Uterine myoma intelligent measurement and evaluation system based on 3D U-Net

Abstract

The invention provides a 3D U-Net-based hysteromyoma intelligent measurement and evaluation system, and the system comprises an intelligent recognition module which is used for constructing a 3D U-Net network model, inputting a real-time image into the 3D U-Net network model for myoma boundary recognition, and judging whether a patient suffers from hysteromyoma or not; the three-dimensional reconstruction module is used for performing three-dimensional reconstruction on the real-time image when the judgment result is that the patient suffers from the hysteromyoma; the growth prediction module is used for performing uterine fibroid growth prediction according to the three-dimensional reconstruction result and outputting a growth prediction result; and the three-dimensional relationship between the myoma and the uterine muscular layer, the blood vessel and other organs can be observed more clearly, so that a more accurate basis is provided for formulating a personalized treatment scheme.

Description

Technical Field

[0001] This invention relates to the field of medical image processing technology, and in particular to an intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net. Background Technology

[0002] Currently, uterine fibroids are the most common benign tumors of the female reproductive system, and their incidence is on the rise worldwide. According to statistics from the International Agency for Research on Cancer, uterine fibroids affect approximately 20% to 40% of women of childbearing age, significantly impacting their health and quality of life.

[0003] In the diagnosis and treatment of uterine fibroids, traditional measurement methods mainly rely on two-dimensional ultrasound or MRI images, estimating volume by manually measuring the length, width, and height of the fibroid. However, this method has significant limitations: First, manual measurement is highly subjective, and the measurement results from different doctors may vary considerably, directly affecting the accuracy of diagnosis and the formulation of treatment plans; second, uterine fibroids often present irregular shapes, and two-dimensional images cannot fully reflect their three-dimensional structure, leading to large errors in volume estimation; while MRI can provide more detailed images, it is costly and time-consuming. In addition, traditional methods cannot monitor the growth rate of fibroids in real time, which is crucial for assessing the growth trend of fibroids and treatment effectiveness.

[0004] Therefore, in order to overcome the above-mentioned technical problems, the present invention provides an intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net. Summary of the Invention

[0005] This invention provides an intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net. It extends the two-dimensional structure of U-Net to three dimensions through the 3D U-Net network model to adapt to the processing of volumetric data. In the detection and segmentation of uterine fibroids, it provides accurate three-dimensional information. By using 3D U-Net and image three-dimensional reconstruction technology, doctors can more clearly observe the three-dimensional relationship between fibroids and the myometrium, blood vessels and other organs, thereby providing a more accurate basis for formulating personalized treatment plans.

[0006] A 3D U-Net-based intelligent measurement and assessment system for uterine fibroids includes:

[0007] The intelligent recognition module is used to construct a 3D U-Net network model and input real-time images into the 3D U-Net network model to identify fibroid boundaries and determine whether a woman has uterine fibroids.

[0008] The 3D reconstruction module is used to perform 3D reconstruction on real-time images when the diagnosis is uterine fibroids.

[0009] The growth prediction module is used to predict the growth of uterine fibroids based on the 3D reconstruction results and output the growth prediction results.

[0010] Preferably, an intelligent measurement and assessment system for uterine fibroids based on 3D U-Net includes an intelligent recognition module comprising:

[0011] Model building unit, used to build 3D U-Net network models;

[0012] The analysis unit is used to acquire real-time images and input them into a 3D U-Net network model for fibroid boundary recognition to determine whether a woman has uterine fibroids.

[0013] Preferably, a model building unit for an intelligent measurement and assessment system for uterine fibroids based on 3D U-Net includes:

[0014] The image acquisition subunit is used to acquire a first sample image dataset of uterine fibroids based on a preset ultrasound device.

[0015] The data preprocessing subunit is used to preprocess the first sample image dataset to obtain the second sample image dataset.

