Artificial intelligence three-dimensional modeling system and method based on CT big data

Abstract

The invention provides an artificial intelligence three-dimensional modeling system and method based on CT big data, and relates to the field of data processing, and the method comprises the steps: building a plurality of three-dimensional reconstruction models based on CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of a plurality of test samples; determining a plurality of reference test samples from the plurality of test samples based on the organization structure information of the to-be-reconstructed target and the organization structure information of the plurality of test samples; based on the optimal CT scanning parameters of the plurality of reference test samples, determining the optimal CT scanning parameters of the to-be-reconstructed target; based on the optimal CT scanning parameters of the to-be-reconstructed target, performing CT scanning on the to-be-reconstructed target to obtain a CT image of the to-be-reconstructed target; the three-dimensional CT image of the to-be-reconstructed target is generated through the plurality of three-dimensional reconstruction models based on the CT image of the to-be-reconstructed target, and the method has the advantage of improving the three-dimensional reconstruction quality of the CT image.

Description

Technical Field

[0001] The present invention relates to the field of medical image processing, and in particular to an artificial intelligence three-dimensional modeling system and method based on CT big data. Background Art

[0002] 3D reconstruction refers to the conversion of 2D images or projection data into the spatial information of 3D objects. The reconstructed model is convenient for computer display and further processing. It has a wide range of applications in medicine, biology, engineering, computer vision and other fields, and is of great significance for studying and analyzing the structure and morphology of objects. With the continuous advancement of imaging technology, high-resolution 3D reconstruction has become an important research direction. In recent years, artificial intelligence technology has developed rapidly, and the research on key technologies of intelligent medical auxiliary diagnosis has a macroscopic significance in modern medical clinics. At present, in the research on 3D reconstruction technology of CT (Computed Tomography) images, due to the objective fact that medical images have less texture and more noise, there are certain difficulties in the research on 3D reconstruction technology of CT images, which brings complexity to the 3D reconstruction of the model and is not conducive to the development of clinical medical auxiliary diagnosis technology.

[0003] Therefore, there is a need to provide an artificial intelligence three-dimensional modeling system and method based on CT big data to improve the quality of three-dimensional reconstruction of CT images. Summary of the invention

[0004] The present invention provides an artificial intelligence three-dimensional modeling method based on CT big data, comprising: obtaining tissue structure information, optimal CT scanning parameters, CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of multiple test samples; establishing multiple three-dimensional reconstruction models based on the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the multiple test samples; obtaining tissue structure information of a target to be reconstructed; determining multiple reference test samples from the multiple test samples based on the tissue structure information of the target to be reconstructed and the tissue structure information of the multiple test samples; determining the optimal CT scanning parameters of the target to be reconstructed based on the optimal CT scanning parameters of the multiple reference test samples; performing a CT scan on the target to be reconstructed based on the optimal CT scanning parameters of the target to be reconstructed to obtain a CT image of the target to be reconstructed; and generating a three-dimensional CT image of the target to be reconstructed based on the CT image of the target to be reconstructed using the multiple three-dimensional reconstruction models.

[0005] Furthermore, obtaining the optimal CT scanning parameters of the test sample includes: generating multiple groups of test CT scanning parameters, wherein the test CT scanning parameters include at least multiple CT scanning angles; obtaining CT images corresponding to each group of the test CT scanning parameters; for each group of the test CT scanning parameters, generating CT reconstruction results corresponding to the test CT scanning parameters according to the CT images corresponding to the test CT scanning parameters through a three-dimensional reconstruction algorithm; based on the CT reconstruction results corresponding to each group of the test CT scanning parameters, screening the multiple groups of test CT scanning parameters to determine the optimal CT scanning parameters of the test sample.

[0006] Further, based on the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the multiple test samples, multiple three-dimensional reconstruction models are established, including: based on the tissue structure information of the multiple test samples, the multiple test samples are clustered by a clustering algorithm to determine multiple first sample clusters; for each of the first sample clusters, based on the optimal CT scanning parameters of the multiple test samples included in the first sample cluster, the multiple test samples included in the first sample cluster are clustered by a clustering algorithm to determine multiple second sample clusters included in the first sample cluster; for each of the second sample clusters, based on the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the multiple test samples included in the second sample cluster, the multiple test samples included in the second sample cluster are clustered by a clustering algorithm to determine multiple third sample clusters included in the second sample cluster; for each of the third sample clusters, multiple three-dimensional reconstruction models corresponding to the third sample cluster are established, and the multiple three-dimensional reconstruction models corresponding to the third sample cluster are jointly trained by the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the multiple test samples included in the third sample cluster, wherein the structures of any two three-dimensional reconstruction models corresponding to the third sample cluster are different.

