Infant standard body position reconstruction system with computer confrontation generation capability

Abstract

The invention provides an infant standard body position reconstruction system with computer adversarial generation capability. A data acquisition module acquires medical image data of multiple groups of body positions of an infant; the preprocessing module is used for preprocessing the medical image data; the adversarial network generation module comprises a generator and a discriminator; the generator obtains the preprocessed medical image data and then generates a first standard body position image; the discriminator receives the first standard body position image and the medical image data, judges the authenticity of the first standard body position image and the medical image data under the condition of unknown sources and assigns values to the authenticity; a training module obtains a sample training set, and the sample training set is circularly trained through a neural network until the authenticity assignment of an output first standard body position image reaches a preset value; after the evaluation module sets fitting parameters of authenticity assignment and returns the parameters to the training module, a second standard body position image is received and evaluation is executed; and the output module is used for outputting the second standard body position image which is evaluated to be qualified. According to the invention, the standard body position image of the baby is obtained by acquiring and reconstructing multiple groups of body position images of the baby.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical imaging, and particularly relates to a baby standard position reconstruction system with computer adversarial generation ability. Background Art

[0002] In pediatric medical imaging diagnosis, accurately obtaining the standard position images of infants is crucial for disease diagnosis and treatment. However, there are many problems in the current infant medical imaging acquisition and processing.

[0003] On the one hand, due to their own particularity, it is difficult for infants to cooperate with the standard position scanning like adults. Infants usually cannot keep still autonomously and are prone to crying and wriggling their bodies, which makes the traditional medical imaging acquisition process extremely difficult. When performing scans such as CT and MRI, the obtained images often have motion artifacts, affecting doctors' accurate judgment of the infant's body structure and lesions. For example, when scanning an infant's brain, even a slight head movement may cause the image to be blurred, making it difficult to clearly display the brain tissue and potential lesions. On the other hand, the existing medical imaging processing methods have limited effects when dealing with infant images. Traditional image reconstruction algorithms usually cannot effectively remove motion artifacts, nor can they convert non-standard position images into standard positions. Moreover, most of the existing software and systems are designed for adults and lack consideration for the small body size, special body proportions, and tissue characteristics of infants. For example, the bones of infants are relatively soft, and the tissue contrast is different from that of adults, which greatly reduces the accuracy of traditional image segmentation and reconstruction methods for infant images.

[0004] In summary, the existing medical imaging technology has obvious deficiencies in the reconstruction of the standard position of infants, and there is an urgent need for a baby standard position reconstruction system with computer adversarial generation ability to improve the quality and diagnostic accuracy of infant medical images. Summary of the Invention

[0005] Therefore, the present invention provides a baby standard position reconstruction system with computer adversarial generation ability to improve the quality and diagnostic accuracy of infant medical images.

[0006] The above technical object of the present invention is achieved through the following solutions:

[0007] A baby standard position reconstruction system with computer adversarial generation ability, comprising:

[0008] A data acquisition module configured to acquire medical image data of multiple positions of an infant;

[0009] A preprocessing module connected to the data acquisition module and configured to perform preprocessing on the medical image data;

[0010] An adversarial network generation module, including a generator and a discriminator;

[0011] The generator is configured to generate a first standard body position image after obtaining the preprocessed medical image data;

[0012] The discriminator is configured to receive the first standard body position image and the medical image data, and judge the authenticity of the two and assign a value to the authenticity under the condition of unknown source;

[0013] A training module, connected to the adversarial network generation module, obtains a sample training set constructed by the first standard body position image and the medical image data, and then cycles through neural network training until the authenticity assignment of the output first standard body position image reaches a preset value;

[0014] An evaluation module is configured to set fitting parameters for the authenticity assignment. After returning the parameters to the training module under the set fitting parameters, it receives the second standard body position image trained by the training module and output by the generator and performs an evaluation;

[0015] An output module outputs the second standard body position image that has passed the evaluation.

[0016] As a preferred method, the data acquisition module includes an image acquisition sub-module and a positioning sub-module;

[0017] The image acquisition sub-module is configured to obtain medical image data of an infant, and the positioning sub-module is used to obtain a marked image in the medical image data.