[0016] The data label addition sub-unit is used to add image data labels to the second sample image dataset based on the image features of the sample images, wherein the image data labels correspond one-to-one with the second sample images;

[0017] The sub-unit is used to divide the image data labels and the corresponding second sample image set into training set, validation set and test set according to a preset ratio;

[0018] The fibroid region extraction subunit is used to extract fibroid regions from the training set, validation set, and test set respectively based on image segmentation techniques.

[0019] The training subunit is used to obtain the initial 3D U-Net network model and train the initial 3D U-Net network model according to the fibroid regions corresponding to the training set, validation set and test set to obtain the 3DU-Net network model.

[0020] Preferably, in a 3D U-Net-based intelligent measurement and evaluation system for uterine fibroids, the data preprocessing subunit performs data preprocessing on the first sample image dataset, including histogram equalization, data augmentation, and normalization.

[0021] Preferably, in a 3D U-Net-based intelligent measurement and evaluation system for uterine fibroids, the sub-units are divided in a preset ratio of 8:1:1.

[0022] Preferably, in a 3D U-Net-based intelligent measurement and assessment system for uterine fibroids, the intelligent recognition module includes:

[0023] The 3D U-Net network model includes: an encoder and a decoder;

[0024] The encoder is composed of multiple alternating three-dimensional convolutional layers and pooling layers;

[0025] The decoder consists of multiple upper sampling layers and three-dimensional convolutional layers.

[0026] Preferably, in a 3D U-Net-based intelligent measurement and evaluation system for uterine fibroids, after obtaining the 3D U-Net network model in the training subunit, the system further includes:

[0027] The network improvement subunit is used to improve the network structure of the 3D U-Net network model based on a full-scale skip connection method.

[0028] The loss optimization and improvement subunit is used to obtain the target loss function by fusing the Dice loss function and the binary cross-entropy loss function. At the same time, the target loss function is used to optimize and train the 3D U-Net network model after network structure improvement to obtain the target 3D U-Net network model.

[0029] Preferably, a loss optimization and improvement subunit for an intelligent measurement and assessment system for uterine fibroids based on 3D U-Net includes:

[0030] The Dice loss function formula is as follows:

[0031]

[0032] Where DiceLoss represents the Dice loss function; X represents the image segmentation sample; Y represents the image segmentation result sample;

[0033] The formula for the binary cross-entropy loss function is as follows:

[0034]

[0035] Where BCELoss represents the binary cross-entropy loss function; N represents the number of groups of objects predicted by the 3D U-Net network model; i represents the group index value of the predicted object; y i p(y) represents the true result value of the i-th predicted object; i ) represents the predicted result value of the i-th group of predicted objects; lg(·) represents the logarithmic function with base 10;

[0036] The Dice loss function and the binary cross-entropy loss function are fused according to the following formula to obtain the target loss function;

[0037] Loss=αDiceLoss+βBCELoss;

[0038] Where Loss represents the objective loss function; α and β represent hyperparameters.

[0039] Preferably, a 3D U-Net-based intelligent measurement and assessment system for uterine fibroids includes a three-dimensional reconstruction module, comprising:

[0040] The spatial coordinate system construction unit is used to acquire real-time image sequences and construct a spatial coordinate system when the judgment result is that the patient has uterine fibroids. At the same time, the image data of the acquired real-time image sequence is read into the spatial coordinate system in sequence.

[0041] 3D reconstruction unit, used for:

[0042] Based on the input results, the target cube is determined in the spatial coordinate system with every 8 adjacent coordinate points as vertices, and the target cube is taken as a voxel, with the 8 adjacent coordinate points as the 8 corner points of the voxel.

[0043] The edges of the boundary voxels and the corners of the isosurfaces are calculated according to the first preset method;

[0044] The normal vectors of each corner point of the voxel are calculated based on the edges of the boundary voxel and the corner points of the isosurface, and the triangular facets are determined based on the normal vectors. At the same time, the normal vectors of the vertices of the triangular facets are calculated based on the first preset method.

[0045] Based on the normal vectors of the vertices of the triangular facets and their coordinates, an isosurface image is drawn to complete the 3D reconstruction of the real-time image.

