CN116205844A - A Fully Automatic Cardiac MRI Segmentation Method Based on Expanded Residual Networks - Google Patents

Abstract

The invention discloses a fully automatic heart magnetic resonance imaging segmentation method based on an expanded residual network. This includes: acquiring a cardiac magnetic resonance image; inputting the cardiac magnetic resonance image into a trained segmentation network to segment the right ventricle area, myocardial area and left ventricle area, wherein the segmentation network is constructed based on the residual network U-Net , the bottleneck layer of the residual network uses an expansion convolution block with a set expansion rate to combine the encoding path and the decoding path. The invention can accurately segment regions such as right ventricle, left ventricle and myocardium from cardiac magnetic resonance images, realize automatic segmentation of cardiac images, and improve the performance of cardiac region image segmentation.

Description

A Fully Automatic Cardiac MRI Segmentation Method Based on Dilated Residual Networks

technical field

The invention relates to the technical field of biomedical engineering, and more specifically, to a fully automatic cardiac magnetic resonance imaging segmentation method based on an expanded residual network.

Background technique

Heart diseases are a serious threat to human life. In order to effectively treat and prevent such diseases, accurate calculation, modeling and analysis of the entire heart structure are essential for research and application in the medical field. However, during CMRI (Magnetic Resonance Cine Imaging) acquisition, the beating heart makes it difficult to obtain clear images, especially for patients with cardiovascular disease, who are more likely to experience arrhythmia, difficulty in holding their breath, etc. As a result, images from MRI (magnetic resonance imaging) scanners may contain various image artifacts, making it difficult to assess image quality. Clinicians may draw wrong conclusions from imaging data if the imaging data is not properly segmented. The existing manual image segmentation is not only time-consuming, but also difficult to guarantee the accuracy. Therefore, it is necessary to realize the automatic segmentation of cardiac regions to solve practical problems in the field of cardiac medicine.

Cardiac image segmentation refers to the division of cardiac images into anatomically meaningful regions based on which quantitative measures such as myocardial mass, wall thickness, left ventricle (LV) and right ventricle (RV) volumes can be extracted. Therefore, it is particularly important to design an accurate automatic heart segmentation algorithm. In recent years, deep convolutional neural networks (DCNNs) have been shown to segment left and right ventricles and myocardium better than traditional computer vision methods. For example, the U-Net architecture is task-independent and has been applied to various biomedical segmentation tasks with only some minor or substantial modifications, and U-Net is the underlying network model for most of the most effective ventricular segmentation algorithms ( backbone).

In the prior art, patent application CN202210321078.8 provides a CT image heart segmentation method and system based on artificial intelligence semantic segmentation. This technology reduces the influence of noise by optimizing the class probability, and realizes accurate image segmentation. However, this solution does not involve the segmentation of the left ventricle, right ventricle, and myocardium in the heart, and cannot obtain independent heart tissue images.

Patent application CN202110391121.3 describes a cardiac segmentation model and pathological classification model training, cardiac segmentation, and pathological classification method and device based on cardiac MRI. This technology can greatly suppress background interference and promote rapid convergence of neural network training. However, for image Segmentation accuracy and robustness did not suggest improvements.

After analysis, the existing technology lacks research on the bottleneck layer of U-Net. Since the background area in the image is much larger than the mask, the pixel degradation and the loss of time and space information caused by the deepening of the network layer lead to The network has insufficient ability to extract sparse features of images.

Contents of the invention

The purpose of the present invention is to overcome the defects of the above-mentioned prior art, and to provide a fully automatic cardiac magnetic resonance imaging segmentation method based on an expanded residual network. The method includes the following steps:

Obtain cardiac magnetic resonance images;

The cardiac magnetic resonance image is input to the trained segmentation network to segment the right ventricle area, myocardial area and left ventricle area;

Wherein, the segmentation network is constructed based on the residual network U-Net, and the bottleneck layer of the residual network uses an expansion convolution block with a set expansion rate to combine the encoding path and the decoding path.