[0007] Furthermore, the tissue structure information of the target to be reconstructed is obtained, including: S11, obtaining ultrasonic detection data of multiple positions of the target to be reconstructed; S12, calculating the ultrasonic fluctuation value of the target to be reconstructed based on the ultrasonic detection data of multiple positions of the target to be reconstructed in each round; S13, judging whether the detection condition is met according to the ultrasonic fluctuation value of the target to be reconstructed, if not, executing S14, if yes, executing S15; S14, performing the next round of ultrasonic detection to obtain the ultrasonic detection data of multiple positions of the target to be reconstructed, wherein the multiple positions of any two rounds are different, and S12 is executed; S15, generating the tissue structure information of the target to be reconstructed, wherein the tissue structure information of the target to be reconstructed at least includes the ultrasonic detection data of multiple positions of the target to be reconstructed in each round.

[0008] Furthermore, based on the tissue structure information of the target to be reconstructed and the tissue structure information of the multiple test samples, multiple reference test samples are determined from the multiple test samples, including: based on the tissue structure information of the target to be reconstructed and the tissue structure information of the test sample corresponding to the cluster center of each first sample cluster, a reference first sample cluster is determined from the multiple first sample clusters; based on the tissue structure information of the target to be reconstructed and the tissue structure information of each test sample included in the reference first sample cluster, multiple reference test samples are determined from the multiple test samples included in the reference first sample cluster.

[0009] Furthermore, based on the optimal CT scanning parameters of the multiple reference test samples, the optimal CT scanning parameters of the target to be reconstructed are determined, including: determining the optimal CT scanning parameters of the target to be reconstructed based on the optimal CT scanning parameters of the multiple reference test samples through a parameter determination model.

[0010] Furthermore, based on the optimal CT scanning parameters of the target to be reconstructed, a CT scan is performed on the target to be reconstructed, including: based on the optimal CT scanning parameters of the target to be reconstructed, a CT scan is performed on the target to be reconstructed; during the CT scanning, scanning status information of the target to be reconstructed is acquired; based on the scanning status information of the target to be reconstructed, it is determined whether to perform a secondary CT scan, and if so, a target CT scanning angle is determined, and a secondary CT scan is performed according to the target CT scanning angle.

[0011] Furthermore, obtaining the scanning status information of the target to be reconstructed includes: setting a plurality of markers on the target to be reconstructed according to the optimal CT scanning parameters and tissue structure information of the target to be reconstructed; acquiring a scanning status image of the target to be reconstructed; and determining the scanning status information of the target to be reconstructed according to the scanning status image of the target to be reconstructed.

[0012] Furthermore, generating a three-dimensional CT image of the target to be reconstructed based on the CT image of the target to be reconstructed by using the multiple three-dimensional reconstruction models includes: determining a target second sample cluster from the multiple second sample clusters based on the optimal CT scanning parameters of the target to be reconstructed; determining a target third sample cluster from the multiple third sample clusters included in the target second sample cluster based on the CT image of the target to be reconstructed; and generating a three-dimensional CT image of the target to be reconstructed based on the CT image of the target to be reconstructed by using multiple three-dimensional reconstruction models corresponding to the target third sample cluster.

[0013] The present invention provides an artificial intelligence three-dimensional modeling system based on CT big data, which is used to execute the above-mentioned artificial intelligence three-dimensional modeling method based on CT big data, including: a data acquisition module, which is used to acquire tissue structure information, optimal CT scanning parameters, CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of multiple test samples; a model establishment module, which is used to establish multiple three-dimensional reconstruction models based on the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the multiple test samples; an information acquisition module, which is used to acquire tissue structure information of a target to be reconstructed; a sample determination module, which is used to determine multiple reference test samples from the multiple test samples based on the tissue structure information of the target to be reconstructed and the tissue structure information of the multiple test samples; a parameter determination module, which is used to determine the optimal CT scanning parameters of the target to be reconstructed based on the optimal CT scanning parameters of the multiple reference test samples; a CT scanning module, which is used to perform CT scanning on the target to be reconstructed based on the optimal CT scanning parameters of the target to be reconstructed to obtain the CT image of the target to be reconstructed; and a three-dimensional reconstruction module, which is used to generate the three-dimensional CT image of the target to be reconstructed based on the CT image of the target to be reconstructed through the multiple three-dimensional reconstruction models.

