CN115797218A

CN115797218A - CT image reconstruction method and device and computer equipment

Info

Publication number: CN115797218A
Application number: CN202211620440.8A
Authority: CN
Inventors: 刘晓鸣; 乔治
Original assignee: Beijing Lianying Intelligent Imaging Technology Research Institute
Current assignee: Beijing Lianying Intelligent Imaging Technology Research Institute
Priority date: 2022-12-15
Filing date: 2022-12-15
Publication date: 2023-03-14

Abstract

The application relates to a CT image reconstruction method, a CT image reconstruction device and computer equipment. The method comprises the following steps: acquiring at least one first CT image; processing the first CT image by adopting at least one contrast adjusting model to obtain at least one second CT image; extracting the features of the second CT images to obtain a feature vector group; and decoding and reconstructing the first CT image and the feature vector group, and outputting a CT reconstructed image. By adopting the method, the display contrast can be improved, and the reconstruction effect of the CT image is further improved.

Description

CT image reconstruction method and device and computer equipment

Technical Field

The present application relates to the field of medical image processing technologies, and in particular, to a CT image reconstruction method, an apparatus, and a computer device.

Background

With the development of CT (Computed Tomography) technology, CT imaging is widely applied to spinal imaging examination items due to its advantages of rapid imaging, low examination cost, and clear visualization of intervertebral disc pneumatosis and calcification. However, CT imaging tends to blur the visualization of disc degeneration, spinal cord deformities, dural sac compression, nerve root compression, and the like.

The current CT imaging mode or the traditional method has the problems of low display contrast and the like.

Disclosure of Invention

In view of the above, it is desirable to provide a CT image reconstruction method, apparatus and computer device capable of improving display contrast.

In a first aspect, the present application provides a CT image reconstruction method, including:

acquiring at least one first CT image;

processing the first CT image by using at least one contrast adjusting model to obtain at least one second CT image;

extracting the features of each second CT image to obtain a feature vector group;

and decoding and reconstructing the first CT image and the feature vector group, and outputting a CT reconstructed image.

In one embodiment, the method further comprises:

acquiring a plurality of original CT images and a plurality of groups of target contrast images obtained by aiming at the original CT images;

and inputting the original CT image and the target contrast image into a generation countermeasure network for training to obtain at least one contrast adjustment model.

In one embodiment, each set of target contrast images includes at least one contrast-adjusted image; inputting an original CT image and a target contrast image into a generation confrontation network for training to obtain at least one contrast regulation model, and the method comprises the following steps:

and inputting the original CT image and each group of target contrast images into a countermeasure network for training to obtain at least one contrast adjustment model.

In one embodiment, acquiring at least one first CT image includes:

acquiring a target CT image; the target CT image includes at least one intervertebral disc; positioning intervertebral discs of the target CT image to obtain central points of the intervertebral discs;

sequentially rotationally correcting the target CT image based on the central point of each intervertebral disc and a preset rotational correction rule to obtain an interested area of each intervertebral disc;

and cutting the interested area of each intervertebral disc to obtain at least one first CT image aiming at the intervertebral disc.

In one embodiment, the intervertebral disc positioning processing on the target CT image to obtain the central point of each intervertebral disc includes:

performing multi-target image segmentation on the target CT image to obtain a vertebral body segmentation result; the vertebral body segmentation result comprises CT images of a plurality of vertebral bodies and a vertebral body label corresponding to each vertebral body;

fitting a minimum external cuboid aiming at the vertebral body to the vertebral body segmentation result to obtain a central point of the vertebral body; the central point of the vertebral body is the gravity center of the minimum external cuboid;

obtaining a spine fitting curve based on the central point of each vertebral body;

and obtaining the central point of each intervertebral disc according to the spine fitting curve and the central point of each vertebral body.

In one embodiment, cropping the region of interest of each intervertebral disc to obtain each first CT image comprises:

acquiring the length of each end plate adjacent to the intervertebral disc;

generating a cutting cuboid based on the maximum value of the lengths of the end plates;

and cutting the region of interest of the intervertebral disc by using a cutting cuboid to obtain a first CT image.

In one embodiment, extracting features of each second CT image to obtain a feature vector set includes:

processing each second CT image of the intervertebral disc by adopting a feature extractor to generate a feature vector group; the feature vector group comprises a plurality of feature vectors aiming at the second CT image; the number of column vectors of the feature vector group is the same as the number of contrast adjustment models.

In one embodiment, the decoding and reconstructing the first CT image and the feature vector group to output a CT reconstructed image includes: the soft tissue window CT image and the feature vector group are paired and then input into a variational self-encoder, and a CT reconstructed image is output; the soft tissue window CT image is an image obtained by adjusting the window width and the window position of the first CT image based on preset parameters; the CT reconstructed images include a sequence of images of each disc at the target contrast.