[0018] As a preferred method, the positioning sub-module is configured to obtain an image within a preset range of an identifier pasted at the marked position of the infant to determine the marked image.

[0019] As a preferred method, the positioning sub-module is configured to obtain an image within a preset range of the marked feature position obtained from the medical image data processed by the preprocessing module to determine the marked image.

[0020] As a preferred method, when the data acquisition module is configured to obtain medical image data of multiple body positions of an infant, the marked image is obtained at least within the field of view of the image from at least two perspectives.

[0021] As a preferred method, when the generator generates the first standard body position image, it specifically includes the following steps:

[0022] Establish the skeleton of the first standard body position image, and attach the marked image at the position corresponding to the marked image on the skeleton;

[0023] Obtain the perspective views of the marked image within at least two perspectives, and generate a pixel-generated pre-filled image of the marked image within the maximum range from the standard image to the perspective;

[0024] Attach the image in the medical image data to the bone again, obtain the maximum boundary position of each marked image in any perspective and the perspective position in another perspective, calculate its spatial position, and obtain the maximum plane fitting boundary in one perspective;

[0025] Fill pixels from the position of the marked image to the maximum plane fitting position, and both the average pixel change rate and the pixel value of the pixels are the average values of the marked image and the medical image data;

[0026] Maintain the marked image within this perspective and after generating the image;

[0027] Obtain the same marked image in another perspective, and after spatial transformation, obtain a perspective degree adjustment that keeps the marked image at the center of the perspective and makes the distance between the marked image and the true boundary of the medical image data in the other perspective equal to the maximum plane fitting edge, and after correspondingly scaling the size of the pixel and the value of the pixel;

[0028] Resample to obtain the average pixel change rate and pixel value of the pixels at the other time and then average-fill them into the previously generated values;

[0029] Loop and execute until the positions of each marked image in all the perspectives where it exists are completed, and obtain an intermediate generated image;

[0030] Increase the number of marked images processed simultaneously, and loop to regenerate the intermediate generated image. The steps of the regeneration are the same as the steps of generating one marked image;

[0031] Until all the marked images of the bone are executed, generate a first standard body position image.

[0032] As a preferred method, the fitting parameters include the average change rate of pixels in the randomly sampled image plane, the average change rate of pixels in the image plane of the specified sampled real part and the generated part, and the pixel continuity of the boundary.

[0033] The second aspect of the present invention provides a method, including the following steps:

[0034] Obtain the medical image data of an infant under multiple body positions;

[0035] Preprocess the medical image data;

[0036] Establish the bones of the first standard body position image, and attach the marker image at the position of the bones corresponding to the marker image;

[0037] Obtain the perspective views of the marker image within at least two perspectives, and generate a pixel-generated pre-filled image within the maximum range from the standard image to the perspective;

[0038] Attach the images in the medical image data to the bones again, obtain the maximum boundary position of each marker image in any perspective and the perspective position in another perspective, calculate its spatial position, and obtain the maximum plane fitting boundary in one perspective;

[0039] Fill pixels from the position of the marker image to the maximum plane fitting position, and both the average pixel change rate and the pixel value of the pixels are the average values of the marker image and the medical image data;

[0040] Maintain the marker image within this perspective and after generating the image;

[0041] Obtain the same marker image in another perspective, and after spatial transformation, obtain a perspective degree adjustment that keeps the marker image at the center of the perspective and makes the distance between the marker image and the true boundary of the medical image data in the other perspective equal to the maximum plane fitting edge, and after correspondingly scaling the size of the pixel and the value of the pixel;

[0042] Resample to obtain the average pixel change rate and pixel value of the pixels at the other time, and then average-fill them to the values generated last time;

[0043] Loop and execute until the positions of each marker image in all the perspectives where it exists are completed, and obtain an intermediate generated image;

[0044] Increase the number of marker images processed simultaneously, and loop to regenerate the intermediate generated image. The steps of the regeneration are the same as the steps of generating one marker image;

[0045] Until all the marker images on the bones are executed, generate the first standard body position image.

[0046] The third aspect of the present invention provides an electronic device.

[0047] The fourth aspect of the present invention provides a computer-readable storage medium.