[0014] Compared with the prior art, the artificial intelligence 3D modeling system and method based on CT big data provided in this specification has at least the following beneficial effects:

[0015] 1. By determining the optimal CT scanning parameters for each test sample, it is possible to ensure that the scanning process is optimized for specific tissue structures, thereby improving the accuracy and image quality of the scan. Optimizing scanning parameters also helps reduce unnecessary radiation exposure, which is safer for both patients and operators. CT images obtained based on the optimal CT scanning parameters can generate more accurate three-dimensional CT images.

[0016] 2. Through multiple clustering (forming the first sample cluster, the second sample cluster, and the third sample cluster), the test samples are finely divided according to the similarity of their tissue structure information and the characteristics of the CT scanning parameters. This ensures that the samples in each sample cluster have a high degree of similarity in multiple dimensions, thereby improving the pertinence and accuracy of the subsequent three-dimensional reconstruction model. Within each sample cluster, further clustering is performed according to the optimal CT scanning parameters, and the scanning parameter range suitable for the samples in the cluster can be more accurately identified. These parameters will be directly used for three-dimensional reconstruction, which helps to improve the quality and accuracy of the reconstructed image. For each third sample cluster, multiple three-dimensional reconstruction models with different structures are established, and joint training is performed through the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of multiple test samples in the cluster. This training method can make full use of the diversity of samples in the cluster, improve the generalization ability of the model, and enable it to better adapt to the reconstruction needs of different samples. By dividing the samples into smaller clusters through clustering, and independently establishing and training the three-dimensional reconstruction model in each cluster, the overall computational cost can be significantly reduced. This is because the number of samples in each cluster is relatively small and has a high similarity, so data processing and model training can be more efficient. The multi-layer clustering algorithm makes the entire modeling process modular. Each sample cluster and sub-cluster can be regarded as an independent module, which is convenient for subsequent management and optimization of the model. When the model needs to be updated or adjusted, it can be operated on a specific sample cluster without having to reprocess the entire data set. Since each third sample cluster has established multiple 3D reconstruction models with different structures, the most appropriate model can be selected according to specific needs in practical applications. This flexibility makes this method more adaptable to complex and changeable application scenarios.

[0017] 3. Ultrasonic testing is a non-invasive imaging technology that does not require cutting or injection of contrast agents to the target to be reconstructed, thus greatly reducing the risk of harm to patients or samples. Through multiple rounds of testing and calculation of ultrasonic fluctuation values, detailed information of the target to be reconstructed at different positions and angles can be obtained, thereby constructing more comprehensive and accurate tissue structure information. Through continuous iteration of the testing process, the test data can be gradually optimized to ensure that the final generated tissue structure information meets the test conditions and improve the reliability and accuracy of the data. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] This specification will be further described in the form of exemplary embodiments, which will be described in detail by the accompanying drawings. These embodiments are not restrictive, and in these embodiments, the same number represents the same structure, wherein:

[0019] Figure 1It is a flow chart of an artificial intelligence three-dimensional modeling method based on CT big data shown in one embodiment of the present application;

[0020] Figure 2 It is a module diagram of an artificial intelligence three-dimensional modeling system based on CT big data shown in one embodiment of the present application. DETAILED DESCRIPTION

[0021] In order to more clearly illustrate the technical solutions of the embodiments of this specification, the following briefly introduces the drawings required for describing the embodiments.

[0022] Figure 1 is a flow chart of an artificial intelligence 3D modeling method based on CT big data according to some embodiments of this specification, such as Figure 1 As shown, an artificial intelligence three-dimensional modeling method based on CT big data may include the following steps.

[0023] Step 110, obtaining tissue structure information, optimal CT scanning parameters, CT images corresponding to the optimal CT scanning parameters, and three-dimensional CT images of multiple test samples.

[0024] Specifically, the tissue structure information of the test sample may include the name (eg, head, chest, abdomen and pelvis, upper limbs and lower limbs, etc.), and ultrasonic detection data of multiple positions.

[0025] The optimal CT scanning parameters may include the scanning angle of the CT scanning device, etc.

[0026] The three-dimensional CT image can be a three-dimensional CT image of the test sample reconstructed according to the CT image corresponding to the optimal CT scanning parameters by a reconstruction algorithm or the like.

[0027] Furthermore, the optimal CT scanning parameters of the test sample are obtained, including:

[0028] Generate multiple groups of test CT scanning parameters, wherein the test CT scanning parameters at least include multiple CT scanning angles, and the test CT scanning parameters may also include X-ray tube voltage, tube current, etc.;

[0029] Obtaining CT images corresponding to each set of test CT scanning parameters;

[0030] For each set of test CT scanning parameters, a CT reconstruction result corresponding to the test CT scanning parameters is generated according to the CT images corresponding to the test CT scanning parameters by a three-dimensional reconstruction algorithm;

[0031] Based on the CT reconstruction results corresponding to each set of test CT scanning parameters, multiple sets of test CT scanning parameters are screened to determine the optimal CT scanning parameters of the test samples.