Compared with the prior art, the present invention has the advantage of proposing a fully automatic cardiac MRI (magnetic resonance imaging) segmentation method based on the expanded residual network, which can accurately segment the right ventricle, left ventricle, myocardium, etc. from cardiac MRI images region, to achieve fully automatic segmentation of heart images, and to improve the performance of heart region image segmentation.

Other features of the present invention and advantages thereof will become apparent from the following detailed description of exemplary embodiments of the present invention with reference to the accompanying drawings.

Description of drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention.

Fig. 1 is a flow chart of a fully automatic cardiac magnetic resonance imaging segmentation method based on an expanded residual network according to an embodiment of the present invention;

Fig. 2 is a schematic diagram of the process from raw magnetic resonance image data to image segmentation according to an embodiment of the present invention;

Fig. 3 is a U-Net-based automatic image segmentation architecture diagram according to one embodiment of the present invention;

Fig. 4 is a schematic diagram of an architecture of an expanded residual block according to an embodiment of the present invention;

Fig. 5 is a schematic diagram of the image segmentation results for the ACDC test data set according to one embodiment of the present invention;

In the figure, Conv-convolution; Norm-regularization; Maxpool-maximum pooling; UpConv-convolution; Deconvolution-deconvolution; Skip-connection-jump connection; Pixel-wise addition-pixel-by-pixel addition; Kernel - Nucleus; Concatenation - cascade; Stride - step length; End Systole - end systole; End Diastole - end diastole.

Detailed ways

Various exemplary embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be noted that the relative arrangements of components and steps, numerical expressions and numerical values set forth in these embodiments do not limit the scope of the present invention unless specifically stated otherwise.

The following description of at least one exemplary embodiment is merely illustrative in nature and in no way taken as limiting the invention, its application or uses.

Techniques, methods and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, such techniques, methods and devices should be considered part of the description.

In all examples shown and discussed herein, any specific values should be construed as exemplary only, and not as limitations. Therefore, other instances of the exemplary embodiment may have different values.

It should be noted that like numerals and letters denote like items in the following figures, therefore, once an item is defined in one figure, it does not require further discussion in subsequent figures.

The present invention develops a fully automatic segmentation method for segmenting right ventricle (RV), myocardium (MYO) and left ventricle (LV) by combining short-axis CMRI (cine magnetic resonance imaging) sequence images. This method captures multi-resolution features in U-Net by dilating convolutional residual network (DRN), which significantly increases spatial and temporal information and maintains localization accuracy. And, the output of each expansion path is summed pixel-by-pixel to improve the training response.

As shown in Figure 1 and Figure 2, the provided fully automatic cardiac magnetic resonance imaging segmentation method based on the expanded residual network includes the following steps:

Step S110, preprocessing the data set to obtain training samples.

Taking magnetic resonance cine imaging as an example, the size of the three-dimensional image is L×W×H, where L is the length of the image sequence, W is the width of the image, and H is the length of the image. In the data set, the label value of the image is set to four labels by means of mapping, namely: black background=0, RV=1, MYO=2, LV=3.

Considering that there are significant differences in the display space size H×W and the range of intensity distribution of MRI cine images. In one embodiment, training samples are obtained through a data preprocessing process, specifically, ACDC (Adverse Conditions Dataset with Correspondences) images are taken as an example. First, the input image is resampled. There is a voxel spacing problem in the ACDC dataset. Since the convolutional neural network cannot explain the voxel spacing, all images are resampled to the same voxel spacing of 1.52×1.52×6.35mm.

The data preprocessing process is to take into account that the voxel spacing directly affects the overall voxel size of the image, and also affects the amount of contextual information extracted by the convolutional neural network from the image patch. Furthermore, if the voxel spacing is increased substantially, the image size decreases to such an extent that details are lost, so it is necessary to ensure that there is a trade-off between the amount of contextual information contained in the network patch size and the amount of detail preserved in the image data to obtain the best results. best performance.

In one embodiment, for the training data, all images are resampled to a median of 256x256 pixels. Magnetic resonance images of the ACDC dataset were then acquired using the multislice magnetic resonance cine images. For example, extract 2D-MRI (magnetic resonance imaging) slices and their associated annotations for each patient. And perform normalization on a slice-by-slice basis for each timeframe.