[0046] Preferably, a 3D U-Net-based intelligent measurement and assessment system for uterine fibroids includes a growth prediction module, comprising:

[0047] The diagnostic result output unit is used to visualize the three-dimensional reconstruction results, determine the image features corresponding to the three-dimensional reconstruction results, and output the diagnostic results of uterine fibroids based on the image features.

[0048] The growth prediction unit is used to pre-train the fibroid growth rate prediction model, input the 3D reconstruction results into the fibroid growth rate prediction model to predict the growth of uterine fibroids, and output the growth prediction results.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0050] By extending the two-dimensional structure of U-Net to three dimensions through the 3D U-Net network model, it can be adapted to the processing of volumetric data. In the detection and segmentation of uterine fibroids, it provides accurate three-dimensional information. Using 3D U-Net and image 3D reconstruction technology, doctors can more clearly observe the three-dimensional relationship between fibroids and the uterine myometrium, blood vessels, and other organs, thus providing a more accurate basis for developing personalized treatment plans.

[0051] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in this application.

[0052] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0053] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0054] Figure 1 This is a structural diagram of an intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net, as described in an embodiment of the present invention.

[0055] Figure 2 This is a flowchart of the system's operation in an embodiment of the present invention;

[0056] Figure 3 This is a flowchart illustrating the training and testing process in an embodiment of the present invention;

[0057] Figure 4 This is a structural diagram of the volumetric data space coordinate system in an embodiment of the present invention;

[0058] Figure 5 This is a diagram of the improved 3D U-Net network structure in an embodiment of the present invention. Detailed Implementation

[0059] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0060] Example 1:

[0061] This embodiment provides an intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net, such as... Figure 1 As shown, it includes:

[0062] The intelligent recognition module is used to construct a 3D U-Net network model and input real-time images into the 3D U-Net network model to identify fibroid boundaries and determine whether a woman has uterine fibroids.

[0063] The 3D reconstruction module is used to perform 3D reconstruction on real-time images when the diagnosis is uterine fibroids.

[0064] The growth prediction module is used to predict the growth of uterine fibroids based on the 3D reconstruction results and output the growth prediction results.

[0065] In this embodiment, the specific workflow diagram is as follows: Figure 2 As shown.

[0066] The working principle and beneficial effects of the above technical solution are as follows: By extending the two-dimensional structure of U-Net to three dimensions through the 3D U-Net network model, it can adapt to the processing of volume data, providing accurate three-dimensional information in the detection and segmentation of uterine fibroids. With the help of 3D U-Net and image three-dimensional reconstruction technology, doctors can more clearly observe the three-dimensional relationship between fibroids and uterine myometrium, blood vessels and other organs, thereby providing a more accurate basis for formulating personalized treatment plans.

[0067] Example 2:

[0068] Based on Example 1, this example provides an intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net, including an intelligent recognition module:

[0069] Model building unit, used to build 3D U-Net network models;

[0070] The analysis unit is used to acquire real-time images and input them into a 3D U-Net network model for fibroid boundary recognition to determine whether a woman has uterine fibroids.

[0071] The beneficial effects of the above technical solution are: by constructing a 3D U-Net network model, the boundary of fibroids can be effectively identified, thereby providing accurate data for judgment.

[0072] Example 3:

[0073] Based on Example 2, this example provides an intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net, including a model building unit comprising:

[0074] The image acquisition subunit is used to acquire a first sample image dataset of uterine fibroids based on a preset ultrasound device.

[0075] The data preprocessing subunit is used to preprocess the first sample image dataset to obtain the second sample image dataset.

[0076] The data label addition sub-unit is used to add image data labels to the second sample image dataset based on the image features of the sample images, wherein the image data labels correspond one-to-one with the second sample images;

[0077] The sub-unit is used to divide the image data labels and the corresponding second sample image set into training set, validation set and test set according to a preset ratio;

[0078] The fibroid region extraction subunit is used to extract fibroid regions from the training set, validation set, and test set respectively based on image segmentation techniques.