[0032] Specifically, for each set of test CT scan parameters, the scores of the test CT scan parameters in multiple evaluation indicators are determined. For example, the clarity, contrast, noise level, etc. of the CT reconstruction results corresponding to the test CT scan parameters can be evaluated by artificial or machine learning models to determine the scores of the test CT scan parameters in the CT reconstruction quality evaluation indicators. The radiation dose corresponding to the test CT scan parameters is evaluated by artificial or machine learning models to determine the scores of the test CT scan parameters in the radiation dose evaluation indicators. For the scores of the test CT scan parameters in multiple evaluation indicators, the screening scores of the test CT scan parameters are calculated, and the test CT scan parameters with the largest screening scores are used as the optimal CT scan parameters.

[0033] Step 120 , establishing a plurality of three-dimensional reconstruction models based on the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the plurality of test samples.

[0034] Furthermore, step 120 specifically includes:

[0035] Based on the organizational structure information of the multiple test samples, clustering the multiple test samples by using a clustering algorithm (for example, a K-means clustering algorithm, a hierarchical clustering algorithm, etc.) to determine multiple first sample clusters;

[0036] For each first sample cluster, based on the optimal CT scanning parameters of the multiple test samples included in the first sample cluster, clustering the multiple test samples included in the first sample cluster by a clustering algorithm (for example, a K-means clustering algorithm, a hierarchical clustering algorithm, etc.) to determine a plurality of second sample clusters included in the first sample cluster;

[0037] For each second sample cluster, based on the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the multiple test samples included in the second sample cluster, the multiple test samples included in the second sample cluster are clustered by a clustering algorithm to determine the multiple third sample clusters included in the second sample cluster. Specifically, the CT images and three-dimensional CT images of the multiple test samples in the second sample cluster are preprocessed, including image denoising, contrast enhancement, normalization and other operations to eliminate the differences between the images and improve the clustering effect. Effective feature information is extracted from the preprocessed CT images and three-dimensional CT images, and these features may include grayscale values, texture features, shape features, etc. The multiple test samples included in the second sample cluster are clustered by a clustering algorithm (for example, a K-means clustering algorithm, a hierarchical clustering algorithm, etc.) according to the feature similarity between the feature information of the preprocessed CT images and three-dimensional CT images of any two test samples included in the second sample cluster, to determine the multiple third sample clusters included in the second sample cluster;

[0038] For each third sample cluster, multiple three-dimensional reconstruction models corresponding to the third sample cluster are established, and the multiple three-dimensional reconstruction models corresponding to the third sample cluster are jointly trained through CT images and three-dimensional CT images corresponding to optimal CT scanning parameters of multiple test samples included in the third sample cluster, wherein the structures of any two three-dimensional reconstruction models corresponding to the third sample cluster are different, for example, one three-dimensional reconstruction model may be a generative adversarial network (GAN) model, and the other three-dimensional reconstruction model may be a convolutional neural network (CNN) model, etc.

[0039] Specifically, the multiple three-dimensional reconstruction models corresponding to the third sample cluster may be jointly trained according to the following process:

[0040] S21, initializing each 3D reconstruction model corresponding to the third sample cluster, including setting parameters such as initial weights, bias items, and learning rates;

[0041] S22, assigning the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the plurality of test samples included in the pre-processed third sample cluster to each three-dimensional reconstruction model;

[0042] S23, within each 3D reconstruction model, the input data is forward propagated through each layer of the network to calculate a 3D CT image;

[0043] S24, calculating a joint loss function value according to the three-dimensional CT image output by each three-dimensional reconstruction model and the three-dimensional CT image corresponding to the optimal CT scanning parameters through a joint loss function;

[0044] S25, updating the parameters of each 3D reconstruction model through a back propagation algorithm according to the joint loss function value;

[0045] S26. Repeat the above steps S22-S25 until the stopping condition is met (for example, a preset number of iterations is reached, the loss value is less than a certain threshold, etc.).

