Embryo development stage classification method and system based on residual perception and wavelet coding

Abstract

The invention provides an embryonic development stage classification method and system based on residual perception and wavelet coding, and belongs to the field of embryonic development stage classification. The method comprises the following steps: acquiring and preprocessing an embryo image; carrying out blocking operation on the pre-processed embryo image, inputting blocked image data into a residual error perception attention module, and extracting a first feature map of the embryo image; inputting the first feature map into a wavelet position coding module, and extracting a second feature map containing frequency domain information; adding the first feature map and the second feature map, inputting the added first feature map and second feature map into a trained deep neural network, and performing global feature extraction and coding to obtain a final feature map; and inputting the final feature map into a classifier for developmental stage classification, and outputting a classification result of the embryonic developmental stage. The problem of spatial information loss in a down-sampling method is effectively solved, the sensitivity of the model to frequency domain information is enhanced, and the classification accuracy of the embryonic development stage is improved.

Description

Technical Field

[0001] The present invention belongs to the field of embryonic development stage classification, and in particular relates to an embryonic development stage classification method and system based on residual perception and wavelet coding. Background Art

[0002] The statements in this section merely provide background information related to the present invention and do not necessarily constitute prior art.

[0003] Infertility is a reproductive health issue affecting millions of couples worldwide. Currently, in vitro fertilization (IVF) has become one of the most effective assisted reproductive technologies (ART) for addressing infertility. This technique combines eggs and sperm in the laboratory to form embryos, which are then transferred back into the uterus. During ART, the health and viability of the embryos directly impact implantation success rates and pregnancy stability. Embryonic development is a complex and delicate process divided into 16 sequential stages. This detailed classification system enables embryologists to more accurately assess embryonic development, thereby selecting embryos with the greatest potential for implantation and ensuring the success of ART.

[0004] In recent years, with the rapid development of deep learning neural networks, research on applying deep learning technology to the field of assisted reproduction has begun to attract attention from the academic community. For example, convolutional neural networks (CNNs) and Transformers have made their mark in the field of assisted reproductive technology with their superior feature extraction capabilities and modeling flexibility. However, these CNN-based models primarily focus on local feature extraction and may ignore global contextual information related to the overall morphology, structure, and tissue characteristics of the embryo, which is crucial for accurately identifying and classifying embryos at different developmental stages. Related Transformer-based methods may lose key spatial information during spatial downsampling (i.e., when dividing the image into fixed-size convolutional blocks for processing). This information plays a decisive role in distinguishing embryo categories, and its loss weakens the model's discriminative ability in embryo image classification.

[0005] Furthermore, current deep learning-based embryo classification methods primarily rely on spatial features and may fail to fully utilize valuable information in the frequency domain. This can cause the deep learning model to miss subtle morphological changes and developmental features in embryo images, reducing the effectiveness of distinguishing different embryonic developmental stages and ultimately affecting overall classification accuracy. Summary of the Invention

[0006] To overcome the deficiencies of the above-mentioned prior art, the present invention provides a method and system for classifying embryonic development stages based on residual perception and wavelet coding. Based on the Transformer deep neural network, residual perception attention and wavelet position coding are combined for the classification of embryonic development stages. This method can not only effectively reduce the loss of important features in the Transformer downsampling process and maintain the integrity of information as much as possible, but also integrate the frequency domain information in the embryonic image, improve the model's ability to perceive the detailed information and differences of embryos at different developmental stages, and improve the accuracy of embryonic development stage classification.

[0007] To achieve the above objectives, one or more embodiments of the present invention provide the following technical solutions: The first aspect of the present invention provides an embryonic development stage classification method based on residual perception and wavelet coding; Embryonic development stage classification method based on residual perception and wavelet coding, including: Acquiring an embryo image, and preprocessing the embryo image to obtain a preprocessed embryo image; Performing a block operation on the preprocessed embryo image to obtain block-wise image data, inputting the block-wise image data into the residual perception attention module to extract the first feature map of the embryo image; Inputting the first feature map into a wavelet position encoding module to extract a second feature map containing frequency domain information; Adding the first feature map and the second feature map together and inputting the result into a trained deep neural network to perform global feature extraction and encoding to obtain a final feature map; The final feature map is input into a classifier for developmental stage classification, and the classification result of the embryonic developmental stage is output.

[0008] As a further technical solution, the process of preprocessing the embryo image to obtain a preprocessed embryo image includes: Perform scaling operations on the embryo image to reduce the pixel value; Data augmentation is performed on the scaled embryo image, and random rotation is used to obtain the scaled and enhanced embryo image.

