A method, system, and related apparatus for image correction of SPECT images.

Abstract

This application discloses a method, system, and related apparatus for correcting SPECT images, relating to the field of image processing. The method includes: acquiring and aligning SPECT and CT images of the same patient; extracting functional distribution features of the SPECT image and structural features of the CT image using a pre-defined deep learning model to determine the artifact probability values ​​of pixels in the SPECT image; classifying the pixels of the SPECT image using an artifact probability threshold; generating a binary mask image of the SPECT image based on the classification results; determining the artifact regions of the SPECT image using the binary mask image; correcting the pixel values ​​of the artifact regions using structural information from the CT image; and fusing the corrected artifact regions with the non-artifact regions of the SPECT image to obtain the corrected SPECT image. This application utilizes the structural information of the CT image to correct SPECT images, improving the accuracy of SPECT image correction.

Description

Technical Field

[0001] This application relates to the field of image processing technology, and in particular to an image correction method, system and related apparatus for SPECT images. Background Technology

[0002] SPECT (Single Photon Emission Computed Tomography) is a nuclear medicine imaging technique. Its principle is as follows: a radiopharmaceutical is injected into the patient's body. The radiopharmaceutical releases single photons (gamma rays) in different tissues or organs. A detector receives the signals released by the radiopharmaceutical, obtaining raw projection data. A computer then reconstructs a tomographic image (SPECT image) of the tissue or organ based on this raw projection data. The reconstruction process involves converting the raw projection data acquired by the detector into a tomographic image using mathematical algorithms.

[0003] Because radiopharmaceuticals can accumulate in specific tissues or organs, and the distribution of radiopharmaceuticals is related to the blood, metabolism, nerve conduction, and other functions of the tissues or organs, SPECT images can include the functional state of tissues or organs (such as the blood distribution and metabolic state of tissues or organs).

[0004] Due to factors such as defects in scanning equipment and acquisition time, some projection data may be missing or incomplete, leading to artifacts in SPECT images that do not belong to human tissues or organs. These artifacts can affect the accuracy of diagnosis. Currently, artifact correction in SPECT images is mainly achieved through optimization algorithms. These algorithms deduce the missing projection data from the existing data and correct the artifacts by supplementing the SPECT image.

[0005] Since optimization algorithms often rely on prior assumptions to supplement SPECT images and correct them, if the prior assumptions do not conform to reality, the supplemented SPECT images will be distorted, resulting in low accuracy of SPECT image correction. Summary of the Invention

[0006] In view of the above problems, this application provides a method, system, and related apparatus for correcting SPECT images, so as to improve the accuracy of SPECT image correction. The specific solution is as follows:

[0007] The first aspect of this application provides a method for image correction of a SPECT image, the method comprising:

[0008] Acquire single-photon emission computed tomography (SPECT) images and computed tomography (CT) images of the same patient, and align the coordinate systems of the SPECT images and the CT images so that the pixels of the SPECT images and the pixels of the CT images have a corresponding relationship.

[0009] The functional distribution features of the SPECT image and the structural features of the CT image are extracted by a preset deep learning model, and the artifact probability value of each pixel in the SPECT image is determined based on the functional distribution features and the structural features.

[0010] By comparing the artifact probability threshold and the artifact probability value of each pixel, the pixels in the SPECT image are divided into artifact category and non-artifact category, and a binary mask image of the SPECT image is generated based on the classification results of the pixels.

[0011] At least one artifact region in the SPECT image is determined by the binary mask image, and the pixel values ​​of the artifact region are corrected by the structural information of the CT image.

[0012] The corrected artifact region is fused with the non-artifact region of the SPECT image to obtain the corrected SPECT image.

[0013] In one possible implementation, acquiring single-photon emission computed tomography (SPECT) images and computed tomography (CT) images of the same patient includes:

[0014] The SPECT and CT images of the same patient can be acquired simultaneously using a SPECT / CT integrated machine.

[0015] In one possible implementation, the training data for the preset deep learning model includes: multiple sets of SPECT images and CT images paired with the SPECT images, wherein the SPECT images in each set are labeled with multiple functional regions and the CT images in each set are labeled with multiple structural regions.

[0016] In one possible implementation, classifying pixels in the SPECT image into artifact and non-artifact categories by comparing an artifact probability threshold with the artifact probability value of each pixel includes:

[0017] For each pixel in the SPECT image, if the artifact probability value of the pixel is greater than the artifact probability threshold, then the pixel is classified as the artifact category; if the artifact probability value of the pixel is not greater than the artifact probability threshold, then the pixel is classified as the non-artifact category.

[0018] Obtain the classification results of the pixels in the SPECT image.