[0079] The training subunit is used to obtain the initial 3D U-Net network model and train the initial 3D U-Net network model according to the fibroid regions corresponding to the training set, validation set and test set to obtain the 3DU-Net network model.

[0080] In this embodiment, the preset ultrasound device is a high-resolution ultrasound device to obtain detailed images of the uterine fibroids (i.e., the first sample image set). The first sample image set must have sufficient quality so that the subsequent algorithm can accurately identify and measure the size and shape of the fibroids.

[0081] In this embodiment, the first sample image dataset is preprocessed, including histogram equalization, data augmentation, and normalization. Based on the preprocessing of the first sample image dataset, a second sample image dataset is obtained. Histogram equalization enhances image contrast, and standardization techniques are used to ensure that image data from different patients remain consistent. Data augmentation techniques, such as rotation, scaling, and flipping, increase the algorithm's generalization ability for different uterine fibroid shapes. Normalization is used to avoid oscillations caused by gradient updates and to accelerate model convergence so that the optimal solution can be obtained in a shorter time.

[0082] In this embodiment, the image data is tagged, and the tag content includes information such as the size, location, and shape of the uterine fibroid. For example, when the fibroid appears gravelly or has a hard shell, the tag content is uterine fibroid calcification.

[0083] In this embodiment, the labels and corresponding image data are divided into training set, validation set and test set in a ratio of 8:1:1. This ensures that the training set and test set do not overlap, making the experimental data more rigorous.

[0084] In this embodiment, image segmentation technology is applied to separate the fibroid region from the surrounding tissue. By employing an advanced segmentation algorithm—the 3D U-Net network model—accurate identification of the fibroid boundary can be achieved.

[0085] In this embodiment, the training and testing flowchart is as follows: Figure 3 As shown.

[0086] The beneficial effects of the above technical solution are as follows: Through data preprocessing, image contrast is effectively enhanced, ensuring that image data from different patients can remain consistent, increasing the algorithm's generalization ability for different morphologies of uterine fibroids, avoiding oscillations caused by gradient updates, and accelerating model convergence so as to obtain the optimal solution in a shorter time. By adding image data labels and dividing the second sample image set into training, validation, and test sets according to a preset ratio, it can be ensured that the training and test sets do not overlap, making the experimental data more rigorous, thereby ensuring that the constructed 3D U-Net network model is more accurate and achieving accurate identification of fibroid boundaries.

[0087] Example 4:

[0088] Based on Example 1, this example provides an intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net. In the intelligent recognition module,

[0089] The 3D U-Net network model includes: an encoder and a decoder;

[0090] The encoder is composed of multiple alternating three-dimensional convolutional layers and pooling layers;

[0091] The decoder consists of multiple upper sampling layers and three-dimensional convolutional layers.

[0092] The working principle and beneficial effects of the above technical solution are as follows: The 3D U-Net network model utilizes convolutional neural network technology in deep learning, consisting of an encoder and a decoder. The encoder is typically composed of multiple alternating 3D convolutional layers and pooling layers. The convolutional layers are mainly used to extract features from 3D images and further obtain deeper information from the image using the spatial dimension information. The pooling layers can reduce the spatial dimension of the feature maps to extract higher-level features. Through multiple convolutional and pooling operations, automatic extraction of image features can be achieved. The decoder consists of a series of upsampling layers and 3D convolutions, mainly responsible for mapping the high-level feature maps extracted from the encoding part back to the original volume size, while restoring the detailed information of the image, thereby achieving accurate segmentation. The skip connections in 3D U-Net are used to achieve feature fusion between corresponding layers of the encoder and decoder, which helps to preserve global and local information and improve the accuracy of segmentation.

[0093] Example 5:

[0094] Based on Example 3, this example provides an intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net. In the training subunit, after obtaining the 3D U-Net network model, it further includes:

[0095] The network improvement subunit is used to improve the network structure of the 3D U-Net network model based on a full-scale skip connection method.