In a second aspect, the present application provides a CT image reconstruction apparatus, comprising:

the first CT image acquisition module is used for acquiring at least one first CT image;

the local adjusting module is used for processing the first CT image by adopting at least one contrast adjusting model to obtain at least one second CT image;

the characteristic extraction module is used for extracting the characteristics of each second CT image to obtain a characteristic vector group;

and the fusion reconstruction module is used for decoding and reconstructing the first CT image and the characteristic vector group and outputting a CT reconstructed image.

In a third aspect, the present application provides a computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the steps of the method when executing the computer program.

According to the CT image reconstruction method, the CT image reconstruction device and the computer equipment, at least one first CT image is obtained; processing the first CT image by using at least one contrast adjusting model to obtain at least one second CT image; extracting the characteristics of each second CT image to obtain a characteristic vector group; the first CT image and the feature vector group are decoded and reconstructed, the CT reconstructed images are output, the features of each second CT image generated by at least one contrast adjusting model can be fused, the adjusting time and the reading time of a doctor for a target bony structure in the initial CT image in the traditional scheme are reduced, the contrast of the CT image for the target bony structure is improved, and meanwhile, the efficiency of obtaining the optimal contrast image for the target bony structure is improved.

Drawings

FIG. 1 is a flowchart illustrating a method for reconstructing CT images according to an embodiment;

FIG. 2 is a flowchart illustrating the CT image reconstruction process according to one embodiment;

FIG. 3 is a flowchart illustrating a step of acquiring at least one first CT image according to one embodiment;

FIG. 4 is a flowchart illustrating the steps of performing a disc positioning process on a target CT image according to one embodiment;

FIG. 5 (a) is a schematic representation of a curve fitted to the spine in one embodiment;

FIG. 5 (b) is a schematic view of an embodiment for rotational alignment of an intervertebral disc;

FIG. 5 (c) is a schematic view of a region of interest tailored to an intervertebral disc in one embodiment;

FIG. 6 is a flowchart illustrating the steps of cropping a region of interest of each disc in one embodiment;

FIG. 7 is a flowchart illustrating an exemplary step of acquiring at least one first CT image;

FIG. 8 is an overall flow chart of CT image reconstruction in one example;

FIG. 9 is a block diagram illustrating an exemplary CT image reconstruction device;

fig. 10 is an internal structural diagram of a computer device in one embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

The degenerative diseases of the spine encompass a wide variety of diseases, of which, prolapse of the intervertebral disc is one of the most advanced degenerative diseases of the spine. The level diagnosis of the intervertebral disc protrusion is important reference information for the definite diagnosis and the selection of a treatment scheme of the spinal degenerative diseases. CT imaging has become an indispensable examination item in the process of diagnosing the herniated disk due to the advantages of rapid imaging, low examination cost and clear visualization of the pneumatosis of the intervertebral disk and calcification. Taking the diagnosis of the type of disc herniation in CT images as an example, the clinical workflow at the present stage is to observe 23 discs of each patient one by imaging physicians and record qualitative descriptions of disc categories in the imaging reports. The increasing examination volume of degenerative diseases of the spine of each medical institution year by year is continuously increasing the workload of radiologists, and doctors in a fatigue state may also miss diagnosis and misdiagnosis, and delay the treatment of patients. The CT image is often blurred in visualization of intervertebral disc degeneration, spinal cord deformation, dural sac compression, nerve root compression and other conditions, when a clinician diagnoses intervertebral disc bulging, herniation and herniation, the window width and the window position of the CT image are often required to be dynamically adjusted in real time, and depending on the adjustment experience and the adjustment efficiency of the clinician, the clinician needs to respectively observe the shapes and the positions of the intervertebral disc, the dural sac and the spinal cord through adjustment, and finally comprehensively judges to obtain category diagnosis. The application provides a CT image reconstruction method for improving the contrast of a CT image, manual adjustment by a doctor is not needed, meanwhile, the contrast adjustment efficiency of the CT image is improved, and the reconstruction effect of the CT image is improved.

In one embodiment, as shown in fig. 1, a CT image reconstruction method is provided, the method comprising:

step 110, acquiring at least one first CT image;

specifically, the first CT image may be a CT image of a target bony structure, such as a CT image of an intervertebral disc, so as to perform contrast adjustment for each target bony structure.

In some examples, the initial CT image may be segmented into a first CT image of a plurality of target bony structures by acquiring the initial CT image and performing an image segmentation process on the initial CT image; the initial CT image may be a spine CT image and the first CT image may be a CT image for an intervertebral disc.

Step 120, processing the first CT image by using at least one contrast adjustment model to obtain at least one second CT image;

in particular, the contrast adjustment model may be an optimal contrast model for a local target, e.g. a contrast adjustment model for a single target tissue in the first CT image; by processing the first CT image using the at least one contrast adjustment model, contrast adjustments may be made for a plurality of target tissues in the first CT image, respectively; the second CT image is an image obtained by processing the first CT image through at least one contrast adjusting model, and then a plurality of local optimal contrast images aiming at the target bony structure can be obtained. The CT images of all target bony structures are adjusted through at least one contrast model, so that a plurality of second CT images for adjusting the contrast of different target tissues can be obtained, the adjusting time and the reading time of a doctor for the target bony structures in the initial CT images in the traditional scheme are reduced, the optimal developed images for the target bony structures are generated conveniently, and the efficiency of obtaining the optimal contrast images for the target bony structures is improved.