[0048] The above technical solutions of the present invention have the following advantages compared with the prior art:

[0049] The present invention realizes the reconstruction of the body position images of infants in the process of medical image processing. Traditionally, obtaining standard body position images may require multiple positioning and reshooting of infants, which not only consumes time but also may cause unnecessary discomfort to the infants. This system uses the computer adversarial generation ability to reconstruct standard body position images by using existing multiple groups of body position medical image data, greatly reducing the need for reshooting and saving medical resources and time costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] Figure 1 FIG. is a schematic structural diagram of the system provided in Embodiment 1 of the present invention.

[0051] Figure 2 FIG. is a schematic structural diagram of the electronic device provided in Embodiment 2 of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0053] Embodiment 1

[0054] The present disclosure provides a system for reconstructing the standard body position of infants with computer adversarial generation ability, as Figure 1 shown, including:

[0055] A data acquisition module configured to acquire medical image data of multiple body positions of an infant;

[0056] A preprocessing module connected to the data acquisition module and configured to perform preprocessing on the medical image data;

[0057] An adversarial network generation module including a generator and a discriminator;

[0058] The generator is configured to generate a first standard body position image after acquiring the preprocessed medical image data;

[0059] The discriminator is configured to receive the first standard body position image and the medical image data, and judge the authenticity of the two and assign a value to the authenticity under the condition of unknown source;

[0060] A training module connected to the adversarial network generation module, obtaining a sample training set constructed by the first standard body position image and the medical image data, and then cyclically training through a neural network until the authenticity value of the output first standard body position image reaches a preset value;

[0061] An evaluation module, configured to set fitting parameters for the authenticity assignment, and after returning the parameters to the training module under the set fitting parameters, receive a second standard body position image trained by the training module and output by the generator and perform an evaluation;

[0062] An output module, outputting the second standard body position image that has passed the evaluation.

[0063] As a preferred method, the data acquisition module includes an image acquisition sub-module and a positioning sub-module;

[0064] The image acquisition sub-module is configured to acquire medical image data of an infant, and the positioning sub-module is used to acquire a marker image in the medical image data.

[0065] As a preferred method, the positioning sub-module is configured to acquire an image within a preset range of a marker attached to a marker position of an infant to determine the marker image.

[0066] As a preferred method, the positioning sub-module is configured to acquire an image within a preset range of a marker feature position obtained from the medical image data processed by the preprocessing module to determine the marker image.

[0067] As a preferred method, when the data acquisition module is configured to acquire medical image data of multiple body positions of an infant, at least two perspectives are used to acquire an image in which the marker image is within the field of view of the image.

[0068] As a preferred method, when the generator generates the first standard body position image, it specifically includes the following steps:

[0069] Establish the skeleton of the first standard body position image, and attach the marker image at the position corresponding to the marker image on the skeleton;

[0070] Obtain the perspective views of the marker image from at least two perspectives, and generate a pre-filled image of the pixels of the marker image within the maximum range from the standard image to the perspective;

[0071] Attach the image in the medical image data to the skeleton again, obtain the maximum boundary position of each marker image in any perspective and the perspective position in another perspective, calculate its spatial position, and obtain the maximum plane fitting boundary in one perspective;

[0072] Fill pixels from the marker image position to the maximum plane fitting position, and the average pixel change rate and pixel value of the pixels are both the average of the marker image and the medical image data;

[0073] After maintaining the marked image within this perspective and generating an image;

[0074] Obtain the same marked image from another perspective, and after spatial transformation, obtain a result such that the marked image is maintained at the center of the perspective and its perspective degree is adjusted so that the distance between the marked image and the true boundary of the medical image data in the other perspective is equal to the maximum plane fitting edge, and after correspondingly scaling the size of the pixels and the value of the pixels;

[0075] After resampling to obtain the average pixel change rate and pixel values of the pixels at the other time, average and fill them into the values generated last time;

[0076] Loop and execute until the positions of each marked image in all the perspectives where it exists are processed, and obtain an intermediate generated image;

[0077] Increase the number of marked images processed simultaneously, and loop to regenerate the intermediate generated image. The steps of the regeneration are the same as the steps of generating one marked image;

[0078] Until all the marked images of the bone are executed completely, generate a first standard body position image.

[0079] As a preferred method, the fitting parameters include the average change rate of pixels in a randomly sampled image plane, the average change rate of pixels in the image planes of the specified sampled real part and the generated part, and the pixel continuity of the boundary.