Step S120, expanding training samples through data augmentation, and constructing a training set.

Due to the limited training data, the model cannot learn the desired invariant and robust features, resulting in overfitting. Therefore, various data augmentation techniques can be applied to the training data to expand the sample size. For example, basic image transformation techniques are employed, including random rotation, random elastic deformation, scaling, flipping, and gamma correction. When applied to original training images, this data augmentation technique can efficiently generate multiple views of the same image. By using a variety of data enhancement methods to expand the training samples, the problems of overfitting and class imbalance can be solved.

Step S130, constructing a segmentation network based on the expanded residual network.

In this article, the heart segmentation network based on the U-net network is used as an example to illustrate, as shown in Figure 3 and Figure 4, where Figure 4 corresponds to the expanded residual block architecture of Figure 3. From the input image to the final output, the segmentation network follows the overall architecture of the encoder-decoder throughout the segmentation process. For example, the shrinkage path is constructed using 5 encoding blocks; each block consists of 2 convolutional layers with 3×3 kernels and 2×2 max-pooling operations with a stride of 2. Initially, 32 convolution kernels are selected. After each max pooling operation, the convolution kernels will be increased, resulting in 320 convolution kernels in the bottleneck layer of U-Net. Similarly, the spatial dimension of feature maps is reduced by a factor of 2 through the downsampling operation. The rectified linear unit (ReLU) is replaced by a leaky linear rectification, and instance regularization is used instead of normalization (BN).

The encoding and decoding paths are combined at the bottleneck layer of U-Net by extending the residual network (DRN), which captures the global context and recovers the spatial and temporal information without compromising the resolution of the segmentation map. In addition, the extended residual network can effectively adjust the depth of the convolutional layer without degrading the network performance. For example, by using dilated convolutions with different dilation rates (d = 1, 3, and 5), the receptive field in the dilated residual network block is enlarged. Then, the previously generated features are concatenated with the current features via residual connections. After every 3×3 convolution in the DRN (Dilated Residual Network) block, a dropout operation with a forgetting rate of 0.5 is performed to prevent overfitting. Therefore, the extended residual network captures contextual image information, high spatial resolution and multi-texture features. The process of the decoding path is similar to that of the encoding path, however the order of operations is reversed. The U-Net architecture offers the advantage of reusing encoded feature maps from encoding blocks to their corresponding levels, where the spatial dimensions match. This can be achieved through channel-specific cascading. A 1×1 kernel projection operation is used at the last stage of the decoding path to align the output channel dimensions to the classified classes (left ventricle, myocardium, and right ventricle). Finally, all expansion path outputs are aggregated via upsampling and pixel-wise addition to enhance the training response.

Typically, natural images contain many objects whose identities and relative positions are important for understanding the scene. However, segmentation becomes more difficult when the target object is not spatially salient, for example, when the target object is small compared to the background. If the features of the target object are lost during downsampling, they cannot be easily recovered during training. But if highly (substantial) spatial and temporal information is maintained throughout the network and provided output features that densely cover input features, backpropagation can learn important features from smaller and less prominent objects. Therefore, the present invention uses an expanded convolutional network to predict small and dense image features by increasing the receptive field and extracting more spatial information. The discrete dilated convolution is as follows:

是输入和输出离散函数，k为大小为(2d+1)²的离散核，*_l为扩张卷积，在求和过程中，需满足s+lt＝p，s表示扩张步幅，l表示缩放因子，p表示感受野，t表示整数序列，即t＝1,2,3...n。in,

Is the input and output discrete function, k is the discrete kernel of size (2d+1) ² , * _l is the expansion convolution, in the summation process, it needs to satisfy s+lt=p, s represents the expansion step, l represents Scaling factor, p represents the receptive field, t represents the sequence of integers, that is, t=1,2,3...n.

A dilated residual network can better expand the receptive field to achieve a promising result and avoid image information loss at the bottleneck of U-Net. Dilated convolution introduces a new parameter called "dilation rate" to the convolutional layer, which defines the distance between values when the convolution kernel processes data, and expands the receptive field by adding holes. Dilated convolutional layers are based on regular convolutions with dilation factors (d=1, 3 and 5). For example, choose a 1×1 kernel for a normal convolutional layer and a 3×3 kernel for a dilated convolution.