[0019] In one possible implementation, the mask pixels of the binary mask image correspond one-to-one with the pixels of the SPECT image;

[0020] The step of generating a binary mask image of the SPECT image based on the pixel classification results includes:

[0021] Traverse each pixel in the SPECT image. If the pixel belongs to the artifact category, set the pixel value of the mask pixel corresponding to the pixel to 1.

[0022] If the category of the pixel is the non-artifact category, then the pixel value of the mask pixel corresponding to the pixel is set to 0;

[0023] Obtain the binary mask image.

[0024] In one possible implementation, correcting the pixel values ​​of the artifact region using the structural information of the CT image includes:

[0025] The grayscale value of the artifact region is corrected by using the structural information and image correlation of the CT image. The image correlation is the feature correlation between the functional regions in the SPECT image and the structural regions in the CT image.

[0026] One possible implementation also includes:

[0027] The image quality of the corrected SPECT image is evaluated. If the image quality does not meet the requirements, the corrected SPECT image is then subjected to a second iteration of correction.

[0028] One possible implementation also includes:

[0029] During the reconstruction of the SPECT image, the structural information of the CT image is used as a constraint condition for the reconstruction process.

[0030] A second aspect of this application provides an image correction system for SPECT images, the image correction system for SPECT images comprising:

[0031] The acquisition unit is used to acquire single-photon emission computed tomography (SPECT) images and computed tomography (CT) images of the same patient, and align the coordinate systems of the SPECT images and the CT images so that the pixels of the SPECT images and the pixels of the CT images have a corresponding relationship.

[0032] The analysis unit is used to extract the functional distribution features of the SPECT image and the structural features of the CT image through a preset deep learning model, and to determine the artifact probability value of each pixel in the SPECT image based on the functional distribution features and the structural features.

[0033] The classification unit is used to classify the pixels in the SPECT image into artifact categories and non-artifact categories by comparing the artifact probability threshold and the artifact probability value of each pixel, and to generate a binary mask image of the SPECT image based on the classification results of the pixels.

[0034] The correction unit is used to determine at least one artifact region in the SPECT image through the binary mask image, and to correct the pixel value of the artifact region through the structural information of the CT image.

[0035] The fusion unit is used to fuse the corrected artifact region with the non-artifact region of the SPECT image to obtain the corrected SPECT image.

[0036] A third aspect of this application provides an electronic device, comprising at least one processor and a memory connected to the processor, wherein:

[0037] The memory is used to store computer programs;

[0038] The processor is used to execute the computer program so that the electronic device can implement the image correction method for SPECT images described in the first aspect or any implementation thereof.

[0039] Based on the above technical solution, this application provides a method, system, and related apparatus for image correction of SPECT images. The method acquires SPECT and CT images of the same patient and aligns them so that there is a correspondence between the pixels of the SPECT image and the pixels of the CT image. It extracts the functional distribution features of the SPECT image and the structural features of the CT image, and determines the artifact probability value of each pixel in the SPECT image based on these features. Each pixel in the SPECT image is then classified into artifact and non-artifact categories using an artifact probability threshold, and a binary mask image of the SPECT image is generated based on the classification results. At least one artifact region in the SPECT image is identified using the binary mask image, and the pixel values ​​in the artifact region are corrected using the structural information of the CT image. The corrected artifact region is then fused with the non-artifact region of the SPECT image to obtain the corrected SPECT image. This method learns and utilizes the image information correlation between CT images and SPECT images to help identify artifact regions in SPECT images and correct these artifact regions. It uses the actual structural information of CT images to correct SPECT images, eliminating dependence on prior assumptions and effectively improving the accuracy of SPECT image correction. Attached Figure Description

[0040] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and the originals and elements are not necessarily drawn to scale.

[0041] Figure 1 A flowchart illustrating an image correction method for a SPECT image provided in an embodiment of this application;

[0042] Figure 2 A schematic diagram of the structure of an image correction system for SPECT images provided in an embodiment of this application;

[0043] Figure 3 This application provides a hardware structure block diagram of an electronic device. Detailed Implementation

[0044] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.

[0045] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.

[0046] The terms “comprising” and “having”, and any variations thereof, in the specification and accompanying drawings of this application are intended to cover non-exclusive inclusion, such that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not expressly listed or inherent to such process, method, product, or apparatus.

[0047] In single-photon emission computed tomography (SPECT), factors such as scanning equipment noise, acquisition time, or patient movement can cause some projection data to be missing or incomplete, resulting in artifacts in the SPECT image. Alternatively, when the computer converts the original projection data, the computer's reconstruction algorithm has limited ability to supplement and correct missing data, leading to image distortion and artifacts. Artifacts are not actual structures or signals existing in the human body; rather, they are "illusions" appearing in the image. Artifacts can mask lesions or create false lesions, thus affecting the patient's diagnosis.

