for generating tracer distribution images.

[0070] In some embodiments, the first training model can be a head training model, and the corresponding attenuation image is a head attenuation image. The first training model can also be a body training model, and the corresponding attenuation image is a body attenuation image. The parameters of the body training model and the head training model may be different or can be shared. The specific implementation method can be determined according to actual needs, and the embodiments of the present invention do not limit this.

[0071] After obtaining the tracer distribution image through the first training model, the scattering estimate of TOF-PET imaging can be obtained based on the attenuation image and the tracer distribution image. In the same scan, when there is no significant motion influence, the attenuation image remains unchanged. Therefore, a scattering correction kernel can be extracted for the reconstruction of dynamic multi-frame images. In some embodiments, estimation methods such as Single Scatter Simulation (SSS), Monte Carlo Simulation (MCS), or Double Scatter Simulation (DSS) can be used to obtain the scattering estimate. This invention does not limit this method, and the obtained scattering estimate can be used for the reconstruction of TOF images or non-TOF images. The generated scattering estimate can be in the form of sinogram data, list-mode data, or other relevant projection domain data formats.

[0072] Step S140 is to perform scattering correction on the TOF-PET data, specifically by correcting the TOF-PET data using scattering estimation data. Since the TOF-PET scattering estimation is close to the true value, it can effectively improve the accuracy of the reconstructed TOF-PET image.

[0073] In this embodiment of the invention, a backprojection image is obtained by direct backprojection of Time-of-Flight (TOF) information. This method is fast, has high throughput, and can achieve real-time imaging. A conversion model from the backprojection image to the actual tracer distribution image is constructed using a first training model, thereby quickly and accurately estimating the tracer distribution image used for scattering based on each generated backprojection image. This method effectively reduces the number of iterations used for scattering estimation and ensures the accuracy of the scattering estimation.

[0074] Based on the above scattering correction method, this invention also provides a PET imaging method. Please refer to [link to relevant documentation]. Figure 2 The PET imaging method includes the following steps:

[0075] S210. Acquire TOF-PET data and still images of the target object during the scanning process;

[0076] S220. Back-projection image of TOF information obtained based on TOF-PET data;

[0077] S230. Input the static image and the back-projected image into the fully trained second training model to obtain the PET reconstructed image.

[0078] In some embodiments, the static image is a static image obtained by the scattering correction method as described in the above embodiments. Since the scattering correction method has been described in detail above, it will not be repeated here.

[0079] In some embodiments, the static image can also be obtained by conventional methods. Specifically, the reconstructed image without scattering correction is used as the initial tracer distribution image. Combined with the attenuation image, the preliminary scattering projection domain distribution is estimated. This preliminary scattering estimate is then used to perform the next image reconstruction including the scattering estimate, and a new tracer distribution image is reconstructed. This process is repeated several times until the scattering distribution converges, and then the static image is reconstructed.

[0080] The PET reconstruction method provided in this embodiment uses a combination of static images and TOF-PET data for PET reconstruction imaging. The static images have already undergone scattering correction, so they can provide a scattering correction kernel for the TOF-PET backprojection image. Therefore, the image reconstruction speed can be accelerated during PET imaging. Moreover, by training the second training model, the static images and backprojection images can be directly input into the second training model during reconstruction to obtain the PET reconstruction image, which can further accelerate the image reconstruction speed.

[0081] In some embodiments, in step S210, the static image is acquired during the scan. A static image refers to an image reconstructed for routine clinical diagnosis, and the scan time is relatively long. For example, if the duration of a dynamic scan is 30 minutes, it can be considered as multiple static images sequentially linked together, with each static image scanned for 2 minutes or longer. We can select a static scan image of 2 minutes or longer as the model input. The static images generally refer to images that have undergone scattering correction and can be directly used for PET image reconstruction.