Among them, y _ij represents the expanded convolution whose input is x _ij , which is a convolution kernel with a length of M and a width of N, and m and n are the input variables of the expanded convolution. w(i, j) is the corresponding weight, i represents the image length index, j represents the image width index, and d represents the dilation rate.

Step S140, using the set loss function to train the segmentation network.

The purpose of segmentation is to detect target objects and draw contours around them. The automatically segmented contour Cp (predicted) is compared with the corresponding annotated image to measure the accuracy of the proposed method. In this paper, the pixels surrounded by the outline are referred to as A _p and A _g .

In segmentation network training, a variety of loss functions can be employed. For example, dice similarity coefficient or Hausdorff distance or other loss function types.

For example, for the Dice Similarity Coefficient (DSC), the ratio between the predicted profile and the ground truth profile represents the DSC score, usually expressed as a percentage between 0 and 1. A high dice value indicates a good match.

Among them, A _p represents the pixels surrounded by the predicted contour, and A _g represents the pixels surrounded by the real contour.

Hausdorff distance (HD) is the symmetric distance between predicted and actual contours compared and provides spatial resolution of MR cine images. The lower the HD value, the better the segmentation matching performance.

Among them, C _p is the predicted automatic segmentation contour, C _g is the corresponding ground truth contour, d(i, j) represents the distance between the ground truth value and the predicted contour, i represents the pixel point of the predicted contour, and j represents the ground truth Contour pixels. Consider images with significant class imbalance between regions of interest (ROI) and background. To address this issue, different loss functions were tested, including dice loss and weighted cross-entropy loss.

In a preferred embodiment, the segmentation network is trained using a dual loss function comprising dice loss and cross-entropy loss. Specifically, the cross-entropy loss is defined as follows:

表示对应于输入x的预测类别c的激活函数值。例如，对于c所表示的类别，黑色背景＝0，RV＝1，MYO＝2，LV＝3。Among them, C represents the total number of categories; c represents the category indication, W=(w ₁ , w ₂ , w ₃ ... w _n ) is a series of learnable weights, w _n is the weight matrix of the nth layer; p(Y _i |X _i , W) represents the probability that a predicted pixel X _i is misclassified relative to the ground truth label pixel; Y(c, x) represents the target label corresponding to the input x;

Indicates the activation function value corresponding to the predicted category c of the input x. For example, for the class represented by c, black background=0, RV=1, MYO=2, LV=3.

The training of the model was carried out for 500 iterations in total. In each iteration of the training set, 250 images were randomly selected from the data set until all the image data were traversed. To improve generalization, slices are randomly cropped from the training images and the network is evaluated after each iteration on the validation set. For example, train a segmentation network using a multiclass variant of the dice loss of

Among them, u and v are the one-hot encoded vector output by the activation function softmax and the image segmentation label value corresponding to the category identifier, i represents the image length index, k represents the image width index; c∈C is the category identifier, that is, the left side of the heart Ventricle, right ventricle, myocardium and background; ε is a small constant. After each pass, the learning rate lr is recalculated according to the following formula. Finally, the best model was selected to evaluate the test set to ensure that the validation of RV (right ventricle), MYO (myocardium) and LV (left ventricle) achieved the highest DSC (Dice Similarity Coefficient). The network provides consistent and stable performance across all folds.

Among them, initial_learning_rate is the initial learning rate, currentepoch is the current number of iterations, and totalepoch is the total number of iterations.

Step S150 , using the trained segmentation network to identify regions such as right ventricle, myocardium, and left ventricle for the acquired target magnetic resonance image.

After completing the training of the segmentation network, the optimized model parameters can be obtained, and then the trained segmentation network can be used to accurately distinguish the regions of interest such as the right ventricle, myocardium and left ventricle, as well as the complete heart contour, and then based on these regions can be Extract quantitative measures such as myocardial mass, left and right ventricle volumes, etc.