[0048] Current SPECT image correction methods rely on interpolation or optimization algorithms, but both suffer from unsatisfactory correction results. For example, optimization algorithms often rely on prior assumptions to supplement the SPECT image, but if the prior assumptions are not accurate, it may lead to SPECT image distortion. Furthermore, optimization algorithms also depend on parameter selection; improper parameter selection can also affect the correction effect of the SPECT image.

[0049] To address the aforementioned issues, this application provides an image correction method for SPECT images. This method utilizes structural information from CT images to correct artifacts in SPECT images, eliminating dependence on prior assumptions or parameters and effectively improving the accuracy of SPECT image correction. CT (Computed Tomography) is a medical imaging technique that uses X-rays to rotate around the patient, acquiring attenuation data of X-rays from multiple angles through detectors, and then using a computer to convert the attenuation data of X-rays into tomographic images of human tissues or organs.

[0050] CT images can include structural information about internal tissues or organs, such as their shape, location, size, and density, reflecting the anatomical structure of the human body. Therefore, in this embodiment, CT images can provide high-resolution anatomical information, especially in soft tissues and bones, while SPECT images can reflect the functional state of tissues or organs (such as blood distribution and metabolic status). Since each tissue or organ has a fixed structure and functional distribution that can be correlated, there is a certain image correlation between CT and SPECT images. By comparing the tissue distribution characteristics of CT and SPECT images at the same location (such as the same tissue or organ), it is possible to identify areas in the SPECT image that do not match the anatomical structure in the CT image, and these areas are highly likely to be artifact areas in the SPECT image. For example, the internal anatomical structure of a normal heart is fixed, as is the blood distribution within it. Furthermore, the structural information of the heart can be correlated with its blood distribution. If the structural information of the heart in a CT image does not correspond to the blood distribution in the heart region of a SPECT image (there is a difference between the original blood distribution and the structural information), it indicates the presence of artifacts in the heart region of the SPECT image, affecting the correlation between the CT and SPECT images. Therefore, the structural information of a CT image can serve as reference data for a SPECT image, helping to identify and correct artifacts in the SPECT image. The image correction method for SPECT images according to embodiments of this application will be described in detail below with reference to the accompanying drawings.

[0051] Reference Figure 1 , Figure 1 This is a flowchart illustrating an image correction method for SPECT images provided in an embodiment of this application, as shown below. Figure 1 As shown in the embodiment of this application, an image correction method for a SPECT image may include steps S10 to S14, which are described in detail below.

[0052] S10. Acquire single-photon emission computed tomography (SPECT) images and computed tomography (CT) images of the same patient, and align the coordinate systems of the SPECT images and the CT images so that the pixels of the SPECT images and the pixels of the CT images have a corresponding relationship.

[0053] In this context, a SPECT image can refer to a tomographic image reconstructed by a computer from original projection data, or it can refer to a SPECT image generated by using the structural information of a CT image as a constraint during the computer-based SPECT image reconstruction process. Specifically, this embodiment can optimize the spatial matching relationship between the projection data of the SPECT image and the structural data of the CT image during the SPECT image reconstruction process. This allows for the reduction of SPECT image artifacts by referencing the structural information of the CT image while maintaining the functional information of the SPECT image. Alternatively, the structural information of the CT image can be introduced as a regularization constraint during the SPECT image reconstruction process to ensure that the reconstruction result of the SPECT image is consistent with the structural information of the CT image. This guides the SPECT image reconstruction process with the structural information of the CT image, reducing the generation of SPECT image artifacts. This embodiment uses the structural information of the CT image to constrain the SPECT image reconstruction process, first reducing the generation of artifacts, and then using a pre-set deep learning model for artifact recognition and a repair network for artifact correction, which can effectively improve the image quality of the SPECT image.

[0054] To improve the accuracy of SPECT image correction, this embodiment selects to simultaneously acquire SPECT and CT images of the same patient using a SPECT / CT integrated machine. The SPECT / CT integrated machine can provide both types of imaging data simultaneously, ensuring that the acquired CT and SPECT images have the same size and resolution. A SPECT / CT integrated machine is a medical imaging device that combines single-photon emission computed tomography (SPECT) and computed tomography (CT). Alternatively, in another optional embodiment, CT and SPECT images can be acquired separately from different data sources.