[0082] Step S220 involves converting the acquired raw TOF-PET data into a back-projection image. In some embodiments, to reduce scattering data in the raw data and increase the accuracy and speed of scattering correction, pre-correction processing is performed on the TOF-PET data before acquisition. This pre-correction process includes, but is not limited to, random correction, normalization correction, and preliminary scattering correction. Random correction refers to projecting random data into the image domain. Normalization correction utilizes a PET device to acquire data from a uniform source. After acquiring a sufficient amount of data, statistical analysis is performed to form the efficiency factor of the detector unit. Then, during TOF-PET data acquisition, the data acquired by the detector is multiplied by the efficiency factor of the detector unit to achieve normalization correction. Preliminary scattering correction refers to projecting the scattering projection data corresponding to different energy spectra into the image domain.

[0083] Step S230 is used to reconstruct PET images using a trained model, resulting in fast reconstruction speed. To ensure that the second training model can obtain more accurate reconstructed images, it needs to be trained first. In some embodiments of the present invention, the specific process of establishing the second training model is as follows: First, an initial model is established using an algorithm. Then, the initial model is trained based on traditional PET reconstructed images, historical backprojection images, and static images. After training, the second training model is obtained. Based on the second training model, the PET reconstructed image can be output using the input backprojection image and static image. The static image used for training can be obtained using traditional methods or the scattering correction method described in the above embodiments, and used as model input.

[0084] Preferably, the PET imaging method further includes:

[0085] Obtain a second training set and construct a second initial training model. The second training set includes several second target parameters and several sets of second input parameters corresponding to the second target parameters. The second input parameters are static images and back-projected images, and the second target parameters are PET reconstructed images.

[0086] The second initial training model is trained using the second training set to obtain a fully trained second training model.

[0087] In this embodiment, each individual in the second training set used for training includes two second input parameters and one second target parameter. The two second input parameters are the static reconstructed image and the back-projected image, which can be obtained through steps S210 to S220. The second target parameter is the PET reconstructed image, which can be obtained using traditional reconstruction methods. Therefore, in this embodiment of the invention, the static image and the back-projected image are used as inputs to the second initial training model, and the traditional PET reconstructed image is used as the output of the second initial training model for model training. This is equivalent to the model automatically learning the PET reconstruction process, and the PET reconstructed image can be obtained directly from the static image and the back-projected image, which speeds up the reconstruction process compared to existing technologies.

[0088] In some embodiments, the second training model is a machine learning model or a deep learning model. The second initial training model also corresponds to a machine learning model or a deep learning model. The machine learning model may include, but is not limited to, linear regression models, ridge regression models, support vector regression models, support vector machines, decision trees, fully connected neural networks, recurrent neural networks, etc. The deep learning model may include, but is not limited to, deep belief networks (DBNs), convolutional neural networks (CNNs), recurrent neural networks (RNNs), recurrent neural networks (RNNs), generative adversarial networks (GANs), etc. In a specific embodiment of the present invention, the network structure of the first training model may be any one of 2DU-net, 3DU-net, U-net++, U-net3+, and V-net. Of course, in other embodiments, other models may be used to implement the functions of the present invention, and the present invention is not limited thereto.

[0089] It should be understood that a deep learning model can learn certain knowledge and capabilities from existing data (historical medical images) to process new data, and can be designed to perform various tasks, in the embodiments of the present invention, for generating PET reconstructed images.

[0090] It should be noted that the PET imaging method provided in this embodiment of the invention can be used to reconstruct dynamic images. Dynamic images can be regarded as being formed by connecting multiple static images in a temporal sequence. Therefore, when reconstructing dynamic images, each frame of back-projected image and static image are sequentially input into the second training model, and the resulting multi-frame PET reconstructed images are the dynamic reconstructed images.

[0091] Based on the above scattering correction method, this embodiment of the invention also provides a scattering correction device 300. Please refer to [link to relevant documentation]. Figure 3 The scattering correction device 300 includes an initial data acquisition module 310, a first back-projection image acquisition module 320, a scattering estimation acquisition module 330, and a scattering correction module 340.

[0092] The initial data acquisition module 310 is used to acquire TOF-PET data and attenuation images of the target object during the scanning process.

[0093] The first back-projection image acquisition module 320 is used to acquire back-projection images based on TOF-PET data to obtain TOF information.