[0096] The loss optimization and improvement subunit is used to obtain the target loss function by fusing the Dice loss function and the binary cross-entropy loss function. At the same time, the target loss function is used to optimize and train the 3D U-Net network model after network structure improvement to obtain the target 3D U-Net network model.

[0097] In this embodiment, the loss optimization improvement subunit includes:

[0098] The Dice loss function formula is as follows:

[0099]

[0100] Where DiceLoss represents the Dice loss function; X represents the image segmentation sample; Y represents the image segmentation result sample;

[0101] The formula for the binary cross-entropy loss function is as follows:

[0102]

[0103] Where BCELoss represents the binary cross-entropy loss function; N represents the number of groups of objects predicted by the 3D U-Net network model; i represents the group index value of the predicted object; y i p(y) represents the true result value of the i-th predicted object; i ) represents the predicted result value of the i-th group of predicted objects; lg(·) represents the logarithmic function with base 10;

[0104] The Dice loss function and the binary cross-entropy loss function are fused according to the following formula to obtain the target loss function;

[0105] Loss=αDiceLoss+βBCELoss;

[0106] Where Loss represents the objective loss function; α and β represent hyperparameters.

[0107] The working principle and beneficial effects of the above technical solution are as follows: To ensure the accuracy of recognition, the 3D U-Net network structure is improved by using full-scale skip connections. This approach allows for the full extraction of image information features at different scales, enabling the fusion of low-level details and high-level semantics, thus further enhancing the segmentation ability of the network model. The improvement is particularly noticeable for segmenting lesions near the core region. The improved 3D U-Net network structure is shown in the diagram below. Figure 5 As shown, in the 3D U-Net architecture, the loss function needs to accurately reflect the difference between the segmentation result and the true label, directly affecting the model's performance in medical image segmentation tasks. The cross-entropy loss function may not be effective enough in dealing with class imbalance, while the Dice loss function can handle class imbalance and can better measure the similarity between the segmentation result and the true label. Therefore, the weighted cross-entropy loss function is adopted to alleviate the class imbalance problem by assigning different weights to different classes. The composite loss function combining cross-entropy and Dice loss can further improve the segmentation accuracy. By optimizing the loss function, the 3D U-Net network model can better adapt to the complexity of medical image segmentation, thereby providing more accurate diagnostic support in clinical applications.

[0108] Example 6:

[0109] Based on Example 1, this example provides an intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net, including a three-dimensional reconstruction module:

[0110] The spatial coordinate system construction unit is used to acquire real-time image sequences and construct a spatial coordinate system when the judgment result is that the patient has uterine fibroids. At the same time, the image data of the acquired real-time image sequence is read into the spatial coordinate system in sequence.

[0111] 3D reconstruction unit, used for:

[0112] Based on the input results, the target cube is determined in the spatial coordinate system with every 8 adjacent coordinate points as vertices, and the target cube is taken as a voxel, with the 8 adjacent coordinate points as the 8 corner points of the voxel.

[0113] The edges of the boundary voxels and the corners of the isosurfaces are calculated according to the first preset method;

[0114] The normal vectors of each corner point of the voxel are calculated based on the edges of the boundary voxel and the corner points of the isosurface, and the triangular facets are determined based on the normal vectors. At the same time, the normal vectors of the vertices of the triangular facets are calculated based on the first preset method.

[0115] Based on the normal vectors of the vertices of the triangular facets and their coordinates, an isosurface image is drawn to complete the 3D reconstruction of the real-time image.