[0055] After acquiring CT and SPECT images, this embodiment preprocesses and aligns their coordinate systems to improve image data quality while ensuring accurate alignment of the CT and SPECT images in the image regions, facilitating the determination of the correspondence between the image regions of the CT and SPECT images. Image coordinate system alignment refers to adjusting the coordinate systems of two or more images through a series of geometric transformations (such as translation and rotation) to establish a correspondence between the spatial positions (pixel coordinates of the two-dimensional image). Specifically, in this embodiment, the purpose of coordinate system alignment is to establish the pixel correspondence between the CT and SPECT images, thereby achieving the correspondence of image regions. The alignment algorithm used in this embodiment can be a rigid registration algorithm (in the registration process, the transformations between images only include translation and rotation, without any deformation (such as scaling)) or a non-rigid registration algorithm (in the registration process, the transformations between images include not only translation and rotation, but may also involve deformation).

[0056] Since CT images and SPECT images are acquired simultaneously from the SPECT / CT integrated machine, they have the same image resolution. A simple alignment operation can be performed using a rigid registration algorithm to ensure a one-to-one correspondence between the pixels in the CT image and the pixels in the SPECT image. Furthermore, two pixels with a one-to-one correspondence have the same pixel coordinates in the image.

[0057] Furthermore, if the data sources for CT images and SPECT images are different, their resolutions may differ. Therefore, without affecting the image quality of CT and SPECT images, a non-rigid registration algorithm is needed to align the coordinate systems of CT and SPECT images so that there is a one-to-many, many-to-one, or many-to-many correspondence between the pixels of the CT image and the pixels of the SPECT image.

[0058] S11. Extract the functional distribution features of SPECT images and the structural features of CT images through a preset deep learning model, and determine the artifact probability value of each pixel in the SPECT image based on the functional distribution features and structural features.

[0059] S12. By comparing the artifact probability threshold and the artifact probability value of each pixel, the pixels in the SPECT image are divided into artifact category and non-artifact category, and a binary mask image of the SPECT image is generated based on the classification results of the pixels.

[0060] The preset deep learning model can be a model specifically designed for identifying the correlation between CT images and SPECT images in this embodiment. The functional distribution features of the SPECT image can refer to the numerical values ​​or vectors representing the functional distribution extracted from the image region of the SPECT image by the preset deep learning model. The structural features of the CT image can refer to the numerical values ​​or vectors representing the anatomical structures extracted from the image region of the CT image by the preset deep learning model. The artifact probability value of a pixel can refer to the confidence or probability that the pixel belongs to an artifact region, predicted by the preset deep learning model based on the functional distribution features and structural features.

[0061] The pre-defined deep learning model can be a convolutional neural network (CNN), a generative adversarial network (GAN), or a deep autoencoder. Specifically, when the pre-defined deep learning model is a CNN, CT images and SPECT images can be used as input to identify artifact regions in the SPECT image. The artifacts are then corrected based on the structural information of the CT image (in the network's output layer, regions in the SPECT image that do not conform to the structural information are adjusted to eliminate artifacts). During the training of the CNN, CT images can be used as prior knowledge to supervise the CNN's learning of artifact recognition and correction. When the pre-defined deep learning model is a GAN, the generator network accepts CT images and SPECT images as input to generate artifact-free SPECT images. The discriminator judges whether the artifact-free SPECT images generated by the generator are genuine SPECT images. The generator and discriminator compete against each other, pushing the generator to produce more accurate and natural SPECT images. When the pre-defined deep learning model is a deep autoencoder, the encoder encodes the CT images and SPECT images into latent space representations, extracts high-dimensional features from the CT images and SPECT images respectively, and the decoder generates corrected SPECT images. Compared to optimization algorithms for artifact correction in SPECT images,

[0062] The specific network of the preset deep learning model used in this embodiment is an encoder-decoder structure network, such as U-Net (a deep learning architecture) or its variants (such as Swin-UNet (a Transformer-based medical image segmentation model)). This encoder-decoder structure can effectively capture multi-scale features of CT images and SPECT images and perform accurate pixel-level localization and segmentation.

[0063] The training data for the preset deep learning model may include: multiple sets of SPECT images and CT images paired with the SPECT images. Each set of SPECT images is labeled with multiple functional regions, and each set of CT images is labeled with multiple structural regions. During the training process of this preset deep learning model, the model learns the correct feature relationships between structural regions of CT images and functional regions of SPECT images through a large amount of training data. This allows the model to infer and correct artifact regions based on the structural information of the CT images, and optimizes the training of the deep learning model using the cross-entropy loss function. After the deep learning model is trained, the preset deep learning model of this embodiment is obtained. In the practical application of this preset deep learning model, if the relationship between a certain structural region of a CT image and the corresponding functional region of a SPECT image identified by the preset deep learning model differs from the correct relationship learned during training, it can be considered that artifacts exist in the functional region of the SPECT image.