[0080] Specifically, in the embodiments of the present disclosure, the image acquisition sub-module: uses professional medical imaging devices, such as X-ray machines, CT scanners or ultrasonic devices, etc., to take images of the baby in different body positions. These devices need to have high resolution and low radiation dose (for X-ray and CT) to ensure obtaining clear and safe medical image data. During the shooting process, it is necessary to ensure that the baby is in a comfortable and safe state, and auxiliary fixing devices can be used to avoid blurred images caused by the baby's movement. At the same time, images at different angles and postures are collected as needed to ensure the diversity of the data.

[0081] The positioning sub-module has two setting methods. If the method of pasting identification objects is adopted to determine the marked image, first paste special identification objects at specific marked positions on the baby's body (such as joints, bone protrusions, etc.). These identification objects should have the characteristics of being easily recognizable in medical images, for example, using identification objects with specific shapes and materials (such as materials containing a small amount of metal components but harmless to the baby). Then, obtain the image within the preset range of the identification object through the imaging acquisition device as the marked image.

[0082] If the marked feature positions are obtained from the preprocessed medical image data, the image processing algorithm is used to identify the positions with specific features in the medical image (such as specific turning points of the bone contour, special soft tissue texture features, etc.), and the images within its preset range are determined as marked images. When collecting data, at least two perspectives are used for shooting to ensure that the marked images are within the field of view of the image in each perspective. This can be achieved by adjusting the angle and position of the image acquisition device, such as using a rotatable and movable scanning frame.

[0083] The preprocessing module performs a variety of preprocessing operations on the obtained medical image data. First, denoising processing is carried out. Algorithms such as median filtering and Gaussian filtering can be used to remove the noise points in the image and improve the image quality. Then, image enhancement operations are performed, such as adjusting the contrast and brightness to make the details in the image clearer. For images obtained from different perspectives or different devices, normalization processing is carried out to unify the image size, gray scale range, etc. into a standard range for convenient subsequent processing.

[0084] Generator: When generating the first standard body position image, follow the following detailed steps.

[0085] First, establish a bone model of the first standard body position image according to the anatomical knowledge of the infant. This can be based on the existing medical bone template library and select the bone structure that matches the infant's age and approximate body type. Then, accurately attach the marked image to the corresponding position on the bone. For example, if the marked image is an identifier at the joint, it is corresponding to the joint position of the bone model.

[0086] Next, obtain the perspective views of the marked image in at least two perspectives. This is achieved through the perspective transformation algorithm. According to the known perspective parameters and the position relationship of the marked image in different perspectives, the perspective effect is calculated. Within the maximum range from the standard image to the perspective, a pre-filled image is generated according to a certain pixel generation rule. For example, based on the distribution law of the pixels around the marked image, pre-filled pixels can be generated by interpolation and other methods.

[0087] Attach the images in the medical image data to the bone again. In this process, the maximum boundary position of each marked image in any perspective and the perspective position in another perspective are used to calculate its accurate spatial position through the three-dimensional space coordinate calculation method, so as to obtain the maximum plane fitting boundary in one perspective.

[0088] Fill the pixels from the marked image position to the maximum plane fitting position. During the filling process, ensure that the average pixel change rate and pixel value of the pixels are both the average values of the marked image and the medical image data. This can be achieved by statistically analyzing the pixel features in the corresponding areas of the marked image and the medical image data. During the filling process, keep the marked image in this perspective unchanged to ensure its accuracy as a positioning and reference.

[0089] Then, obtain the same marked image from another perspective and maintain it at the center of the perspective through spatial transformation operations (such as rotation, translation, scaling, etc.). At the same time, adjust its perspective degree so that the distance between the marked image and the real boundary of the medical image data in the other perspective is equal to the maximum plane fitting edge. In this process, appropriate transformation parameters are calculated according to the perspective principle and geometric relationship. Then, correspondingly scale the size of the pixel and the value of the pixel to match the characteristics such as the proportion and brightness of the overall image.

[0090] Then, obtain the average pixel change rate and pixel value of the pixels at another time through resampling operation, and evenly fill them into the values generated last time. This resampling process can adopt methods such as uniform sampling or sampling according to pixel distribution weights.