In some examples, the first CT image may be a CT image for an intervertebral disc, and the plurality of second CT images may include one or more of an intervertebral disc optimal contrast image, a dural sac optimal contrast image, and a spinal cord optimal contrast image.

Step 130, extracting the features of each second CT image to obtain a feature vector group;

specifically, feature extraction may be performed on each second CT image to obtain a feature vector group including a feature vector corresponding to each second CT image. By respectively extracting the features of the second CT images obtained after the processing of the at least one contrast adjustment model, the features of the second CT images subjected to contrast adjustment aiming at different target tissues are conveniently integrated, and the contrast of the finally generated CT images aiming at the target bony structure is favorably improved.

In some examples, a feature extraction algorithm or a feature extraction model may be used to perform feature extraction on each second CT image, so as to obtain a feature vector for each second CT image, and the obtained feature vectors are combined into the feature vector group.

And 140, decoding and reconstructing the first CT image and the feature vector group, and outputting a CT reconstructed image.

Specifically, the first CT image may be decoded and reconstructed based on the feature vector group to obtain a CT reconstructed image, and since the feature vector group extracts features of each second CT image, each second CT image is obtained by processing with at least one contrast adjustment model, so that the CT reconstructed image is fused with the contrast adjustment features of each second CT image, and the contrast of the CT image with respect to the target bony structure is improved.

In some examples, after the contrast of the first CT image is adjusted by using a conventional soft tissue window, a depth generation model for decoding reconstruction, such as a variational self-encoder, may be input together with the feature vector group to obtain a CT reconstructed image; the CT reconstructed image may be a reconstructed image for an intervertebral disc.

According to the embodiment of the application, a plurality of first CT images are obtained; processing the first CT image by using at least one contrast adjusting model to obtain at least one second CT image; extracting the characteristics of each second CT image to obtain a characteristic vector group; the first CT image and the feature vector group are decoded and reconstructed, the CT reconstructed images are output, the features of each second CT image generated by at least one contrast adjusting model can be fused, the adjusting time and the reading time of a doctor for a target bony structure in the initial CT image in the traditional scheme are reduced, the contrast of the CT image for the target bony structure is improved, and meanwhile, the efficiency of obtaining the optimal contrast image for the target bony structure is improved.

In one embodiment, as shown in fig. 2, the method further comprises:

step 210, obtaining a plurality of original CT images and a plurality of sets of target contrast images obtained by adjusting the original CT images;

step 220, inputting the original CT image and the target contrast image into a generation confrontation network for training, and obtaining at least one contrast adjustment model.

Specifically, a plurality of original CT images can be subjected to contrast adjustment for different targets, so as to obtain a plurality of sets of target contrast images obtained through adjustment; the original CT image and the target contrast image can be input into a generation countermeasure network for training, wherein the generation countermeasure network can comprise a generator and a discriminator; the generator may include an encoder, which may include several convolutional layers; the discriminator may comprise a decoder, which may comprise several deconvolution layers. For example, the original CT image and each set of target contrast images are input into a countermeasure network for training, and at least one contrast adjustment model is obtained. The plurality of contrast adjustment models may be models for which contrast adjustment is performed for different target tissues, respectively. By acquiring a target contrast image for training and inputting the target contrast image and an original CT image together to generate an confrontation network for training, at least one contrast adjusting model for improving the contrast adjusting efficiency can be obtained, and the adjusted target contrast image training model is based on the acquired target contrast image, so that the optimal developing CT image of the target bony structure can be generated accurately, and the contrast of the CT image for the target bony structure can be improved.

In some examples, the constructed generation countermeasure network may include one or more of a conditional generation countermeasure network CGAN, an image translation Pix2PixGAN, and a cyclic generation countermeasure network CycleGAN. The discriminator can select a convolutional neural network frame; generating a Loss function (Loss function) against the network may include a regularization function.