[0064] This embodiment introduces a deep learning model and combines it with structural information from CT images to achieve efficient identification and accurate correction of artifacts in SPECT images. Furthermore, the deep neural network used in this embodiment can automatically learn the structure-function mapping relationship through large-scale SPECT / CT paired data, eliminating dependence on prior assumptions and improving the generalization ability and robustness of the preset deep learning model. Simultaneously, CT images, as structural guidance information, are embedded in the SPECT image correction process, effectively improving the structural consistency and functional accuracy of the images. Rapid output can be achieved through forward inference, meeting the needs of real-time clinical processing.

[0065] Specifically, the trained pre-defined deep learning model can have dual-channel input: one channel for CT images and the other for SPECT images. The encoder (multiple downsampling layers) progressively reduces the spatial dimension of the image feature maps, extracting the functional distribution features and structural features of the CT images. These features are then fused to obtain fused high-level features. The pre-defined deep learning model identifies the correlation between CT and SPECT images from these fused high-level features. The decoder (multiple upsampling layers) upsamples and refines the fused high-level features, resulting in a richer feature representation of the image for subsequent pixel classification.

[0066] In the output layer of the preset deep learning model, a probability map with the same size as the SPECT image (pixels correspond one-to-one) is output. The pixel value of each pixel in the probability map can represent the artifact probability value (between 0 and 1) of each pixel in the SPECT image. By setting the artifact probability threshold, the probability map is converted into a mask map to realize the recognition of artifact regions. The artifact probability threshold can be flexibly set according to the actual situation.

[0067] Specifically, this embodiment classifies each pixel in the SPECT image into an artifact category and a non-artifact category using an artifact probability threshold, and generates a mask image of the SPECT image based on the pixel classification results. For each pixel in the SPECT image, if the pixel's artifact probability value is greater than the artifact probability threshold, the pixel is classified as an artifact; if the pixel's artifact probability value is not greater than the artifact probability threshold, the pixel is classified as a non-artifact, thus obtaining the classification result of the pixels in the SPECT image. Iterating through each pixel in the SPECT image, if the pixel's category is an artifact, the pixel value of the corresponding mask pixel is set to 1 (image color is white); if the pixel's category is a non-artifact, the pixel value of the corresponding mask pixel is set to 0 (image color is black), thus obtaining a binary mask image.

[0068] S13. Determine at least one artifact region in the SPECT image using a binary mask image. In the artifact region, correct the pixel values ​​of the artifact region using the structural information of the CT image.

[0069] S14. The corrected artifact region is fused with the non-artifact region of the SPECT image to obtain the corrected SPECT image.

[0070] In this context, the artifact region refers to the image region containing artifacts determined from the SPECT image based on the binary mask image. This embodiment determines the artifact region from the SPECT image based on information in the binary mask image. Specifically, since the binary mask image and the SPECT image have the same size (one-to-one pixel correspondence), and in the binary mask image, the pixel value of the mask pixels for artifact regions is set to 1, while the pixel value of the mask pixels for non-artifact regions is set to 0, this embodiment can determine the corresponding pixels in the SPECT image based on the positions of the mask pixels with a pixel value of 1 in the binary mask image, thereby determining the artifact region in the SPECT image composed of these pixels (corresponding to a pixel value of 1 in the mask). During artifact correction, the binary mask image can directly guide the repair network to correct only the artifact regions in the SPECT image composed of these pixels (corresponding to a pixel value of 1 in the mask).

[0071] Of course, in another optional embodiment, this embodiment can directly fuse the binary mask image with the SPECT image to obtain the artifact region in the SPECT image. Specifically, since the binary mask image and the SPECT image have the same size (one-to-one pixel correspondence), the pixel values ​​of the mask pixels in the binary mask image can be multiplied by the corresponding pixel values ​​in the SPECT image to extract the artifact region from the SPECT image. Therefore, when the pixel values ​​of the binary mask image and the SPECT image are multiplied, the pixel values ​​of the artifact region in the SPECT image remain unchanged (multiplied by 1), while the pixel values ​​of the non-artifact region are all 0 (multiplied by 0), thus segmenting the artifact region in the SPECT image.