In order to further verify the effect of the present invention, experimental tests have been carried out on cardiac magnetic resonance images of multiple patients. See Figure 5 for a schematic diagram of different slice positions of the heart. Experimental results show that the present invention obtains higher segmentation accuracy and segmentation speed of left ventricle, right ventricle and myocardium, and obtains an overall dice similarity coefficient of 0.92±0.02 and an average Hausdorff distance of 8.06±0.05mm. Moreover, the present invention improves the speed of image segmentation, for example, it takes 0.28 seconds on average to process a 2D magnetic resonance image. Furthermore, our network designed to predict individual MRI images to segment ventricular regions successfully achieved automatic segmentation of cardiac images.

In summary, compared with the prior art, the present invention has the following technical effects:

1) The present invention introduces an expanded convolutional residual network, and enhances the performance of the U-Net bottleneck layer to realize fully automatic and accurate segmentation of cardiac MRI (magnetic resonance imaging) images, which solves the limitation of the U-Net bottleneck layer, Spatial and temporal information is significantly enhanced, improving accuracy while maintaining spatial consistency.

2) The present invention designs the Extended Residual Network (DRN) block to replace the original bottleneck layer of U-Net. And a variety of loss functions are used to better use the cardiac features to train the model and improve the accuracy in the process of segmenting the cardiac image.

3) The present invention has higher calculation speed and robustness, and can be applied to diverse cardiac CMRI (magnetic resonance film imaging) data sets. For example, the processed data is the magnetic resonance images of the patient under two different magnetic intensities. After the data is processed, the patient's complete heart outline, images of the left and right ventricles and myocardial outline can be obtained at the same time.

The present invention can be a system, method and/or computer program product. A computer program product may include a computer readable storage medium having computer readable program instructions thereon for causing a processor to implement various aspects of the present invention.

A computer readable storage medium may be a tangible device that can retain and store instructions for use by an instruction execution device. A computer readable storage medium may be, for example, but is not limited to, an electrical storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of computer-readable storage media include: portable computer diskettes, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), or flash memory), static random access memory (SRAM), compact disc read only memory (CD-ROM), digital versatile disc (DVD), memory stick, floppy disk, mechanically encoded device, such as a printer with instructions stored thereon A hole card or a raised structure in a groove, and any suitable combination of the above. As used herein, computer-readable storage media are not to be construed as transient signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through waveguides or other transmission media (e.g., pulses of light through fiber optic cables), or transmitted electrical signals.

Computer-readable program instructions described herein may be downloaded from a computer-readable storage medium to a respective computing/processing device, or downloaded to an external computer or external storage device over a network, such as the Internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers, and/or edge servers. A network adapter card or a network interface in each computing/processing device receives computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in each computing/processing device .

Computer program instructions for carrying out operations of the present invention may be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine-related instructions, microcode, firmware instructions, state setting data, or Source or object code written in any combination, including object-oriented programming languages—such as Smalltalk, C++, Python, etc., and conventional procedural programming languages—such as the “C” language or similar programming languages. Computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer, or entirely on the remote computer or server implement. In cases involving a remote computer, the remote computer can be connected to the user computer through any kind of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computer (such as via the Internet using an Internet service provider). connect). In some embodiments, an electronic circuit, such as a programmable logic circuit, field programmable gate array (FPGA), or programmable logic array (PLA), can be customized by utilizing state information of computer-readable program instructions, which can Various aspects of the invention are implemented by executing computer readable program instructions.

Aspects of the present invention are described herein with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It should be understood that each block of the flowcharts and/or block diagrams, and combinations of blocks in the flowcharts and/or block diagrams, can be implemented by computer-readable program instructions.

These computer-readable program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine such that when executed by the processor of the computer or other programmable data processing apparatus , producing an apparatus for realizing the functions/actions specified in one or more blocks in the flowchart and/or block diagram. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause computers, programmable data processing devices and/or other devices to work in a specific way, so that the computer-readable medium storing instructions includes An article of manufacture comprising instructions for implementing various aspects of the functions/acts specified in one or more blocks in flowcharts and/or block diagrams.