Specifically, the plurality of contrast adjustment models may be images obtained by performing contrast adjustment for different target tissues, for example, the plurality of contrast adjustment models may be a contrast adjustment model adjusted for a first tissue, a contrast adjustment model adjusted for a second tissue, and a contrast adjustment model adjusted for a third tissue, or may be a contrast adjustment model adjusted for more or less target tissues. For example, the plurality of contrast adjustment models may be a first contrast model for intervertebral disc contrast adjustment, a second contrast model for epidural sac contrast adjustment, and a third contrast model for spinal cord contrast adjustment, respectively; by acquiring a target contrast image for a first tissue, a target contrast image for a second tissue and a needleA target contrast image for a third tissue, wherein the target contrast image for the first tissue may be an intervertebral disc target contrast image, the target contrast image for the second tissue may be a dural sac target contrast image, and the target contrast image for the third tissue may be a spinal cord target contrast image, e.g., an intervertebral disc optimal display image I ₁ Optimal display image I of the dural sac ₂ And the best display image I of the spinal cord ₃ Constructing an intervertebral disc contrast adjustment data set; the target contrast image for the first tissue may be an optimal contrast image of the intervertebral disc obtained based on empirical value adjustment, the first contrast model may be obtained by training an anti-collision network generated by inputting an original CT image and the target contrast image for the first tissue, and the first contrast model may be the optimal contrast model of the intervertebral disc; the target contrast image for the second tissue may be an optimal contrast image of the dural sac adjusted based on empirical values, the second contrast model may be obtained by training an anti-contrast network generated by inputting the original CT image and the target contrast image for the second tissue, and the second contrast model may be an optimal contrast model of the dural sac; the target contrast image for the third tissue may be an optimal spinal contrast image obtained based on empirical value adjustment, the third contrast model may be obtained by inputting the original CT image and the target contrast image for the third tissue to generate an anti-collision network for training, and the third contrast model may be an optimal spinal contrast model. Further, according to step 120, the first CT image I is processed with each generated confrontation model (the first contrast model G1, the second contrast model G2 and the third contrast model G3) of the pre-training, and the generated plurality of second CT images for the intervertebral disc may include the intervertebral disc optimal contrast image I ₁ ', optimal contrast image of dural sac I ₂ ' and spinal cord optimal contrast image I ₃ One or more of. By generating the first contrast model, the second contrast model and the third contrast model, the first CT image can be directly processed, and the clinician is prevented from performing real-time dynamic window width adjustment and window level adjustment on the target CT image to observe respectivelyThe shapes and the positions of the intervertebral disc, the dural sac and the spinal cord reduce the adjusting time and the reading time of a doctor in the traditional scheme, and improve the efficiency of obtaining the intervertebral disc image with the optimal contrast.

In some examples, the discriminator In the generation of the countermeasure network may select a convolutional neural network framework for distinguishing the target contrast image In from the second CT image I' generated from the first CT image I after the generation of the countermeasure network. Regularization constraints may be added to the loss function that generates the counterpoise network to bring the second CT image I' as close as possible to the target contrast image In. The regularization function may be selected to be an L1 function or an L2 function. The target contrast image for the first tissue may be obtained by adjusting the optimal developing state of the intervertebral disc for all the grouped original CT images by an image doctor who passes through more than middle-aged resources (e.g., 7-10 years of clinical examination experience), and recording the window width WW1 and the window level WL1 in the optimal developing state; the target contrast image of the second tissue can be obtained by adjusting the optimal development state of the hard capsule of all the original CT images which are put into the group respectively by an image doctor of more than middle-aged resources, and recording the window width WW2 and the window level WL2 in the optimal development state; the target contrast image for the third tissue may be obtained by adjusting the optimal development state of the spinal cord for all the original CT images entered into the group by an imaging doctor of more than middle-aged resources, and recording the window width WW3 and the window level WL3 in the optimal development state.

In one embodiment, as shown in fig. 3, acquiring at least one first CT image includes:

step 310, acquiring a target CT image; the target CT image includes at least one intervertebral disc;

step 320, intervertebral disc positioning processing is carried out on the target CT image to obtain the central point of each intervertebral disc;

step 330, sequentially rotationally aligning the target CT images based on the central points of the intervertebral discs and a preset rotational alignment rule to obtain the interested areas of the intervertebral discs;

step 340, cutting the region of interest of each intervertebral disc to obtain at least one first CT image for the intervertebral disc.

Specifically, a target CT image may be acquired, which may be an initial CT image for the spine; intervertebral disc positioning processing can be carried out on each intervertebral disc in the target CT image, for example, an image processing method is adopted to carry out intervertebral disc positioning marking on the target CT image, and the central point of each intervertebral disc is obtained; the center point of the intervertebral disc may be a three-dimensional coordinate of the center of gravity of the intervertebral disc; the preset rotation and correction rule can be realized by adopting a computer vision algorithm, and the target CT image can be rotationally corrected based on the central point of the intervertebral disc, for example, the central point of the intervertebral disc is taken as a rotation center, so that the region of interest of the intervertebral disc is obtained, and the influence of the intervertebral disc on different postures in a three-dimensional space is reduced. In the target CT image, the area of each intervertebral disc occupies a smaller proportion of the CT image, the interested area of each intervertebral disc is cut by accurately positioning the related area of the intervertebral disc in the target CT image so as to completely cut the related area of the intervertebral disc, and the obtained first CT images are convenient for pertinently adjusting the contrast of the intervertebral disc and related tissues. Through the preprocessing process, the target CT image can be divided into a plurality of first CT images aiming at the intervertebral discs, so that the contrast of each intervertebral disc can be adjusted conveniently and subsequently.

In some examples, each of the first CT images obtained by the cropping includes a complete image region of the corresponding intervertebral disc.