[0072] This embodiment can correct artifact regions using a trained inpainting network (such as U-Net or a generative adversarial network). During training, this inpainting network learns the correct correlation between structural regions in CT images and functional regions in SPECT images using a large amount of training data, enabling artifact correction based on the structural information of the CT image. Furthermore, the inpainting network needs to be trained on image correction quality. Each output image during training is scored, or L1 / L2 loss is used to calculate the difference between the corrected image and the ground truth reference image, providing feedback to optimize or adjust the network's parameters and improve image correction quality. Structural similarity loss can also be introduced to enhance image quality and structure preservation. When the inpainting network is a generative adversarial network, the discriminator within the network can be used to determine whether the corrected region is natural and consistent with the surrounding image structure.

[0073] Specifically, the input to the trained repair network can include: the original SPECT image, the CT image, and a binary mask. The binary mask guides the repair network to correct only the artifact regions in the original SPECT image. The repair network corrects the pixel values ​​in the artifact regions by using the structural information of the CT image and the image correlation (the feature correlation between the functional regions in the SPECT image and the structural regions in the CT image learned during training). Since the SPECT image is a grayscale image, this embodiment selects to adjust the grayscale values ​​of the pixels in the artifact regions to achieve artifact region correction in the SPECT image.

[0074] After artifact region correction is completed, the repair network performs pixel-level fusion of the corrected artifact region with the non-artifact region of the original SPECT image, achieving region stitching between the corrected artifact region and the non-artifact region of the original SPECT image to obtain a complete, artifact-corrected SPECT image. Alternatively, in another optional embodiment, the CT image and the artifact-corrected SPECT image can be fused to obtain a SPECT / CT fusion image for display, facilitating further assessment by physicians of the structural and functional consistency of tissues or organs.

[0075] Furthermore, after obtaining the artifact-corrected SPECT image, this embodiment can evaluate the image quality of the artifact-corrected SPECT image. If the image quality does not meet the requirements, the artifact-corrected SPECT image is input into the repair network again for a second iteration of correction. This embodiment can select a full-reference image quality evaluation index to evaluate the image quality of the artifact-corrected SPECT image, such as PSNR (Peak Signal-to-Noise Ratio) and SSIM (Structural Similarity Index). Specifically, this embodiment can select a reference "gold standard" SPECT image (which can be a high-quality SPECT image from the training data), calculate the PSNR and SSIM values ​​between the "gold standard" SPECT image and the artifact-corrected SPECT image. If the PSNR and SSIM values ​​do not meet the thresholds, the image quality of the artifact-corrected SPECT image can be considered unsatisfactory. For example, setting the PSNR threshold to 30 dB and the SSIM threshold to 0.95, when the PSNR value > 30 dB and the SSIM value > 0.95, the image quality is excellent; otherwise, the image quality is poor. Alternatively, in another implementation, a no-reference image quality evaluation metric (such as NIQE or BRISQUE) or a discriminator network can be trained to evaluate the image quality of the artifact-corrected SPECT image. NIQE (Natural Image Quality Evaluator) is a no-reference image quality evaluation algorithm that predicts the subjective quality score of an image by analyzing the statistical characteristics of the distorted image itself. BRISQUE (Blind / Referenceless Image Spatial Quality Evaluator) is a no-reference image quality evaluation algorithm that quantifies image quality by analyzing the statistical deviations of natural images.

[0076] This embodiment provides an image correction method for SPECT images. The method acquires SPECT and CT images of the same patient and aligns them to establish a correspondence between pixels in the SPECT and CT images. It extracts the functional distribution features of the SPECT image and the structural features of the CT image, and determines the artifact probability value of each pixel in the SPECT image based on these features. Each pixel in the SPECT image is then categorized into artifact and non-artifact categories using an artifact probability threshold, and a binary mask image of the SPECT image is generated based on the classification results. At least one artifact region in the SPECT image is identified using the binary mask image, and the pixel values ​​in the artifact region are corrected using the structural information from the CT image. The corrected artifact region is then fused with the non-artifact region of the SPECT image to obtain the corrected SPECT image. This method learns and utilizes the image information correlation between the CT and SPECT images to assist in identifying and correcting artifact regions in the SPECT image. By utilizing the actual structural information of the CT image to correct the SPECT image, it eliminates dependence on prior assumptions and effectively improves the accuracy of SPECT image correction. This embodiment also integrates quality assessment and feedback mechanisms, constructs a closed-loop optimization process, and further enhances the stability and intelligence of SPECT image correction effects.

[0077] The above describes an image correction method for SPECT images provided by embodiments of this application. The following will describe a system that applies the above-described image correction method for SPECT images.

[0078] Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of an image correction system for SPECT images provided in an embodiment of this application. Figure 2 As shown, the image correction system for the SPECT image may include:

[0079] The acquisition unit 100 is used to acquire single-photon emission computed tomography (SPECT) images and computed tomography (CT) images of the same patient, and align the coordinate systems of the SPECT images and the CT images so that the pixels of the SPECT images and the pixels of the CT images have a corresponding relationship.