It is also possible to load computer-readable program instructions into a computer, other programmable data processing device, or other equipment, so that a series of operational steps are performed on the computer, other programmable data processing device, or other equipment to produce a computer-implemented process , so that instructions executed on computers, other programmable data processing devices, or other devices implement the functions/actions specified in one or more blocks in the flowcharts and/or block diagrams.

The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, a portion of a program segment, or an instruction that includes one or more Executable instructions. In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks in succession may, in fact, be executed substantially concurrently, or they may sometimes be executed in the reverse order, depending upon the functionality involved. It should also be noted that each block of the block diagrams and/or flowchart illustrations, and combinations of blocks in the block diagrams and/or flowchart illustrations, can be implemented by a dedicated hardware-based system that performs the specified function or action , or may be implemented by a combination of dedicated hardware and computer instructions. It is well known to those skilled in the art that implementation by means of hardware, implementation by means of software, and implementation by a combination of software and hardware are all equivalent.

Having described various embodiments of the present invention, the foregoing description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and alterations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The terminology used herein is chosen to best explain the principle of each embodiment, practical application or technical improvement in the market, or to enable other ordinary skilled in the art to understand each embodiment disclosed herein. The scope of the invention is defined by the appended claims.

Claims (10)

1. A fully automatic cardiac magnetic resonance imaging segmentation method based on an expanded residual network, comprising the following steps: Obtain cardiac magnetic resonance images; The cardiac magnetic resonance image is input to the trained segmentation network to segment the right ventricle area, myocardial area and left ventricle area; Wherein, the segmentation network is constructed based on the residual network U-Net, and the bottleneck layer of the residual network uses an expansion convolution block with a set expansion rate to combine the encoding path and the decoding path. 2. The method according to claim 1, wherein the segmentation network is trained according to the following steps: Constructing a training set, the training set includes multiple pieces of sample data, each piece of sample data is a magnetic resonance image with a label category, and the label category is used to distinguish the right ventricular area, myocardial area and left ventricular area; performing image augmentation on the training set to generate multiple views of the same magnetic resonance image, the image augmentation comprising one or more of random rotation, random elastic deformation, scaling, flipping, and gamma correction; The training set after image enhancement is used to train the segmentation network with a set loss function to obtain optimized parameters. 3. The method according to claim 1, wherein training the segmentation network with a set loss function comprises: Within the range of the number of iterations set, the cross-entropy loss function is used to train the segmentation network, and each iteration extracts a set number of sample images from the training set; Randomly crop slices from the training images and evaluate the segmentation network after each iteration on the validation set, using a multiclass variant of the dice loss to train the segmentation network; After each pass, recalculate the learning rate; A segmentation network meeting the set performance requirements is selected as the trained segmentation network. 4. The method according to claim 3, wherein the cross-entropy loss is expressed as:

Among them, C represents the total number of label categories; c indicates the label category indication, W=(w ₁ , w ₂ , w ₃ ... w _n ) is a series of weights to be learned, w _n is the weight matrix of the nth layer, p (Y _i |X _i , W) represents the probability that a predicted pixel X _i is misclassified relative to the ground truth label pixel Y _i , and Y(c, x) represents the target label corresponding to the input x;

Indicates the activation function value corresponding to the predicted category c of the input x. 5. method according to claim 3, is characterized in that, described dice loss is expressed as:

Among them, u and v are the one-hot encoded vector output by the activation function softmax and the image segmentation label value corresponding to the annotation category indication c∈C, i represents the image length index, k represents the image width index, ε is a set constant, and C represents the label total number of categories. 6. The method according to claim 3, wherein the learning rate is updated according to the following formula:

Among them, initial_learning_rate is the initial learning rate, currentepoch is the current number of iterations, and totalepoch is the total number of iterations. 7. The method of claim 1, wherein the last stage of the decoding path of the segmentation network uses a 1×1 kernel projection operation. 8 . The method according to claim 1 , wherein the dilation rate is set to d=1, 3 or 5. 9 . 9. A computer-readable storage medium on which a computer program is stored, wherein when the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented. 10. A computer device comprising a memory and a processor, on which a computer program capable of running on the processor is stored, wherein claims 1 to 8 are realized when the processor executes the computer program The steps of any one of the methods.

Images