In one embodiment, as shown in fig. 4, the intervertebral disc positioning processing is performed on the target CT image to obtain the central point of each intervertebral disc, and the method includes:

step 410, performing multi-target image segmentation on the target CT image to obtain a vertebral body segmentation result; the vertebral body segmentation result comprises CT images of a plurality of vertebral bodies and a vertebral body label corresponding to each vertebral body;

step 420, fitting a minimum external cuboid aiming at the vertebral body to the vertebral body segmentation result to obtain a central point of the vertebral body; the central point of the vertebral body is the gravity center of the minimum external cuboid;

430, obtaining a spine fitting curve based on the central points of the vertebral bodies;

and step 440, obtaining the central point of each intervertebral disc according to the spine fitting curve and the central point of each vertebral body.

In particular, multi-target image segmentation may label each pixel in an image as one of several predetermined object classes, while performing recognition and segmentation of multiple targets; performing multi-target image segmentation on the target CT image to obtain a vertebral body segmentation result; the vertebral body segmentation result comprises CT images of a plurality of vertebral bodies, and multi-label image classification is carried out on the target CT image, so that the vertebral body segmentation result comprises a vertebral body label corresponding to each vertebral body; as shown in fig. 5 (a), the minimum external cuboid for the vertebral body can be fitted to the CT images of each vertebral body in the vertebral body segmentation result, and then the center point of the vertebral body can be determined according to the center of gravity of the minimum external cuboid; fitting the central points of the vertebral bodies to obtain a spine fitting curve, for example, the central points of the vertebral bodies can form a point set, and the point set can be fitted by adopting an interpolation method to obtain a smooth fitting curve C for the spine; the central points of a plurality of intervertebral discs can be determined according to the spine fitting curve and the central points of the vertebral bodies, for example, the central points between the central points of the adjacent vertebral bodies can be selected; if the midpoint is located on the spine fitting curve, the midpoint can be determined as the central point of the intervertebral disc; if the midpoint is located outside the spine fit curve, the point on the spine fit curve with the smallest distance from the midpoint can be determined as the center point of the intervertebral disc. Through the steps, the central point of the intervertebral disc can be quickly and accurately determined, and the efficiency and the accuracy of intervertebral disc positioning are improved.

In some examples, a trained deep learning segmentation model may be used to perform image segmentation on the target CT image to obtain a vertebral body segmentation result, where the vertebral body segmentation result may represent different vertebral bodies by different label values. The cone segmentation result can be further optimized through a Graph Cut algorithm, a deep learning algorithm and the like, so that the cone segmentation is prevented from being inaccurate due to the fact that the cone segmentation result is under-segmented or over-segmented. The three-dimensional coordinates of the central point of the cone can be obtained by adopting a spine detection and positioning marking method, for example, a deep learning target detection model or deep learning point detection can be adoptedMeasuring a model to obtain the central point of each vertebral body; the interpolation method may include at least one of B-spline interpolation and polynomial interpolation; the central points of 23 vertebral bodies (C2-C7, T1-T12 and L1-L5) from the cervical 2 vertebral body to the lumbar 5 vertebral body can be combined into a point set P [ P ] ₁ ,P ₂ ,…,P ₂₃ ]And fitting the point set P by using an interpolation method to form the smooth fitting curve C. For the center point determined as the intervertebral disc, P in the point set P can be taken sequentially _n And P _n+1 (n =1 to 22) midpoint P _C ', if the midpoint P _C ' just on the fitted curve C, the midpoint P will be _C ' As intervertebral disc center point P _o-min (ii) a Otherwise, selecting P in the fitting curve C _Cn To P _Cn+1 On the segment with P _C ' the point where the absolute distance is smallest is taken as the disc center point P _o-min 。

In some examples, the target CT image may be rotationally rectified based on the fitted curve C, as shown in fig. 5 (b). Can obtain a fitting curve C at the central point P of the intervertebral disc _o-min If the included angle between the tangent vector V 'and the direction vector V of the fitting curve C is theta, the original CT image takes the orthogonal vector obtained by the tangent vector V' and the direction vector V as a rotating shaft and the central point P of the intervertebral disc _o-min And performing rotation correction on the rotation center, for example, rotating by an angle theta, and further obtaining the region of interest of the intervertebral disc.

In one embodiment, as shown in fig. 6, cropping the region of interest of each intervertebral disc to obtain each first CT image comprises:

step 610, acquiring the length of each endplate adjacent to the intervertebral disc;

step 620, generating a cutting cuboid based on the maximum value of the lengths of the end plates;

and 630, cutting the region of interest of the intervertebral disc by using the cutting cuboid to obtain a first CT image.