[0046] Furthermore, the joint loss function is:

[0047]

[0048] Among them, L joint is the joint loss, b 11 、b 12 and b 13 is the preset weight, b 11 、b 12 and b 13 Greater than 0, b 11 +b12 +b 13 =1,a k is the loss weight corresponding to the kth 3D reconstruction model corresponding to the third sample cluster, L k is the loss value of the kth 3D reconstruction model corresponding to the third sample cluster, K is the total number of 3D reconstruction models corresponding to the third sample cluster, N is the total number of training samples used in one round of training for the kth 3D reconstruction model corresponding to the third sample cluster, and Y (n,k) is the 3D CT image of the nth test sample output by the kth 3D reconstruction model corresponding to the third sample cluster, Y (n,true) is a three-dimensional CT image corresponding to the optimal CT scanning parameters of the n-th test sample corresponding to the third sample cluster for training the k-th three-dimensional reconstruction model; The weighted average of the losses of all 3D reconstruction models is calculated. This weighted average method ensures that each model contributes to its total loss, and the weight of each model can be adjusted according to actual needs;. The average of all model losses is calculated, regardless of the weights of the individual models. It provides an indicator of the overall performance of all models, which helps monitor overall progress during training. By incorporating this average loss into the joint loss function, all models are encouraged to maintain a consistent performance level overall, preventing some models from deviating too much; The average of the squares of the difference between the loss value of each model and its average loss value, i.e., the variance of the loss, is calculated. It reflects the degree of dispersion of performance between models. By minimizing this variance term, each model can be encouraged to gradually converge during the training process and reduce the performance differences between them. This is crucial for achieving collaborative work between multiple models and improving the stability and robustness of the overall system.

[0049] It is understandable that due to the different structures of each 3D reconstruction model, they can extract data features from different angles and levels, thereby improving the overall reconstruction effect. Through joint training, the advantages of different machine learning models can be used to complement each other. Since multiple models are used for joint training, even if a model has problems or performance degradation during training, other models can still work normally, thereby improving robustness.

[0050] Step 130: Obtain the organizational structure information of the target to be reconstructed.

[0051] Furthermore, step 130 specifically includes:

[0052] S11, acquiring ultrasonic detection data of multiple positions of the target to be reconstructed;

[0053] S12, calculating the ultrasonic fluctuation value of the target to be reconstructed based on the ultrasonic detection data of multiple positions of the target to be reconstructed in each round;

[0054] S13, judging whether the detection condition is met according to the ultrasonic wave value of the target to be reconstructed, if not, executing S14, if yes, executing S15;

[0055] S14, performing the next round of ultrasonic detection to obtain ultrasonic detection data of multiple positions of the target to be reconstructed, wherein the multiple positions of any two rounds are different, and S12 is executed;

[0056] S15. Generate tissue structure information of the target to be reconstructed, wherein the tissue structure information of the target to be reconstructed at least includes ultrasonic detection data of multiple positions of the target to be reconstructed in each round.

[0057] Specifically, for each position, variational modal decomposition is performed on the ultrasonic detection data of the position to determine multiple modal components corresponding to the ultrasonic detection data of the position, and the characteristics of each modal component (for example, center frequency, bandwidth, mean, standard deviation, etc.) are extracted to generate a feature matrix of the ultrasonic detection data of the position, wherein a row vector of the feature matrix includes the characteristics of a modal component.

[0058] The ultrasonic wave value of the target to be reconstructed is calculated according to the characteristic matrix of the ultrasonic detection data at each position.

[0059] Furthermore, the ultrasonic wave value of the target to be reconstructed can be calculated according to the following formula:

[0060]

[0061] Among them, σ t is the ultrasonic fluctuation value of the target to be reconstructed after completing t rounds of ultrasonic detection, S (i,j) is the matrix similarity between the feature matrix of the ith position and the feature matrix of the jth position for ultrasonic testing, M is the total number of positions for ultrasonic testing after completing t rounds of ultrasonic testing, η is a preset parameter, η is greater than 0, V ((e,f),i) is the value of the element in the eth row and jth column of the feature matrix at the i-th position to be tested by ultrasonic wave, V ((e,f),j) is the value of the element in the e-th row and j-th column of the feature matrix at the j-th position under ultrasonic detection, E is the total number of rows of the feature matrix, and F is the total number of columns of the feature matrix.

[0062] When the ultrasonic fluctuation value of the target to be reconstructed is less than the ultrasonic fluctuation value threshold, it is determined that the detection condition is met.

[0063] Step 140 : determining a plurality of reference test samples from the plurality of test samples based on the tissue structure information of the target to be reconstructed and the tissue structure information of the plurality of test samples.

[0064] Furthermore, step 140 specifically includes:

[0065] Based on the tissue structure information of the target to be reconstructed and the tissue structure information of the test sample corresponding to the cluster center of each first sample cluster, a reference first sample cluster is determined from multiple first sample clusters, for example, the first sample cluster to which the test sample having a tissue structure information similarity greater than a first tissue structure information similarity threshold belongs is taken as the reference first sample cluster;

[0066] Based on the tissue structure information of the target to be reconstructed and the tissue structure information of each test sample included in the reference first sample cluster, multiple reference test samples are determined from the multiple test samples included in the reference first sample cluster. For example, the test samples whose tissue structure information similarity with the reference first sample cluster is greater than a second tissue structure information similarity threshold are used as reference test samples.