[0080] The analysis unit 110 is used to extract the functional distribution features of the SPECT image and the structural features of the CT image through a preset deep learning model, and to determine the artifact probability value of each pixel in the SPECT image based on the functional distribution features and structural features.

[0081] The classification unit 120 is used to classify pixels in the SPECT image into artifact categories and non-artifact categories by comparing the artifact probability threshold and the artifact probability value of each pixel, and to generate a binary mask image of the SPECT image based on the classification results of the pixels.

[0082] The correction unit 130 is used to determine at least one artifact region in the SPECT image through a binary mask image, and to correct the pixel values ​​of the artifact region through the structural information of the CT image.

[0083] The fusion unit 140 is used to fuse the corrected artifact region with the non-artifact region of the SPECT image to obtain the corrected SPECT image.

[0084] In one possible implementation, the acquisition unit 100 acquires single-photon emission computed tomography (SPECT) images and computed tomography (CT) images of the same patient, which can be specifically configured as follows:

[0085] SPECT and CT images of the same patient can be acquired simultaneously using a SPECT / CT integrated machine.

[0086] In one possible implementation, the training data for the pre-defined deep learning model may include: multiple sets of SPECT images and CT images paired with the SPECT images, wherein the SPECT images in each set are labeled with multiple functional regions and the CT images in each set are labeled with multiple structural regions.

[0087] In one possible implementation, the classification unit 120 classifies pixels in the SPECT image into artifact and non-artifact categories by comparing an artifact probability threshold with the artifact probability value of each pixel. This can be specifically configured as follows:

[0088] For each pixel in the SPECT image, if the artifact probability value of the pixel is greater than the artifact probability threshold, the pixel is classified as an artifact; if the artifact probability value of the pixel is not greater than the artifact probability threshold, the pixel is classified as a non-artifact. The classification result of the pixels in the SPECT image is obtained.

[0089] In one possible implementation, the mask pixels of the binary mask image correspond one-to-one with the pixels of the SPECT image;

[0090] The classification unit 120 generates a binary mask image of the SPECT image based on the classification results of the pixels, which can be specifically configured as follows:

[0091] Iterate through each pixel in the SPECT image. If the pixel belongs to the artifact category, set the pixel value of the corresponding mask pixel to 1; if the pixel belongs to the non-artifact category, set the pixel value of the corresponding mask pixel to 0; thus obtaining a binary mask image.

[0092] In one possible implementation, the correction unit 130 corrects the pixel values ​​of the artifact region using the structural information of the CT image, which can be specifically configured as follows:

[0093] The grayscale values ​​of artifact regions are corrected by using structural information and image correlations in CT images. The image correlations are: the feature correlations between functional regions in SPECT images and structural regions in CT images.

[0094] In one possible implementation, the system may also include an evaluation unit:

[0095] The evaluation unit is used to evaluate the image quality of the corrected SPECT image. If the image quality does not meet the requirements, the corrected SPECT image is then subjected to a second iteration for correction.

[0096] In one possible implementation, the system also includes constraint units:

[0097] Constraint units are used to use the structural information of CT images as constraints in the reconstruction process of SPECT images.

[0098] This application also provides an electronic device in its embodiments. (See reference...) Figure 3 The diagram illustrates a structural schematic suitable for implementing the electronic device in the embodiments of this application. The electronic device in the embodiments of this application may include, but is not limited to, fixed terminals such as mobile phones, laptops, PDAs (personal digital assistants), PADs (tablet computers), desktop computers, etc. Figure 3 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0099] like Figure 3 As shown, the electronic device may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. When the electronic device is powered on, the RAM 303 also stores various programs and data required for the operation of the electronic device. The processing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0100] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, memory cards, hard drives, etc.; and communication devices 309. Communication device 309 allows electronic devices to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 Electronic devices with various devices are shown, but it should be understood that it is not required to implement or have all of the devices shown. More or fewer devices may be implemented or have instead.

[0101] This application also provides a computer program product including computer-readable instructions, which, when executed on an electronic device, cause the electronic device to implement any of the SPECT image correction methods provided in this application.

[0102] This application also provides a computer-readable storage medium carrying one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any of the SPECT image correction methods provided in this application.

[0103] It should also be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the system embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.

[0104] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, training equipment, or network device, etc.) to execute the methods described in the various embodiments of this application.

[0105] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.

[0106] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, training device, or data center to another website, computer, training device, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a training device or data center that integrates one or more available media. The available media may be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state drives (SSDs)).