[0091] Loop through the above steps until the positions of each marked image in all its existing perspectives are processed to obtain the intermediate generated image. Then, gradually increase the number of marked images processed simultaneously, and loop to regenerate the intermediate generated image. The steps of each regeneration are the same as those of generating a marked image. Until all the marked images of the bone are executed, the first standard body position image is finally generated.

[0092] The discriminator receives the first standard body position image generated by the generator and the original medical image data. Using a deep neural network structure, such as a convolutional neural network (CNN), it extracts features from the input image. Under the condition of unknown source, by learning a large number of training samples, a model that can distinguish real and generated images is established. For the input image, judge the authenticity between the two according to its image features (such as texture, edge, brightness distribution, etc.), and assign a value to the authenticity according to the preset evaluation criteria. This value assignment can be a numerical value between 0 and 1, where 0 represents completely false and 1 represents completely true.

[0093] The training module obtains the first standard body position image and the original medical image data from the generator to construct a sample training set. Input this training set into the neural network for loop training. During the training process, according to the value assignment result of the discriminator for the authenticity of the generated image, use the backpropagation algorithm to adjust the network parameters of the generator and the discriminator. For example, if the value assignment for the authenticity of the generated image is low, adjust the parameters of the generator to make it generate a more realistic image. Continuously repeat this training process until the value assignment for the authenticity of the output first standard body position image reaches the preset value. This preset value can be set according to the actual application requirements and the accuracy requirements for reconstructing the standard body position image, for example, set to 0.9.

[0094] The evaluation module sets fitting parameters for authenticity assignment. These fitting parameters include the average change rate of pixels in the randomly sampled image plane, the average change rate of pixels in the image plane of the specified sampled real part and the generated part, the pixel continuity of the boundary, etc. The second standard body position image output by the trained generator is evaluated according to these fitting parameters. By comparing the differences between the generated image and the real medical image in these parameters, the quality of the generated image is judged. If the generated image has a high degree of matching with the real image in these parameters, the evaluation is considered qualified. If the evaluation is unqualified, the relevant parameter information is returned to the training module to continue adjusting the training process.

[0095] The output module outputs the second standard body position image that has passed the evaluation. The output image can be stored in a local database in a standard medical image format (such as DICOM format) for subsequent medical diagnosis, research, etc. At the same time, the output image can be displayed on the interface of a dedicated medical image viewing software for medical staff or researchers to visually view and analyze.

[0096] The second aspect of the embodiments of the present disclosure provides a method for reconstructing the standard body position of an infant with computer-generated adversarial ability, including the following steps:

[0097] Obtain medical image data of an infant in multiple body positions;

[0098] Preprocess the medical image data;

[0099] Establish the skeleton of the first standard body position image, and attach the marker image at the position of the skeleton corresponding to the marker image;

[0100] Obtain the perspective views of the marker image within at least two perspectives, and generate a pre-filled image of the pixels of the marker image within the maximum range from the standard image to the perspective;

[0101] Attach the image in the medical image data to the skeleton again, obtain the maximum boundary position of each marker image in any perspective and the perspective position in another perspective, calculate its spatial position, and obtain the maximum plane fitting boundary in one perspective;

[0102] Fill pixels from the position of the marker image to the maximum plane fitting position, and the average pixel change rate and pixel value of the pixels are the average values of the marker image and the medical image data;

[0103] After maintaining the marker image and the generated image within this perspective;

[0104] Obtain the same marked image from another perspective, and after spatial transformation, obtain the marked image that maintains at the center of the perspective and adjusts its perspective degree so that the distance between the marked image and the true boundary of the medical image data in the other perspective is equal to the maximum plane fitting edge, and after correspondingly scaling the size of the pixels and the value of the pixels;

[0105] After resampling to obtain the average pixel change rate and pixel value of the pixels at the other time, average and fill them into the values generated last time;

[0106] Loop and execute until the positions of each marked image in all its existing perspectives are completed, and obtain an intermediate generated image;

[0107] Increase the number of marked images processed simultaneously, and loop to regenerate the intermediate generated image. The steps of the regeneration are the same as the steps of generating one marked image;

[0108] Until all the marked images of the bone are executed, generate a first standard body position image.