[0094] The scattering estimation acquisition module 330 is used to acquire scattering estimates from TOF-PET imaging based on the back-projection image and the attenuation image.

[0095] The scattering correction module 340 is used to perform scattering correction on TOF-PET data based on scattering estimation of TOF-PET imaging.

[0096] In this embodiment, the TOF-PET data and attenuation image of the target object during the scanning period are first acquired. Then, the TOF-PET data is back-projected to obtain the back-projected image of the TOF information. Next, the scattering estimate is obtained based on the back-projected image and the attenuation image. Finally, the TOF-PET data is scattered corrected based on the scattering estimate. This can effectively reduce the number of iterations used for scattering estimation, speed up the scattering correction while ensuring the accuracy of the scattering estimate.

[0097] In some embodiments, the scattering estimation acquisition module includes a tracer distribution image acquisition unit and a scattering estimation unit.

[0098] The tracer distribution image acquisition unit is used to input the attenuation image and the back projection image into the fully trained first training model to obtain the tracer distribution image;

[0099] The scattering estimation unit is used to obtain scattering estimates from TOF-PET imaging based on tracer distribution and attenuation images.

[0100] In some embodiments, the scattering correction device 300 further includes a first training module, the first training module being used for:

[0101] A first training set is acquired and a first initial training model is constructed. The first training set includes several first target parameters and several sets of first input parameters corresponding to the first target parameters. The first input parameters are back-projected images and their corresponding attenuation images. The first target parameters are tracer distribution images.

[0102] The first initial training model is trained using the first training set to obtain a fully trained first training model.

[0103] In some embodiments, the first training model is a machine learning model or a deep learning model.

[0104] In some embodiments, the scattering correction device 300 further includes a pre-correction module for pre-correcting TOF-PET data.

[0105] Based on the above-described scattering correction method, apparatus, and PET imaging method, this embodiment of the invention also provides a corresponding PET imaging apparatus 400. Please refer to [link to relevant documentation]. Figure 4The PET imaging device 400 includes a static image acquisition module 410, a second back-projection image acquisition module 420, and a reconstruction module 430.

[0106] The static image acquisition module 410 is used to acquire TOF-PET data and static images of the target object during scanning;

[0107] The second back-projection image acquisition module 420 is used to acquire back-projection images of TOF information based on TOF-PET data.

[0108] The reconstruction module 430 is used to input the static image and the back-projected image into the fully trained second training model to obtain the PET reconstructed image.

[0109] In this embodiment, the static image acquisition module 410 can acquire a static image either by performing scattering correction on the TOF-PET data as described in the above embodiments, or by using conventional methods. In this embodiment, the static image acquired after scattering correction and the back-projection image are combined for PET reconstruction imaging. The static image acquired after scattering correction provides a scattering correction kernel for the dynamic TOF-PET data. Therefore, the image reconstruction speed can be accelerated during PET imaging. Furthermore, by training the second training model, the static image and the back-projection image can be directly input into the second training model during reconstruction to obtain the PET reconstructed image, which can further accelerate the image reconstruction speed.

[0110] In some embodiments, the PET imaging device 400 further includes a second training module, the second training module being used for:

[0111] Obtain a second training set and construct a second initial training model. The second training set includes several second target parameters and several sets of second input parameters corresponding to the second target parameters. The second input parameters are static images and back-projected images, and the second target parameters are PET reconstructed images.

[0112] The second initial training model is trained using the second training set to obtain a fully trained second training model.

[0113] In some embodiments, the second training model is a machine learning model or a deep learning model.

[0114] like Figure 5 As shown, based on the aforementioned scattering correction method and PET imaging method, the present invention also provides an electronic device, which can be a control device for a medical imaging system, a mobile terminal, a desktop computer, a laptop, a handheld computer, or a server, etc. The electronic device includes a processor 10, a memory 20, and a display 30. Figure 5Only some components of the electronic device are shown; however, it should be understood that it is not required to implement all of the components shown, and more or fewer components may be implemented instead.