Specifically speaking, the endplate adjacent with the intervertebral disc constitutes the upper and lower bony border of intervertebral disc, including the lower edge endplate of the vertebra body on the intervertebral disc and the upper edge endplate of the vertebra body under the intervertebral disc, can select the maximum value in each endplate length adjacent with the intervertebral disc as the benchmark, generate and cut the cuboid, for example, as shown in fig. 5 (c), can all set up the length and the width of cutting the cuboid to the twice maximum value, the height of cutting the cuboid sets up to the maximum value, and then obtain and cut the cuboid, adopt and cut the cuboid and cut the region of interest of intervertebral disc, can obtain first CT image. The length, width and height of the cut cuboid can also be set according to actual requirements so as to contain a complete intervertebral disc image area. Through the steps, the region of interest can contain similar intervertebral disc information, and the influence of other image information is reduced.

In some examples, the length of the inferior endplate of the superior vertebral body of the intervertebral disc, and the length of the superior endplate of the inferior vertebral body of the intervertebral disc, may be determined by computer vision or other related image algorithms based on CT images of the intervertebral disc in the median sagittal position.

Specifically, a feature extractor may be adopted to process each second CT image of the intervertebral disc, and a feature vector group may be generated, and each feature vector in the feature vector group may correspond to each second CT image one to one; the number of column vectors of the feature vector group may be the number of feature vectors in the feature vector group, and is the same as the number of contrast adjustment models. In the above steps, the characteristics of contrast adjustment are accurately extracted by improving the comprehensiveness of the characteristic vector group to the characteristic description of each second CT image, so that the contrast of the final development of the intervertebral disc CT image is improved, and the reconstruction effect of the CT image is improved.

In some examples, the depth feature extractor may be employed to separately process the intervertebral disc best contrast image I ₁ ', optimal contrast image of the dural sac I ₂ ' and optimal spinal contrast image I ₃ ' 3 XN dimensional features are abstractedA set of vectors F, which may comprise 3 feature vectors, e.g. feature vectors F of dimension 1 XN ₁ 1 XN-dimensional feature vector F ₂ And a feature vector F of 1 XN dimensions ₃ . The depth feature extractor comprises a plurality of convolution layers and a full connection layer which are connected in series.

Specifically, the soft tissue window CT image Is may be an image obtained by adjusting the window width and the window position of the first CT image I of the intervertebral disc; the soft tissue window CT image Is may be paired with the feature vector set F and then input to a Variational Auto Encoder (VAE), for example, the soft tissue window CT image Is and the feature vector F in the feature vector set F ₁ Feature vector F ₂ And a feature vector F ₃ The reconstructed images Io of the intervertebral discs are obtained as input from the encoder, and may include image sequences of each intervertebral disc of the spine at a target contrast (e.g., an optimal contrast), for example, the reconstructed images Io of the intervertebral discs may be displayed in a reading system of a imaging physician in a sub-view form by taking an object to be measured as a unit, including a sequence of each intervertebral disc Ion (n = 1-23). The above steps can fuse each feature vector, improve the contrast adjustment effect on the soft tissue window CT image, and generate the reconstructed image of the intervertebral disc with better developing effect on the intervertebral disc, the dural sac and the spinal cord.

In some examples, adjusting the preset parameters of the window width and the window level of the first CT image may include: the window width WW of the soft tissue window is 300-500 HU, and the window level WL of the soft tissue window is 40-60 HU. The variational self-encoder may include an encoder and a decoder; wherein random noise (e.g., gaussian noise, etc.) may be added to the encoder to enhance the robustness of the generated image, a KL function may be chosen as the loss function of the variational self-encoder. The reconstructed image Io of the intervertebral disc may be displayed in a multi-planar reconstruction (MPR) manner, which facilitates viewing the axial position, the coronal position, and the sagittal position of the reconstructed CT image of the cervical spine at the same time.

In some examples, as shown in fig. 7, to obtain a flow chart of the step of obtaining at least one first CT image, a multi-objective segmentation may be performed on a vertebral body of a spine to obtain a vertebral body segmentation result; optimizing the vertebral body segmentation result, for example, adopting Graph Cut algorithm or deep learning algorithm, etc., avoiding inaccurate cone segmentation caused by under-segmentation or over-segmentation of the vertebral body segmentation result, and obtaining a vertebral body multi-target segmentation result; the minimum external cuboid aiming at the vertebral body can be fitted to the vertebral body segmentation result to obtain the central point coordinate of the vertebral body; the intervertebral disc can be positioned based on the centrum central point coordinates and the centrum multi-target segmentation result, and the central point of the intervertebral disc can be obtained; performing rotational correction on the intervertebral discs in the CT image, for example, the central point of the intervertebral disc may be used as a rotational center to perform rotational correction on each intervertebral disc in the CT image; and (4) cutting the intervertebral discs in the CT images to obtain the CT images of the intervertebral discs. By adopting the flow to preprocess the target CT image, the target CT image can be accurately divided into the CT images of a plurality of intervertebral discs, so that the contrast of each intervertebral disc can be adjusted conveniently and subsequently.