[0067] Step 150 : determining optimal CT scanning parameters of the target to be reconstructed based on the optimal CT scanning parameters of a plurality of reference test samples.

[0068] Furthermore, step 150 specifically includes:

[0069] The optimal CT scanning parameters of the target to be reconstructed are determined by a parameter determination model based on the optimal CT scanning parameters of multiple reference test samples, wherein the parameter determination model can be a Convolutional Neural Networks (CNN) model.

[0070] Step 160 , based on the optimal CT scanning parameters of the target to be reconstructed, perform CT scanning on the target to be reconstructed to obtain a CT image of the target to be reconstructed.

[0071] Furthermore, step 160 specifically includes:

[0072] Based on the optimal CT scanning parameters of the target to be reconstructed, performing CT scanning on the target to be reconstructed;

[0073] During the CT scanning process, the scanning status information of the target to be reconstructed is obtained;

[0074] Based on the scanning status information of the target to be reconstructed, it is determined whether to perform a second CT scan. If so, the target CT scanning angle is determined, and a second CT scan is performed according to the target CT scanning angle.

[0075] Furthermore, the scanning status information of the target to be reconstructed is obtained, including:

[0076] Setting a plurality of markers on the target to be reconstructed according to the optimal CT scanning parameters and tissue structure information of the target to be reconstructed;

[0077] Acquire a scanned state image of the target to be reconstructed;

[0078] The scanning state information of the object to be reconstructed is determined according to the scanning state image of the object to be reconstructed.

[0079] Specifically, the position determination model can be used to determine the position that has less impact on the CT scan and is more likely to be displaced according to the optimal CT scan parameters and tissue structure information of the target to be reconstructed, and a marker can be set at the position. The position determination model can be a convolutional neural network model.

[0080] The relative position relationship of the multiple markers is determined according to the scanning state image of the target to be reconstructed, thereby determining the displacement of the target to be reconstructed during the scanning process. The scanning state information of the target to be reconstructed may include the displacement of the target to be reconstructed during the scanning process.

[0081] If the displacement of the target to be reconstructed during the scanning process is greater than the displacement threshold, it is determined to perform a second CT scan.

[0082] Step 170 : Generate a three-dimensional CT image of the target to be reconstructed based on the CT image of the target to be reconstructed by using multiple three-dimensional reconstruction models.

[0083] Furthermore, step 170 specifically includes:

[0084] Determining a target second sample cluster from a plurality of second sample clusters based on an optimal CT scanning parameter of the target to be reconstructed;

[0085] Based on the CT image of the target to be reconstructed, determining a target third sample cluster from a plurality of third sample clusters included in the target second sample cluster;

[0086] A three-dimensional CT image of the target to be reconstructed is generated based on the CT image of the target to be reconstructed by using multiple three-dimensional reconstruction models corresponding to the third sample cluster of the target.

[0087] Specifically, the parameter difference between the optimal CT scanning parameters of the target to be reconstructed and the optimal CT scanning parameters of the test sample corresponding to the cluster center of each second sample cluster is calculated, and the second sample cluster to which the test sample having a parameter difference less than a parameter difference threshold belongs is taken as the target second sample cluster.

[0088] Calculate the feature similarity between the CT image of the target to be reconstructed and the CT image of the test sample corresponding to the cluster center of each third sample cluster included in the target second sample cluster, and take the third sample cluster to which the test sample with a feature similarity less than a feature similarity threshold belongs as the target third sample cluster.

[0089] Figure 2 is a module schematic diagram of an artificial intelligence 3D modeling system based on CT big data according to some embodiments of this specification, such as Figure 2 As shown, an artificial intelligence three-dimensional modeling system based on CT big data may include a data acquisition module, a model building module, an information acquisition module, a sample determination module, a parameter determination module, a CT scanning module and a three-dimensional reconstruction module.