[0115] In some embodiments, memory 20 may be an internal storage unit of the electronic device, such as a hard disk or memory. In other embodiments, memory 20 may be an external storage device of the electronic device, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc. Furthermore, memory 20 may include both internal and external storage units. Memory 20 is used to store application software and various types of data installed on the electronic device, such as program code installed on the electronic device. Memory 20 may also be used to temporarily store data that has been output or will be output. In one embodiment, memory 20 stores a computer program 40, which can be executed by processor 10 to implement the scattering correction method or PET imaging method of the embodiments of this application.

[0116] In some embodiments, processor 10 may be a central processing unit (CPU), microprocessor or other data processing chip, used to run program code stored in memory 20 or process data, such as performing scattering correction methods or PET imaging methods.

[0117] In some embodiments, the display 30 may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen. The display 30 is used to display information on the electronic device and to display a visual user interface. The components of the electronic device, processor 10, memory 20, and display 30, communicate with each other via a system bus.

[0118] In one embodiment, when the processor 10 executes the computer program 40 in the memory 20, the following steps are performed:

[0119] Acquire TOF-PET data and attenuation images of the target object during the scan;

[0120] Back-projection images of TOF information obtained from TOF-PET data;

[0121] Scattering estimates from TOF-PET imaging are obtained based on back-projection and attenuation images;

[0122] Scattering correction is performed on TOF-PET data based on scattering estimation from TOF-PET imaging.

[0123] In one embodiment, when the processor 10 executes the computer program 40 in the memory 20, the following steps are also performed:

[0124] The attenuation image and back-projection image are input into the fully trained first training model to obtain the tracer distribution image;

[0125] Scattering estimates from TOF-PET imaging are obtained based on tracer distribution and attenuation images.

[0126] In one embodiment, when the processor 10 executes the computer program 40 in the memory 20, the following steps are also performed:

[0127] A first training set is acquired and a first initial training model is constructed. The first training set includes several first target parameters and several sets of first input parameters corresponding to the first target parameters. The first input parameters are back-projected images and their corresponding attenuation images. The first target parameters are tracer distribution images.

[0128] The first initial training model is trained using the first training set to obtain a fully trained first training model.

[0129] In one embodiment, the first training model is a machine learning model or a deep learning model.

[0130] In one embodiment, when the processor 10 executes the computer program 40 in the memory 20, the following steps are also performed:

[0131] Pre-calibration processing was performed on the TOF-PET data.

[0132] In one embodiment, when the processor 10 executes the computer program 40 in the memory 20, the following steps are also performed:

[0133] Acquire TOF-PET data and still images of the target object during the scanning process;

[0134] Back-projection images of TOF information obtained from TOF-PET data;

[0135] The static image and the back-projected image are input into the fully trained second training model to obtain the PET reconstructed image.

[0136] In one embodiment, when the processor 10 executes the computer program 40 in the memory 20, the following steps are also performed:

[0137] A second training set is obtained and a second initial training model is constructed. The second training set includes several second target parameters and several sets of second input parameters corresponding to the second target parameters. The second input parameters are static images and back-projected images, and the second target parameters are PET reconstructed images.

[0138] The second initial training model is trained using the second training set to obtain a fully trained second training model.

[0139] In one embodiment, the second training model is a machine learning model or a deep learning model.

[0140] In summary, the scattering correction method, PET imaging method, apparatus, device, and storage medium provided by this invention first acquire TOF-PET data and attenuation images of the target object during scanning. Then, the TOF-PET data is back-projected to obtain a back-projected image of the TOF information. Subsequently, a trained first training model is used to calculate the attenuation image and the back-projected image to obtain a tracer distribution image. This tracer distribution image is the true tracer distribution image after preliminary scattering correction. Then, a scattering estimate is obtained based on the tracer distribution image and the attenuation image. Finally, the TOF-PET data is scattered based on the scattering estimate, thereby effectively reducing the number of iterations used for scattering estimation, accelerating the scattering correction speed while ensuring the accuracy of the scattering estimate. During PET imaging, a combination of scatter-corrected static images and back-projection images is used for PET reconstruction imaging. The scatter-corrected static images provide a scatter correction kernel for the dynamic TOF-PET data. Therefore, the image reconstruction speed can be accelerated during PET imaging. Furthermore, by training a second training model, the static images and back-projection images can be directly input into the second training model during reconstruction to obtain the PET reconstructed image, which can further accelerate the image reconstruction speed.