[0009] As a further technical solution, the pre-processed embryo image is divided into blocks to obtain block image data, and the block image data is input into the residual perception attention module to extract the first feature map of the embryo image, including: Downsampling the preprocessed embryo image to obtain a downsampled feature map; The downsampled feature maps are input into the channel attention branch and spatial attention branch of the residual perception attention module respectively to obtain channel attention features and spatial attention features; The channel attention feature and the spatial attention feature are multiplied and fused with the block-wise image data through a jump connection to obtain a first feature map of the embryo image.

[0010] As a further technical solution, the process of obtaining spatial attention features using the spatial attention branch includes: The downsampled feature map is subjected to mean operation and reshape operation in sequence to obtain the reshaped feature map; Normalize and standardize each sample of the reshaped feature map to obtain a feature map with zero mean and unit variance; Multiply the feature map with zero mean and unit variance by the learnable weight parameter, add the bias term, and obtain the spatial attention mapping coefficient through the Sigmoid function; The spatial attention mapping coefficient is multiplied element-by-element with the block-by-block image data to obtain the spatial attention feature.

[0011] As a further technical solution, the first feature map is input into a wavelet position encoding module to extract a second feature map containing frequency domain information, including: The first feature map is processed by the frequency information extraction branch of the wavelet position coding module to obtain a first branch feature map that integrates frequency domain information; the first feature map is processed by the feature enhancement convolution branch to obtain a second branch feature map that enhances detail features; the first feature map is processed by the detail preservation convolution branch to obtain a third branch feature map that retains the original detail information; Fusing the first branch feature map and the second branch feature map by element-by-element multiplication to obtain a fused feature map; The fused feature map is fused with the third branch feature map by element-by-element addition to obtain a second feature map containing frequency domain information.

[0012] As a further technical solution, the first feature map and the second feature map are added and input into a trained deep neural network to perform global feature extraction and encoding to obtain a final feature map, including: The trained deep neural network uses a Transformer neural network to normalize the result of adding the first feature map and the second feature map; The normalized result is input into the multi-head self-attention mechanism, the attention score is calculated, and the output of the multi-head self-attention mechanism is obtained. The output of the multi-head self-attention mechanism is a weighted feature representation that integrates the attention information of different attention heads; The weighted feature representation is nonlinearly transformed and modeled using a feedforward neural network to obtain the final feature map.

[0013] As a further technical solution, the final feature map is input into a classifier for developmental stage classification, and the classification result of the embryonic developmental stage is output, including: The final feature map is linearly transformed through the fully connected layer, the features are mapped to the output space, and the output layer is used to output the classification results of the embryonic development stage.

[0014] The second aspect of the present invention provides an embryonic development stage classification system based on residual perception and wavelet coding.

[0015] Embryonic development stage classification system based on residual perception and wavelet coding, including: The image preprocessing module is configured to: acquire an embryo image, preprocess the embryo image, and obtain a preprocessed embryo image; The feature map extraction module is configured to: perform a block operation on the preprocessed embryo image to obtain block image data, input the block image data into the residual perception attention module to extract a first feature map of the embryo image; input the first feature map into the wavelet position coding module to extract a second feature map containing frequency domain information; A global feature extraction module is configured to: add the first feature map and the second feature map and input the result into a trained deep neural network to perform global feature extraction and encoding to obtain a final feature map; The stage classification module is configured to: input the final feature map into a classifier to perform developmental stage classification, and output the classification result of the embryonic development stage.

[0016] A third aspect of the present invention provides a computer-readable storage medium having a program stored thereon, which, when executed by a processor, implements the steps of the embryonic development stage classification method based on residual perception and wavelet coding as described in the first aspect of the present invention.

[0017] The fourth aspect of the present invention provides an electronic device, comprising a memory, a processor, and a program stored in the memory and runnable on the processor, wherein when the processor executes the program, the steps in the method for classifying embryonic development stages based on residual perception and wavelet coding as described in the first aspect of the present invention are implemented.

[0018] One or more of the above technical solutions have the following beneficial effects: (1) The residual perception attention module in the present invention can minimize the information loss caused by resolution reduction during the downsampling process and efficiently retain the key feature information in the embryo image; the wavelet position encoding module can enhance the model's ability to perceive the local subtle structures and overall morphological features in the embryo image; the encoder in the Transformer module extracts global context features from the input feature map through multi-head self-attention and feedforward neural network. By combining the advantages of these three modules for the classification of embryonic development stages, it can effectively overcome the problem of spatial information loss in the downsampling method and enhance the model's sensitivity to frequency domain information.