[0116] In this embodiment, the specific process includes: (1) Constructing a voxel: defining a Figure 5 The spatial coordinate system shown is used to read the data of the image sequence into the coordinate system one by one, and then a small cube with 8 adjacent coordinate points as vertices is used as a voxel. These 8 adjacent vertices are the 8 corner points of the voxel; (2) Calculate the intersection point: use the linear interpolation method to calculate the intersection point of the edge of the boundary voxel (the boundary voxel that intersects with the isosurface) with the isosurface; (3) Find the normal vector: obtain the normal vector of each corner point of the voxel according to the central difference method, form a triangular facet, and then use the linear interpolation method to calculate the normal vector of the vertex of the triangular facet; (4) Draw the isosurface image: draw the isosurface image according to the normal vector of the vertex of the triangular facet and the coordinate of the vertex. Among them, the linear difference method is the first preset method, and the central difference method is the second preset method.

[0117] In this embodiment, the structure diagram of the volume data space coordinate system is as follows: Figure 4 As shown, the image data corresponding to the X-axis and Y-axis represent the length and width of the acquired image, respectively, and the Z-axis represents the layer thickness direction. The interlayer distance can be obtained from the information of the original image, thus ensuring that each voxel corresponds to the corresponding image grayscale value.

[0118] Example 7:

[0119] Based on Example 1, this example provides an intelligent measurement and assessment system for uterine fibroids based on 3D U-Net, including a growth prediction module:

[0120] The diagnostic result output unit is used to visualize the three-dimensional reconstruction results, determine the image features corresponding to the three-dimensional reconstruction results, and output the diagnostic results of uterine fibroids based on the image features.

[0121] The growth prediction unit is used to pre-train the fibroid growth rate prediction model, input the 3D reconstruction results into the fibroid growth rate prediction model to predict the growth of uterine fibroids, and output the growth prediction results.

[0122] In this embodiment, the diagnostic result of uterine fibroids can be output through prediction. The corresponding diagnostic result can be output according to the image features of different situations. For example, when smooth muscle cells are replaced by tissue fibers and the cross-sectional whorl structure disappears, the output result is hyalinization of uterine fibroids.

[0123] In this embodiment, uterine fibroids can be visualized in a three-dimensional form (the result of 3D reconstruction is presented in the form of a 3D volumetric image) to help doctors observe the pathological condition more intuitively. Furthermore, by adjusting visualization parameters such as transparency, color mapping, and lighting, a clear and discernible volumetric image effect can be obtained.

[0124] In this embodiment, the pre-trained fibroid growth rate prediction model includes: training a model that can predict the growth rate of fibroids by collecting a large amount of clinical data from fibroid patients, including tumor size, growth rate, patient age, hormone levels, etc.

[0125] In this embodiment, the treatment effect can be monitored. For example, by using the 3D U-Net network model regularly, doctors can observe the trend of changes in fibroid volume and thus judge the effectiveness of the drug. If the growth of fibroid volume slows down or shrinks, it may indicate that the treatment is effective. Conversely, if the fibroid volume continues to grow rapidly, it may be necessary to adjust the treatment plan. Therefore, the fibroid growth rate assessment model not only provides strong support for clinical decision-making, but also makes personalized medicine possible.

[0126] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A 3D U-Net-based intelligent measurement and evaluation system for uterine fibroids, characterized in that, include: The intelligent recognition module is used to construct a 3D U-Net network model and input real-time images into the 3D U-Net network model to identify fibroid boundaries and determine whether a woman has uterine fibroids. The 3D reconstruction module is used to perform 3D reconstruction on real-time images when the diagnosis is uterine fibroids. The growth prediction module is used to predict the growth of uterine fibroids based on the 3D reconstruction results and output the growth prediction results.

2. The intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net according to claim 1, characterized in that, The intelligent recognition module includes: Model building unit, used to build 3D U-Net network models; The analysis unit is used to acquire real-time images and input them into a 3D U-Net network model for fibroid boundary recognition to determine whether a woman has uterine fibroids.