[0109] The method part in the embodiments of the present disclosure has been elaborated in detail in the system part, and will not be repeated here.

[0110] Embodiment 2

[0111] Combined Figure 2 As shown, the embodiments of the present disclosure provide an electronic device, including a processor 30 and a memory 31. Optionally, the electronic device may further include a communication interface 32 and a bus 33. Among them, the processor 30, the communication interface 32, and the memory 31 can complete mutual communication through the bus 33. The communication interface 32 can be used for information transmission. The processor 30 can call the logical instructions in the memory 31 to execute the method of the above-mentioned Embodiment 1.

[0112] The embodiments of the present disclosure provide a storage medium storing computer-executable instructions, and the computer-executable instructions are set to execute the method as in Embodiment 1.

[0113] The above storage medium may be a transient computer-readable storage medium or a non-transient computer-readable storage medium. The non-transient storage medium includes: various media such as a USB flash drive, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disc that can store program codes, and may also be a transient storage medium.

[0114] The above description and the accompanying drawings fully illustrate the embodiments of the present disclosure, enabling those skilled in the art to practice them. Other embodiments may include structural, logical, electrical, process, and other changes. Embodiments merely represent possible variations. Unless explicitly required, individual components and functions are optional, and the order of operations may vary. Parts and features of some embodiments may be included in or replace parts and features of other embodiments. Moreover, the terms used in this application are only for describing embodiments and do not limit the claims. As used in the description of embodiments and claims, unless the context clearly indicates otherwise, the singular forms "a", "an", and "the" are intended to also include the plural forms. Similarly, as used in this application, the term "and / or" refers to any and all possible combinations including one or more of the associated listed items. Additionally, when used in this application, the term "comprise" and its variants "comprises" and / or "comprising" etc. mean the presence of the stated features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or groups of these. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, or apparatus including the element. Herein, each embodiment may focus on the differences from other embodiments, and the same or similar parts among the embodiments may be referred to each other. For the methods, products, etc. disclosed in the embodiments, if they correspond to the method parts disclosed in the embodiments, the relevant parts may refer to the description of the method parts.

[0115] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner may depend on the specific application and design constraints of the technical solution. The technician can use different methods for each specific application to implement the described functions, but such implementation should not be considered to exceed the scope of the embodiments of the present disclosure. The technician can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described devices, devices, and units can refer to the corresponding processes in the foregoing method embodiments, and will not be elaborated herein.

[0116] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of code, or a portion thereof that contains one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions noted in the blocks may occur out of the order noted in the figures. For example, two consecutive blocks may in fact be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur out of the order disclosed in the descriptions, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may in fact be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented by a dedicated hardware-based device that performs the specified functions or actions, or can be implemented by a combination of dedicated hardware and computer instructions.

Claims

1. A baby standard posture reconstruction system with computer adversarial generation capability, characterized in that: include: A data acquisition module is configured to obtain medical imaging data of multiple groups of body positions of the infant; a preprocessing module, connected to the data acquisition module and configured to perform preprocessing on the medical image data; Adversarial network generation module, including generator and discriminator; The generator is configured to generate a first standard body position image after acquiring the preprocessed medical image data; The discriminator is configured to receive the first standard body position image and the medical image data, determine the authenticity of the two under the condition of unknown sources and assign a value to the authenticity; A training module, connected to the adversarial network generation module, obtains a sample training set constructed by the first standard body position image and the medical image data, and then performs a neural network cyclic training until the authenticity assignment of the output first standard body position image reaches a preset value; An evaluation module is configured to set fitting parameters for the authenticity assignment, and after returning parameters to the training module under the set fitting parameters, receive the second standard body position image trained by the training module and output by the generator and perform evaluation; The output module outputs the second standard body position image that has been evaluated as qualified.

2. The infant standard body position reconstruction system with computer adversarial generation capability according to claim 1, characterized in that: The data acquisition module includes an image acquisition submodule and a positioning submodule; The image acquisition submodule is configured to acquire medical image data of the infant, and the positioning submodule is used to acquire a marked image in the medical image data.

3. The infant standard body position reconstruction system with computer adversarial generation capability according to claim 2, characterized in that: The positioning submodule is configured to acquire an image within a preset range of a marker affixed to a marking position of the baby to determine the marking image.