[0019] (2) The residual-aware attention module in the present invention includes a channel attention branch, a spatial attention branch, and a skip connection branch. The channel attention branch enables the model to identify and enhance important features within each channel, effectively preventing information loss; the spatial attention branch helps the network focus on key areas in the embryo image, and the skip connection can directly transfer high-resolution features to subsequent network layers, thereby preserving key information.

[0020] (3) The wavelet position encoding module in the present invention includes a frequency information extraction branch, a feature enhancement convolution branch, and a detail preservation convolution branch. The frequency information extraction branch uses wavelet transform to extract low-frequency and high-frequency components from the embryo feature map; the feature enhancement convolution branch is used to enhance the complex detail features in the image; and the detail preservation convolution branch can preserve the original detail information of the input features.

[0021] Advantages of additional aspects of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings, which constitute a part of the present invention, are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute improper limitations on the present invention.

[0023] Figure 1 This is a flow chart of the method of the first embodiment.

[0024] Figure 2 This is a diagram of the overall architecture of the method adopted in the first embodiment.

[0025] Figure 3 This is a structural diagram of the residual perception attention module in the first embodiment.

[0026] Figure 4 This is a structural diagram of the wavelet position coding module in the first embodiment.

[0027] Figure 5This is a structural diagram of the deep neural network module in the first embodiment.

[0028] Figure 6 This is a structural diagram of the multi-head self-attention mechanism in the deep neural network module in the first embodiment.

[0029] Figure 7 This is a system structure diagram of the second embodiment. DETAILED DESCRIPTION

[0030] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which the present invention belongs.

[0031] It should be noted that the terms used herein are for describing particular embodiments only and are not intended to limit the exemplary embodiments according to the present invention.

[0032] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0033] Example 1 This embodiment discloses an embryonic development stage classification method based on residual perception and wavelet coding; Combine Figure 1 and Figure 2 , an embryonic development stage classification method based on residual perception and wavelet coding, including: Step S1, acquiring an embryo image, and preprocessing the embryo image to obtain a preprocessed embryo image; Step S2, performing a block operation on the pre-processed embryo image to obtain block-by-block image data, inputting the block-by-block image data into a residual perception attention module, and extracting a first feature map of the embryo image; Step S3, inputting the first feature map into a wavelet position coding module to extract a second feature map containing frequency domain information; Step S4, adding the first feature map and the second feature map and inputting the result into a trained deep neural network to perform global feature extraction and encoding to obtain a final feature map; Step S5: input the final feature map into a classifier for developmental stage classification, and output the embryonic developmental stage classification result.

[0034] Specifically, it also includes the following: In step S1, the input embryo image is preprocessed as follows: first, the embryo image is scaled from 500×500 pixels to 224×224 pixels to adapt to the network architecture; then, the scaled embryo image is subjected to data augmentation and randomly rotated to obtain the scaled and enhanced embryo image EM.

[0035] In step S2, the pre-processed embryo image EM is processed by convolution and block processing to obtain the block image data The divided embryo image data is input into the residual perception attention module, and the first feature map of the embryo image is output.

[0036] Among them, combined Figure 3 The residual-aware attention module consists of three branches: the first branch, the channel attention branch, which enables the model to identify and enhance important features within each channel; the second branch, the spatial attention branch, which helps the network focus on key areas in the image and retain key spatial information even when the resolution is reduced; and the third branch, the skip connection branch, which directly passes high-resolution features to subsequent network layers without downsampling, thereby effectively retaining key information and reducing the information loss caused by resolution reduction.

[0037] Specifically, the first branch and the second branch are connected in parallel, and the input image data after the block After downsampling, the downsampled feature map is obtained , and input them into the first and second branches for processing respectively. The first branch includes adaptive average pooling, reshaping, normalization, standardization, learnable weight multiplication, bias addition, and Sigmoid activation function. The second branch includes averaging, reshaping, normalization, standardization, learnable weight multiplication, bias addition, Sigmoid activation function, and jump connection with the third branch for element-by-element multiplication of the input signal.

[0038] Furthermore, the output features of the first branch are multiplied by the output features of the second branch and added and fused with the input features of the third branch to obtain a first feature map of the embryo image.

[0039] In the channel attention branch, the downsampled feature map First, after the adaptive average pooling layer, global averaging is performed on the channel dimension to obtain the feature map The specific expression of the calculation process of adaptive average pooling is:

[0040] Among them, H and W represent feature maps respectively. The height and width of the Representation feature map The element at row j and column k in .