[0090] The data acquisition module can be used to obtain tissue structure information, optimal CT scanning parameters, CT images corresponding to the optimal CT scanning parameters, and three-dimensional CT images of multiple test samples;

[0091] The model building module can be used to build multiple three-dimensional reconstruction models based on CT images and three-dimensional CT images corresponding to optimal CT scanning parameters of multiple test samples;

[0092] The information acquisition module can be used to obtain the organizational structure information of the target to be reconstructed;

[0093] The sample determination module may be used to determine a plurality of reference test samples from a plurality of test samples based on the tissue structure information of the target to be reconstructed and the tissue structure information of the plurality of test samples;

[0094] The parameter determination module may be used to determine optimal CT scanning parameters of a target to be reconstructed based on optimal CT scanning parameters of a plurality of reference test samples;

[0095] The CT scanning module can be used to perform CT scanning on the target to be reconstructed based on the optimal CT scanning parameters of the target to be reconstructed, so as to obtain a CT image of the target to be reconstructed;

[0096] The three-dimensional reconstruction module can be used to generate a three-dimensional CT image of the target to be reconstructed based on the CT image of the target to be reconstructed by using multiple three-dimensional reconstruction models.

[0097] An artificial intelligence three-dimensional modeling system based on CT big data can be used to execute an artificial intelligence three-dimensional modeling method based on CT big data. For more descriptions of an artificial intelligence three-dimensional modeling system based on CT big data, please refer to the relevant descriptions of an artificial intelligence three-dimensional modeling method based on CT big data, which will not be repeated here.

[0098] Finally, it should be understood that the embodiments described in this specification are only used to illustrate the principles of the embodiments of this specification. Other variations may also fall within the scope of this specification. Therefore, as an example and not a limitation, alternative configurations of the embodiments of this specification may be considered consistent with the teachings of this specification. Accordingly, the embodiments of this specification are not limited to the embodiments explicitly introduced and described in this specification.

Claims

1. An artificial intelligence 3D modeling method based on CT big data, characterized in that: include: Acquire tissue structure information, optimal CT scanning parameters, CT images corresponding to the optimal CT scanning parameters, and three-dimensional CT images of multiple test samples; Establishing a plurality of three-dimensional reconstruction models based on the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the plurality of test samples; Obtain the organizational structure information of the target to be rebuilt; Determining a plurality of reference test samples from the plurality of test samples based on the tissue structure information of the target to be reconstructed and the tissue structure information of the plurality of test samples; Determining optimal CT scanning parameters of the target to be reconstructed based on the optimal CT scanning parameters of the multiple reference test samples; Based on the optimal CT scanning parameters of the target to be reconstructed, performing CT scanning on the target to be reconstructed to obtain a CT image of the target to be reconstructed; A three-dimensional CT image of the target to be reconstructed is generated based on the CT image of the target to be reconstructed by using the multiple three-dimensional reconstruction models.

2. The artificial intelligence three-dimensional modeling method based on CT big data according to claim 1 is characterized in that: Obtaining optimal CT scanning parameters of the test sample, including: generating a plurality of sets of test CT scanning parameters, wherein the test CT scanning parameters at least include a plurality of CT scanning angles; Obtaining CT images corresponding to each set of the test CT scanning parameters; For each set of the test CT scan parameters, generating a CT reconstruction result corresponding to the test CT scan parameters according to the CT images corresponding to the test CT scan parameters by a three-dimensional reconstruction algorithm; Based on the CT reconstruction results corresponding to each set of the test CT scanning parameters, the multiple sets of test CT scanning parameters are screened to determine the optimal CT scanning parameters of the test sample.

3. The artificial intelligence three-dimensional modeling method based on CT big data according to claim 1 is characterized in that: Based on the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the multiple test samples, multiple three-dimensional reconstruction models are established, including: Based on the organizational structure information of the plurality of test samples, clustering the plurality of test samples by a clustering algorithm to determine a plurality of first sample clusters; For each of the first sample clusters, based on the optimal CT scanning parameters of the multiple test samples included in the first sample cluster, clustering the multiple test samples included in the first sample cluster by a clustering algorithm to determine multiple second sample clusters included in the first sample cluster; For each of the second sample clusters, based on the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the multiple test samples included in the second sample cluster, clustering the multiple test samples included in the second sample cluster by a clustering algorithm to determine multiple third sample clusters included in the second sample cluster; For each of the third sample clusters, multiple three-dimensional reconstruction models corresponding to the third sample cluster are established, and the multiple three-dimensional reconstruction models corresponding to the third sample cluster are jointly trained through CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the multiple test samples included in the third sample cluster, wherein the structures of any two three-dimensional reconstruction models corresponding to the third sample cluster are different.