[0141] Of course, those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware (such as a processor, controller, etc.). The program can be stored in a computer-readable storage medium, and when executed, it can include the processes described in the above method embodiments. The storage medium can be a memory, magnetic disk, optical disk, etc.

[0142] The specific embodiments of the present invention described above do not constitute a limitation on the scope of protection of the present invention. Any other corresponding changes and modifications made in accordance with the technical concept of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A scattering correction method, characterized in that, Includes the following steps: Acquire TOF-PET data and attenuation images of the target object during the scan; A back-projection image of TOF information is obtained based on the TOF-PET data; Based on the back-projection image and the attenuation image, a scattering estimate for TOF-PET imaging is obtained; Scattering correction is performed on the TOF-PET data based on the scattering estimation of the TOF-PET imaging; The step of obtaining the scattering estimate of TOF-PET imaging based on the back-projection image and the attenuation image includes: The attenuation image and the back-projection image are input into the fully trained first training model to obtain the tracer distribution image; Based on the tracer distribution image and the attenuation image, the scattering estimate of the TOF-PET imaging is obtained.

2. The scattering correction method according to claim 1, characterized in that, The method further includes: A first training set is acquired and a first initial training model is constructed. The first training set includes several first target parameters and several sets of first input parameters corresponding to the first target parameters. The first input parameters are back-projected images and their corresponding attenuation images. The first target parameters are tracer distribution images. The first initial training model is trained using the first training set to obtain a fully trained first training model.

3. The scattering correction method according to claim 2, characterized in that, The first training model is a machine learning model or a deep learning model.

4. The scattering correction method according to claim 1, characterized in that, After acquiring the TOF-PET data and attenuation image of the target object during the scan, and before acquiring the back-projection image of TOF information based on the TOF-PET data, the process further includes: The TOF-PET data is pre-corrected.

5. A PET imaging method, characterized in that, Includes the following steps: Acquire TOF-PET data and still images of the target object during scanning, wherein the still images are acquired after scattering correction by the scattering correction method as described in any one of claims 1-4; A back-projection image of TOF information is obtained based on the TOF-PET data; The static image and the back-projected image are input into the fully trained second training model to obtain the PET reconstructed image.

6. The PET imaging method according to claim 5, characterized in that, The method further includes: A second training set is obtained and a second initial training model is constructed. The second training set includes several second target parameters and several sets of second input parameters corresponding to the second target parameters. The second input parameters are static images and back-projected images, and the second target parameters are PET reconstructed images. The second initial training model is trained using the second training set to obtain a fully trained second training model.

7. A scattering correction device, characterized in that, include: The initial data acquisition module is used to acquire TOF-PET data and attenuation images of the target object during the scanning process; The back projection image acquisition module is used to acquire a back projection image of TOF information based on the TOF-PET data; The scattering estimation acquisition module is used to acquire the scattering estimate of the TOF-PET imaging based on the back-projection image and the attenuation image; A scattering correction module is used to perform scattering correction on the TOF-PET data based on the scattering estimation of the TOF-PET imaging; The scattering estimation module includes a tracer distribution image acquisition unit and a scattering estimation unit. The tracer distribution image acquisition unit is used to input the attenuation image and the back projection image into the fully trained first training model to obtain the tracer distribution image; The scattering estimation unit is used to obtain scattering estimates from TOF-PET imaging based on tracer distribution and attenuation images.

8. An electronic device, characterized in that, include: Processor and memory; The memory stores a computer-readable program that can be executed by the processor; When the processor executes the computer-readable program, it implements the steps of the scattering correction method as described in any one of claims 1-4 or the PET imaging method as described in any one of claims 5-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the steps of the scattering correction method as described in any one of claims 1-4 or the PET imaging method as described in any one of claims 5-6.

Images