[0041] Feature Map Perform reshape operation to get , where B is the batch size and C is the number of channels of the input feature. Subsequently, the data of each channel is normalized and standardized to obtain a feature map with zero mean and unit variance. and To ensure the distribution consistency and stability of channel data. Next, the standardized channel features are multiplied by the learnable weights and the bias term is added. Then, the Sigmoid activation function is used to perform nonlinear transformation to obtain the channel attention features of this branch. .

[0042] In the spatial attention branch, the downsampled feature map First, perform the mean operation along the channel dimension to obtain the feature map , to reduce the spatial size of the data. The specific expression for the mean operation calculation is:

[0043] in, Represents the feature map during downsampling The element of the nth channel.

[0044] Reshape the feature map Reorganize and change its dimension to get Then, each sample is normalized and standardized to obtain a feature map with zero mean and unit variance. and , to ensure the consistency and stability of the feature distribution. Then, the processed features are multiplied by the learnable weight parameters and the bias term is added. Subsequently, they are processed by the Sigmoid function to obtain the spatial attention map coefficients. These coefficients reflect the importance and influence of each spatial position in the entire feature map, thereby helping the model to better utilize the spatial structure information of the input data. Finally, the spatial attention map coefficients are combined with the image data after segmentation. Perform element-by-element multiplication to obtain the spatial attention features of the spatial attention branch , where B, H, and W represent the batch normalization size, height, and width of the input features, respectively.

[0045] In the jump connection branch, the channel attention feature and the spatial attention feature are multiplied and then connected again with the block image data through the jump connection. Fusion is performed to obtain the first feature map of the embryo image This jump connection preserves the channel and spatial features while maintaining the integrity of the original input features, ensuring the effective transmission of information.

[0046] In step S3, the first feature map extracted by the residual perception attention module is , which is input into the wavelet position encoding module to capture subtle structural changes in the embryonic feature map.

[0047] Among them, such as Figure 4 As shown in the figure, the wavelet position encoding module also includes the first branch, the second branch, and the third branch. The first branch is the frequency information extraction branch, which can provide structural features at different scales, enabling the model to more comprehensively analyze embryo images; the second branch is the feature enhancement convolution branch, which enables the model to enhance its ability to extract complex information in the image; the third branch module is the detail preservation convolution branch, which can retain the original details of the input features and avoid the detail loss caused by downsampling, nonlinear activation or multi-layer convolution.

[0048] In the frequency information extraction branch, the first feature map of the input Perform reshape operation to make it become , then a discrete wavelet transform (DWT) is applied to it for frequency domain decomposition to obtain four wavelet sub-bands. In this embodiment, Haar wavelet is used for DWT transformation.

[0049] Specifically, first, a low-pass filter and a high-pass filter are used along the row direction to filter Decomposed into two subbands and , and then the two sub-bands are and Continue to apply low-frequency and high-frequency filtering to obtain four wavelet sub-bands. The four wavelet sub-bands are low-frequency-low-frequency signals that capture the overall structure. and low-frequency-high-frequency signals of large-scale information , and high-low frequency signals that capture fine-grained details and high-frequency-high-frequency signals of edge information Subsequently, these subbands are concatenated along the channel dimension to form a new feature map Then, Apply 3×3 two-dimensional convolution to enhance the feature expression ability, and get In order to effectively reintegrate the frequency domain information into the feature map, Perform inverse discrete wavelet transform (iDWT) to obtain a feature map that combines spatial and frequency domain information Finally, the feature map that combines spatial and frequency domain information is Perform the reshape operation to obtain the first branch feature map of the fused frequency domain information , to ensure that it is suitable for processing in the subsequent one-dimensional convolutional layer.

[0050] In the feature enhancement convolution branch and detail preservation convolution branch, the first feature map of the input is first Perform a flattening operation.

[0051] The feature enhancement convolution branch consists of flattening operation, one-dimensional convolution, BN and ReLU activation function, which are used to enhance the complex detail features in the image. 、 and And output the second branch feature map . Then, the second branch feature map With the first branch feature map Perform element-by-element multiplication to obtain the fused feature map .

[0052] The detail-preserving convolution branch consists of a flattening operation and a one-dimensional convolution with a single-layer kernel size of 1, which is used to preserve the original detail information of the input features and obtain the third branch feature map that preserves the original detail information. .

[0053] Furthermore, the fusion feature map And the third branch feature map The second feature map containing frequency domain information is obtained by fusion through element-by-element addition, and then one-dimensional convolution and reshape operation are performed. .

[0054] In step S4, the first feature map and the second feature map are added and input into a trained deep neural network to perform global feature extraction and encoding to obtain a final feature map.

[0055] like Figure 5 As shown in the figure, the trained deep neural network uses a Transformer neural network, specifically a Transformer-based global feature extraction encoding module (also called a Transformer encoder module). The encoder includes a multi-head self-attention mechanism and a feedforward neural network connected in sequence.