[0107] The various embodiments in this specification are described in a related manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, the apparatus embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions of the method embodiments.

[0108] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0109] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for image correction of SPECT images, characterized in that, The image correction method for the SPECT image includes: Acquire single-photon emission computed tomography (SPECT) images and computed tomography (CT) images of the same patient, and align the coordinate systems of the SPECT images and the CT images so that the pixels of the SPECT images and the pixels of the CT images have a corresponding relationship. The functional distribution features of the SPECT image and the structural features of the CT image are extracted by a preset deep learning model, and the artifact probability value of each pixel in the SPECT image is determined based on the functional distribution features and the structural features. By comparing the artifact probability threshold and the artifact probability value of each pixel, the pixels in the SPECT image are divided into artifact category and non-artifact category, and a binary mask image of the SPECT image is generated based on the classification results of the pixels. At least one artifact region in the SPECT image is determined by the binary mask image, and the pixel values ​​of the artifact region are corrected by the structural information of the CT image. The corrected artifact region is fused with the non-artifact region of the SPECT image to obtain the corrected SPECT image.

2. The image correction method for SPECT images according to claim 1, characterized in that, The acquisition of single-photon emission computed tomography (SPECT) images and computed tomography (CT) images of the same patient includes: The SPECT and CT images of the same patient can be acquired simultaneously using a SPECT / CT integrated machine.

3. The image correction method for SPECT images according to claim 1, characterized in that, The training data for the preset deep learning model includes: multiple sets of SPECT images and CT images paired with the SPECT images. In each set, the SPECT images are labeled with multiple functional regions, and in each set, the CT images are labeled with multiple structural regions.

4. The image correction method for SPECT images according to claim 1, characterized in that, The process of classifying pixels in the SPECT image into artifact and non-artifact categories by comparing an artifact probability threshold with the artifact probability value of each pixel includes: For each pixel in the SPECT image, if the artifact probability value of the pixel is greater than the artifact probability threshold, then the pixel is classified as the artifact category; if the artifact probability value of the pixel is not greater than the artifact probability threshold, then the pixel is classified as the non-artifact category. Obtain the classification results of the pixels in the SPECT image.

5. The image correction method for SPECT images according to claim 1, characterized in that, The mask pixels of the binary mask image correspond one-to-one with the pixels of the SPECT image; The step of generating a binary mask image of the SPECT image based on the pixel classification results includes: Traverse each pixel in the SPECT image. If the pixel belongs to the artifact category, set the pixel value of the mask pixel corresponding to the pixel to 1. If the category of the pixel is the non-artifact category, then the pixel value of the mask pixel corresponding to the pixel is set to 0; Obtain the binary mask image.

6. The image correction method for SPECT images according to claim 1, characterized in that, The step of correcting the pixel values ​​of the artifact region using the structural information of the CT image includes: The grayscale value of the artifact region is corrected by using the structural information and image correlation of the CT image. The image correlation is the feature correlation between the functional regions in the SPECT image and the structural regions in the CT image.

7. The image correction method for SPECT images according to claim 1, characterized in that, Also includes: The image quality of the corrected SPECT image is evaluated. If the image quality does not meet the requirements, a second iteration of correction is performed on the corrected SPECT image.

8. The image correction method for SPECT images according to claim 1, characterized in that, Also includes: During the reconstruction of the SPECT image, the structural information of the CT image is used as a constraint condition for the reconstruction process.

9. An image correction system for SPECT images, characterized in that, The image correction system for the SPECT image includes: The acquisition unit is used to acquire single-photon emission computed tomography (SPECT) images and computed tomography (CT) images of the same patient, and align the coordinate systems of the SPECT images and the CT images so that the pixels of the SPECT images and the pixels of the CT images have a corresponding relationship. The analysis unit is used to extract the functional distribution features of the SPECT image and the structural features of the CT image through a preset deep learning model, and to determine the artifact probability value of each pixel in the SPECT image based on the functional distribution features and the structural features. The classification unit is used to classify the pixels in the SPECT image into artifact categories and non-artifact categories by comparing the artifact probability threshold and the artifact probability value of each pixel, and to generate a binary mask image of the SPECT image based on the classification results of the pixels. The correction unit is used to determine at least one artifact region in the SPECT image through the binary mask image, and to correct the pixel value of the artifact region through the structural information of the CT image. The fusion unit is used to fuse the corrected artifact region with the non-artifact region of the SPECT image to obtain the corrected SPECT image.

10. An electronic device, characterized in that, It includes at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is configured to execute the computer program to enable the electronic device to implement the image correction method for SPECT images as described in any one of claims 1 to 8.

Images