3. The intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net according to claim 2, characterized in that, Model building units include: The image acquisition subunit is used to acquire a first sample image dataset of uterine fibroids based on a preset ultrasound device. The data preprocessing subunit is used to preprocess the first sample image dataset to obtain the second sample image dataset. The data label addition sub-unit is used to add image data labels to the second sample image dataset based on the image features of the sample images, wherein the image data labels correspond one-to-one with the second sample images; The sub-unit is used to divide the image data labels and the corresponding second sample image set into training set, validation set and test set according to a preset ratio; The fibroid region extraction subunit is used to extract fibroid regions from the training set, validation set, and test set respectively based on image segmentation techniques. The training subunit is used to obtain the initial 3D U-Net network model and train the initial 3D U-Net network model according to the fibroid regions corresponding to the training set, validation set and test set to obtain the 3DU-Net network model.

4. The intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net according to claim 3, characterized in that, In the data preprocessing subunit, data preprocessing is performed on the first sample image dataset, including histogram equalization, data augmentation, and normalization.

5. The intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net according to claim 3, characterized in that, The preset ratio for dividing the sub-units is 8:1:

1.

6. The intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net according to claim 1, characterized in that, In the intelligent recognition module, The 3D U-Net network model includes: an encoder and a decoder; The encoder is composed of multiple alternating three-dimensional convolutional layers and pooling layers; The decoder consists of multiple upper sampling layers and three-dimensional convolutional layers.

7. The intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net according to claim 3, characterized in that, After obtaining the 3D U-Net network model, the training sub-unit also includes: The network improvement subunit is used to improve the network structure of the 3D U-Net network model based on a full-scale skip connection method. The loss optimization and improvement subunit is used to obtain the target loss function by fusing the Dice loss function and the binary cross-entropy loss function. At the same time, the target loss function is used to optimize and train the 3D U-Net network model after network structure improvement to obtain the target 3D U-Net network model.

8. The intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net according to claim 7, characterized in that, The loss optimization improvement subunit includes: The Dice loss function formula is as follows: Where DiceLoss represents the Dice loss function; X represents the image segmentation sample; Y represents the image segmentation result sample; The formula for the binary cross-entropy loss function is as follows: Where BCELoss represents the binary cross-entropy loss function; N represents the number of groups of objects predicted by the 3D U-Net network model; i represents the group index value of the predicted object; y i p(y) represents the true result value of the i-th predicted object; i ) represents the predicted result value of the i-th group of predicted objects; lg(·) represents the logarithmic function with base 10; The Dice loss function and the binary cross-entropy loss function are fused according to the following formula to obtain the target loss function; Loss=αDiceLoss+βBCELoss; Where Loss represents the objective loss function; α and β represent hyperparameters.

9. The intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net according to claim 1, characterized in that, The 3D reconstruction module includes: The spatial coordinate system construction unit is used to acquire real-time image sequences and construct a spatial coordinate system when the judgment result is that the patient has uterine fibroids. At the same time, the image data of the acquired real-time image sequence is read into the spatial coordinate system in sequence. 3D reconstruction unit, used for: Based on the input results, the target cube is determined in the spatial coordinate system with every 8 adjacent coordinate points as vertices, and the target cube is taken as a voxel, with the 8 adjacent coordinate points as the 8 corner points of the voxel. The edges of the boundary voxels and the corners of the isosurfaces are calculated according to the first preset method; The normal vectors of each corner point of the voxel are calculated based on the edges of the boundary voxel and the corner points of the isosurface, and the triangular facets are determined based on the normal vectors. At the same time, the normal vectors of the vertices of the triangular facets are calculated based on the first preset method. Based on the normal vectors of the vertices of the triangular facets and their coordinates, an isosurface image is drawn to complete the 3D reconstruction of the real-time image.

10. The intelligent measurement and evaluation system for uterine fibroids based on 3D U-Net according to claim 1, characterized in that, The growth prediction module includes: The diagnostic result output unit is used to visualize the three-dimensional reconstruction results, determine the image features corresponding to the three-dimensional reconstruction results, and output the diagnostic results of uterine fibroids based on the image features. The growth prediction unit is used to pre-train the fibroid growth rate prediction model, input the 3D reconstruction results into the fibroid growth rate prediction model to predict the growth of uterine fibroids, and output the growth prediction results.

Images