In some examples, as shown in fig. 8, an overall flowchart of CT image reconstruction is shown. The intervertebral disc CT image I can be obtained by sequentially positioning, rotating and cutting the intervertebral disc in the CT image. The soft tissue window of the intervertebral disc CT image I can be adjusted to obtain an intervertebral disc soft tissue window CT image Is; can obtain the best intervertebral disc display image I obtained by adjusting the CT image of the intervertebral disc by an imaging doctor ₁ Optimal display image of the dural sac I ₂ And the best display image I of the spinal cord ₃ (ii) a Optimal display image I based on intervertebral disc ₁ Optimal display image of the dural sac I ₂ And the best display image I of the spinal cord ₃ A single optimal contrast generation can be constructedThe model comprises a generated intervertebral disc optimal visualization model, a generated dura mater sac optimal visualization model and a generated spinal cord optimal visualization model; respectively inputting the intervertebral disc CT image I into each single optimal contrast generation model, and obtaining corresponding single optimal contrast images including the intervertebral disc optimal contrast image I ₁ ', optimal contrast image of the dural sac I ₂ ' and spinal cord optimal contrast image I ₃ '; fusing the single optimal contrast image, and inputting the fused optimal contrast image and the CT image Is of the intervertebral disc soft tissue window into the constructed VAE generation model together to obtain images before and after the contrast adjustment of the intervertebral disc CT image; the intervertebral disc CT image with the adjusted contrast is used for image display, so that the intervertebral disc display contrast can be improved. In practical application, the whole flow chart only needs to train the model once, the trained model can be directly used, the contrast adjustment efficiency can be improved, and the CT image reconstruction efficiency is improved.

It should be understood that, although the steps in the flowcharts related to the embodiments as described above are sequentially displayed as indicated by arrows, the steps are not necessarily performed sequentially as indicated by the arrows. The steps are not limited to being performed in the exact order illustrated and, unless explicitly stated herein, may be performed in other orders. Moreover, at least a part of the steps in the flowcharts related to the embodiments described above may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the execution order of the steps or stages is not necessarily sequential, but may be rotated or alternated with other steps or at least a part of the steps or stages in other steps.

Based on the same inventive concept, the embodiment of the present application further provides a CT image reconstruction apparatus for implementing the above CT image reconstruction method. The implementation scheme for solving the problem provided by the apparatus is similar to the implementation scheme described in the above method, so the specific limitations in one or more embodiments of the CT image reconstruction apparatus provided below can be referred to the limitations on the CT image reconstruction method in the above description, and are not described again here.

In one embodiment, a CT image reconstruction apparatus is provided, as shown in fig. 9, the apparatus comprising:

a first CT image obtaining module 910, configured to obtain at least one first CT image;

a local adjustment module 920, configured to process the first CT image using at least one contrast adjustment model to obtain at least one second CT image;

a feature extraction module 930, configured to extract features of each second CT image to obtain a feature vector group;

and a fusion reconstruction module 940 for decoding and reconstructing the first CT image and the feature vector group and outputting a CT reconstructed image.

In one embodiment, the apparatus further comprises:

the image acquisition module is used for acquiring a plurality of original CT images and a plurality of groups of target contrast images which are obtained by aiming at the original CT images;

and the model training module is used for inputting the original CT image and the target contrast image into a countermeasure network for training to obtain at least one contrast adjusting model.

In one embodiment, the model training module is further configured to input the original CT image and each set of target contrast images into a countermeasure network for training, so as to obtain at least one contrast adjustment model. In one embodiment, the first CT image acquiring module 910 includes:

the image acquisition unit is used for acquiring a target CT image; the target CT image includes at least one intervertebral disc;

the positioning processing unit is used for carrying out intervertebral disc positioning processing on the target CT image to obtain the central point of each intervertebral disc;

the rotary rectification unit is used for sequentially carrying out rotary rectification on the target CT images based on the central point of each intervertebral disc and a preset rotary rectification rule to obtain an interested area of each intervertebral disc;

and the target cutting unit is used for cutting the interested area of each intervertebral disc to obtain at least one first CT image aiming at the intervertebral disc.

In one embodiment, the positioning processing unit includes:

the cone segmentation subunit is used for carrying out multi-target image segmentation on the target CT image to obtain a cone segmentation result; the vertebral body segmentation result comprises CT images of a plurality of vertebral bodies and a vertebral body label corresponding to each vertebral body;

the first center determining subunit is used for fitting a minimum external cuboid aiming at the vertebral body to the vertebral body segmentation result to obtain a central point of the vertebral body; the central point of the vertebral body is the gravity center of the minimum external cuboid;

the curve fitting subunit is used for obtaining a spine fitting curve based on the central points of the vertebral bodies;

and the second center determining subunit is used for obtaining the center point of each intervertebral disc according to the spine fitting curve and the center point of each vertebral body.

In one embodiment, the object clipping unit includes:

the length acquisition subunit is used for acquiring the length of each endplate adjacent to the intervertebral disc;

a cutting cuboid generating subunit, configured to generate a cutting cuboid based on a maximum value of the lengths of the end plates;

and the target cutting subunit is used for cutting the region of interest of the intervertebral disc by adopting a cutting cuboid to obtain a first CT image.