4. The infant standard body position reconstruction system with computer adversarial generation capability according to claim 2, characterized in that: The positioning submodule is configured to obtain an image in which the position of the marking feature obtained from the medical image data obtained after being processed by the preprocessing module is within a preset range to determine the marking image.

5. The infant standard body position reconstruction system with computer adversarial generation capability according to claim 4, characterized in that: The data acquisition module is configured to acquire the medical image data of multiple groups of infant body positions, and to acquire the images in which the marked images are located within the field of view of the images in at least two viewing angles.

6. The infant standard body position reconstruction system with computer adversarial generation capability according to claim 5, characterized in that: When the generator generates the first standard body position image, the following steps are specifically included: Establishing a skeleton of the first standard body position image, and attaching the marked image to the skeleton at a position corresponding to the marked position; Obtaining the perspective angle of the marked image in at least two viewing angles, and generating a pre-filled image by generating pixels of the marked image from the standard image to the maximum range of the perspective; Attach the image in the medical image data to the skeleton for a second time, obtain the maximum boundary position of each of the marked images at any viewing angle and the perspective position of another viewing angle, calculate its spatial position, and obtain the maximum plane fitting boundary at one viewing angle; Filling pixels from the marked image position to the maximum plane fitting position, wherein the average pixel change rate and pixel value of the pixels are both average values ​​of the marked image and the medical image data; After maintaining the said marked image within the viewing angle and generating the image; Obtaining the same marked image at another viewing angle, and after spatial transformation, maintaining the marked image at the center of the viewing angle and adjusting its perspective so that the distance between the marked image and the real boundary of the medical image data at the other viewing angle is equal to the maximum plane fitting edge, and scaling the size and value of the pixels accordingly; Resampling to obtain the average pixel change rate and pixel value of the pixels at the other time and then filling them with the values ​​generated last time; Execute the loop until each of the marked images is at the position of all existing viewing angles, and obtain an intermediate generated image; Increasing the number of the marked images processed simultaneously, cyclically regenerating the intermediate generated images, the regeneration step being the same as the step of performing the generation of one marked image; After all the labeled images of the skeleton are executed, a first standard body position image is generated.

7. The infant standard body position reconstruction system with computer adversarial generation capability according to claim 1, characterized in that: The fitting parameters include the average change rate of pixels in the randomly sampled image surface, the average change rate of pixels in the image surface of the real part and the generated part of the specified sampling, and the pixel continuity of the boundary.

8. A method for reconstructing standard infant posture with computer adversarial generation capability, characterized in that: The steps include: Obtain medical imaging data of infants in multiple positions; Preprocessing medical imaging data; Establishing the skeleton of the first standard body position image, and attaching the marker image at the position of the skeleton corresponding to the marker image; Obtaining the perspective angle of the marked image in at least two viewing angles, and generating a pre-filled image by generating pixels of the marked image from the standard image to the maximum range of the perspective; Attach the image in the medical image data to the skeleton for a second time, obtain the maximum boundary position of each of the marked images at any viewing angle and the perspective position of another viewing angle, calculate its spatial position, and obtain the maximum plane fitting boundary at one viewing angle; Filling pixels from the marked image position to the maximum plane fitting position, wherein the average pixel change rate and pixel value of the pixels are both average values ​​of the marked image and the medical image data; After maintaining the said marked image within the viewing angle and generating the image; Obtaining the same marked image at another viewing angle, and after spatial transformation, maintaining the marked image at the center of the viewing angle and adjusting its perspective so that the distance between the marked image and the real boundary of the medical image data at the other viewing angle is equal to the maximum plane fitting edge, and scaling the size and value of the pixels accordingly; Resampling to obtain the average pixel change rate and pixel value of the pixels at the other time and then filling them with the values ​​generated last time; Execute the loop until each of the marked images is at the position of all existing viewing angles, and obtain an intermediate generated image; Increasing the number of the marked images processed simultaneously, cyclically regenerating the intermediate generated images, the regeneration step being the same as the step of performing the generation of one marked image; After all the labeled images of the skeleton are executed, a first standard body position image is generated.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method as claimed in claim 8 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method as claimed in claim 8 is implemented.

Images