4. The artificial intelligence three-dimensional modeling method based on CT big data according to claim 3 is characterized in that: Obtain the organizational structure information of the target to be rebuilt, including: S11, acquiring ultrasonic detection data of multiple positions of the target to be reconstructed; S12, calculating the ultrasonic fluctuation value of the target to be reconstructed based on the ultrasonic detection data of multiple positions of the target to be reconstructed in each round; S13, judging whether the detection condition is met according to the ultrasonic wave fluctuation value of the target to be reconstructed, if not, executing S14, if yes, executing S15; S14, performing the next round of ultrasonic detection to obtain ultrasonic detection data of multiple positions of the target to be reconstructed, wherein the multiple positions of any two rounds are different, and S12 is executed; S15. Generate tissue structure information of the target to be reconstructed, wherein the tissue structure information of the target to be reconstructed at least includes ultrasonic detection data of multiple positions of the target to be reconstructed in each round.

5. The artificial intelligence three-dimensional modeling method based on CT big data according to claim 4 is characterized in that: Based on the tissue structure information of the target to be reconstructed and the tissue structure information of the multiple test samples, determining multiple reference test samples from the multiple test samples includes: Determining a reference first sample cluster from the plurality of first sample clusters based on the tissue structure information of the target to be reconstructed and the tissue structure information of the test sample corresponding to the cluster center of each of the first sample clusters; Based on the tissue structure information of the to-be-reconstructed object and the tissue structure information of each test sample included in the reference first sample cluster, a plurality of reference test samples are determined from the plurality of test samples included in the reference first sample cluster.

6. The artificial intelligence three-dimensional modeling method based on CT big data according to any one of claims 1 to 5, characterized in that: Determining the optimal CT scanning parameters of the target to be reconstructed based on the optimal CT scanning parameters of the multiple reference test samples includes: The optimal CT scanning parameters of the target to be reconstructed are determined by using a parameter determination model based on the optimal CT scanning parameters of the multiple reference test samples.

7. The artificial intelligence three-dimensional modeling method based on CT big data according to any one of claims 3 to 5, characterized in that: Performing a CT scan on the target to be reconstructed based on the optimal CT scanning parameters of the target to be reconstructed includes: Performing a CT scan on the target to be reconstructed based on the optimal CT scanning parameters of the target to be reconstructed; During the CT scanning process, obtaining scanning status information of the target to be reconstructed; Based on the scanning status information of the target to be reconstructed, it is determined whether to perform a secondary CT scan. If so, a target CT scanning angle is determined, and a secondary CT scan is performed according to the target CT scanning angle.

8. The artificial intelligence three-dimensional modeling method based on CT big data according to claim 7, characterized in that: Obtaining scanning status information of the target to be reconstructed, including: According to the optimal CT scanning parameters and tissue structure information of the target to be reconstructed, a plurality of markers are set on the target to be reconstructed; Acquiring a scanned state image of the target to be reconstructed; The scanning state information of the object to be reconstructed is determined according to the scanning state image of the object to be reconstructed.

9. The artificial intelligence three-dimensional modeling method based on CT big data according to any one of claims 3 to 5, characterized in that: Generating a three-dimensional CT image of the target to be reconstructed based on the CT image of the target to be reconstructed by using the multiple three-dimensional reconstruction models includes: Determining a target second sample cluster from the plurality of second sample clusters based on the optimal CT scanning parameters of the target to be reconstructed; Based on the CT image of the target to be reconstructed, determining a target third sample cluster from a plurality of third sample clusters included in the target second sample cluster; A three-dimensional CT image of the target to be reconstructed is generated based on the CT image of the target to be reconstructed by using a plurality of three-dimensional reconstruction models corresponding to the target third sample cluster.

10. An artificial intelligence 3D modeling system based on CT big data, characterized in that: The method for performing an artificial intelligence three-dimensional modeling method based on CT big data as described in any one of claims 1 to 9 comprises: A data acquisition module, used to acquire tissue structure information, optimal CT scanning parameters, CT images corresponding to the optimal CT scanning parameters, and three-dimensional CT images of multiple test samples; A model building module, used to build multiple three-dimensional reconstruction models based on the CT images and three-dimensional CT images corresponding to the optimal CT scanning parameters of the multiple test samples; An information acquisition module, used to acquire the organizational structure information of the target to be reconstructed; A sample determination module, configured to determine a plurality of reference test samples from the plurality of test samples based on the tissue structure information of the target to be reconstructed and the tissue structure information of the plurality of test samples; A parameter determination module, configured to determine the optimal CT scanning parameters of the target to be reconstructed based on the optimal CT scanning parameters of the multiple reference test samples; A CT scanning module, configured to perform a CT scan on the target to be reconstructed based on the optimal CT scanning parameters of the target to be reconstructed, so as to obtain a CT image of the target to be reconstructed; The three-dimensional reconstruction module is used to generate a three-dimensional CT image of the target to be reconstructed based on the CT image of the target to be reconstructed by using the multiple three-dimensional reconstruction models.

Images