In one embodiment, the feature extraction module 930 is further configured to process each second CT image of the intervertebral disc with the feature extractor to generate a feature vector set; the feature vector group comprises a plurality of feature vectors aiming at the second CT image; the number of column vectors of the feature vector group is the same as the number of contrast adjustment models.

In one embodiment, the fusion reconstruction module 940 is further configured to: the soft tissue window CT image and the feature vector group are paired and then input into a variational self-encoder, and a CT reconstructed image is output; the soft tissue window CT image is an image obtained by adjusting the window width and the window position of the first CT image based on preset parameters; the CT reconstructed images include a sequence of images of each disc at a target contrast.

All or part of the modules in the CT image reconstruction apparatus can be implemented by software, hardware and a combination thereof. The modules can be embedded in a hardware form or independent of a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In an embodiment, a computer device is provided, comprising a memory storing a computer program and a processor implementing the steps of the method described above when the processor executes the computer program.

In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 10. The computer apparatus includes a processor, a memory, an input/output interface, a communication interface, a display unit, and an input device. The processor, the memory and the input/output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input/output interface. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program to run on the non-volatile storage medium. The input/output interface of the computer device is used for exchanging information between the processor and an external device. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless communication can be realized through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. The computer program is executed by a processor to implement a CT image reconstruction method. The display unit of the computer device is used for forming a visual picture and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.

Those skilled in the art will appreciate that the architecture shown in fig. 10 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.

In an embodiment, a computer-readable storage medium is provided, on which a computer program is stored, which computer program, when being executed by a processor, carries out the steps of the above-mentioned method.

In an embodiment, a computer program product is provided, comprising a computer program which, when executed by a processor, performs the steps of the method described above.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, database, or other medium used in the embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, high-density embedded nonvolatile Memory, resistive Random Access Memory (ReRAM), magnetic Random Access Memory (MRAM), ferroelectric Random Access Memory (FRAM), phase Change Memory (PCM), graphene Memory, and the like. Volatile Memory can include Random Access Memory (RAM), external cache Memory, and the like. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM), for example. The databases referred to in various embodiments provided herein may include at least one of relational and non-relational databases. The non-relational database may include, but is not limited to, a block chain based distributed database, and the like. The processors referred to in the embodiments provided herein may be general purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, quantum computing based data processing logic devices, etc., without limitation.

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present application. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, and these are all within the scope of protection of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.

Claims

1. A method for reconstructing CT images, the method comprising:

acquiring at least one first CT image;

processing the first CT image by adopting at least one contrast adjusting model to obtain at least one second CT image;

extracting the characteristics of each second CT image to obtain a characteristic vector group;

2. The method of claim 1, further comprising:

and inputting the original CT image and the target contrast image into a countermeasure network for training to obtain at least one contrast adjusting model.

3. The method of claim 2, wherein each set of the target contrast images includes at least one contrast-adjusted image; inputting the original CT image and the target contrast image into a confrontation network for training to obtain at least one contrast adjustment model, wherein the confrontation network comprises:

and inputting the original CT image and each group of target contrast images into the generated countermeasure network for training to obtain at least one contrast adjustment model.

4. The method of claim 1, wherein said acquiring at least one first CT image comprises:

acquiring a target CT image; the target CT image includes at least one intervertebral disc;

carrying out intervertebral disc positioning treatment on the target CT image to obtain the central point of each intervertebral disc;

5. The method of claim 4, wherein the subjecting the target CT image to a disc positioning process to obtain a center point of each of the discs comprises:

6. The method of claim 4, wherein said cropping the region of interest of each of the intervertebral discs to obtain each of the first CT images comprises:

obtaining the length of each endplate adjacent to the intervertebral disc;

generating a cutting cuboid based on the maximum value of the length of each end plate;

and cutting the region of interest of the intervertebral disc by using the cutting cuboid to obtain the first CT image.

7. The method of claim 1, wherein the extracting features of each of the second CT images to obtain a set of feature vectors comprises:

processing each second CT image of the intervertebral disc by adopting a feature extractor to generate the feature vector group; the set of feature vectors comprises a plurality of feature vectors for the second CT image; the number of column vectors of the feature vector group is the same as the number of the contrast adjustment models.

8. The method according to claim 7, wherein said decoding and reconstructing the first CT image and the feature vector set and outputting the CT reconstructed image comprises:

the soft tissue window CT image and the feature vector group are paired and then input into a variational self-encoder, and the CT reconstructed image is output; the soft tissue window CT image is an image obtained by adjusting the window width and the window position of the first CT image based on preset parameters; the CT reconstructed images include a sequence of images of each disc at a target contrast.

9. A CT image reconstruction apparatus, comprising:

the feature extraction module is used for extracting features of the second CT images to obtain a feature vector group;

and the fusion reconstruction module is used for decoding and reconstructing the first CT image and the feature vector group and outputting a CT reconstructed image.

10. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor, when executing the computer program, implements the steps of the method of any of claims 1 to 8.