[0056] The feedforward neural network is composed of two fully connected layers separated by a GELU activation function and a dropout layer. Preferably, the fully connected layer, the GELU activation function, the dropout layer, the fully connected layer, and the dropout layer are sequentially connected to obtain a feedforward neural network.

[0057] Before the feature map obtained by adding the first feature map and the second feature map is input into the multi-head self-attention mechanism and the feedforward neural network, the input feature map is normalized.

[0058] like Figure 6 As shown in the figure, the specific steps of the feature map obtained after the addition of the input in the encoder are as follows: In the multi-head self-attention mechanism, the input feature map is first linearly transformed to generate three vectors Q, K, and V. Then the Q, K, and V vectors are equally divided into h parts along the channel dimension, respectively. , where h is the number of attention heads in the multi-head self-attention mechanism.

[0059] Then to and Perform element-wise multiplication, scaling, and softmax operations to calculate the attention score and sum the attention score Aggregation is performed to obtain the feature map of the i-th attention head. Finally, the outputs of all attention heads are concatenated and processed through a linear layer to obtain the output of the multi-head self-attention mechanism, which is a weighted feature representation that integrates the information of different attention heads.

[0060] The feedforward neural network further extracts features by performing nonlinear transformation and modeling on the weighted feature representation of the multi-head self-attention mechanism. After the feedforward neural network processing, a more abstract feature representation is finally obtained, resulting in the final feature map.

[0061] In step S5, the final feature map is input into a classifier for developmental stage classification, and the classification result of the embryonic development stage is output.

[0062] In this embodiment, the final feature map is linearly transformed through a series of fully connected layers to map the features to the output space, and the output layer is used to output the predicted embryonic development stage. The above-mentioned residual perception attention module and wavelet position encoding module can overcome the problem of spatial information loss in the downsampling method and enhance the model's sensitivity to frequency domain information, thereby effectively improving the model's classification performance in the embryonic development stage.

[0063] To verify the effectiveness of the present invention, the present invention used a public embryo image dataset for experiments. As shown in Table 1, the results show that the classification accuracy of the method disclosed in the present invention is 71.0%. Compared with the classification accuracy of 66.7% of the Transformer model alone, the classification effect of the method disclosed in the present invention is better and can effectively improve the accuracy of embryonic development stage classification.

[0064] Table 1 Performance comparison between the method of the present invention and the Transformer model

[0065] The calculation formulas for accuracy (ACC), precision (P), recall (R), and F1 score are as follows: Accuracy = (TP + TN) / (TP + TN + FP + FN) Precision = TP / (TP+FP) Recall = TP / (TP+FN) F1 value = (2 × precision × recall) / (precision + recall) Among them, TP, FN, FP, and TN represent true positive examples, false negative examples, false positive examples, and true negative examples, respectively.

[0066] Example 2 This embodiment discloses an embryonic development stage classification system based on residual perception and wavelet coding; like Figure 7 As shown in FIG, the embryonic development stage classification system based on residual perception and wavelet coding includes: The image preprocessing module is configured to: acquire an embryo image, preprocess the embryo image, and obtain a preprocessed embryo image; The feature map extraction module is configured to: perform a block operation on the preprocessed embryo image to obtain block image data, input the block image data into the residual perception attention module to extract a first feature map of the embryo image; input the first feature map into the wavelet position coding module to extract a second feature map containing frequency domain information; A global feature extraction module is configured to: add the first feature map and the second feature map and input the result into a trained deep neural network to perform global feature extraction and encoding to obtain a final feature map; The stage classification module is configured to: input the final feature map into a classifier to perform developmental stage classification, and output the embryonic developmental stage classification result.

[0067] Example 3 The purpose of this embodiment is to provide a computer-readable storage medium.

[0068] A computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the embryonic development stage classification method based on residual perception and wavelet coding as described in Example 1.

[0069] Example 4 The purpose of this embodiment is to provide an electronic device.

[0070] An electronic device comprises a memory, a processor and a program stored in the memory and executable on the processor, wherein when the processor executes the program, the steps of the embryonic development stage classification method based on residual perception and wavelet coding as described in Example 1 are implemented.

[0071] The steps involved in the apparatuses of Examples 2, 3, and 4 above correspond to those of Method Example 1. For detailed implementations, please refer to the relevant description of Example 1. The term "computer-readable storage medium" should be understood to mean a single medium or multiple media containing one or more instruction sets; it should also be understood to include any medium capable of storing, encoding, or carrying an instruction set for execution by a processor and causing the processor to perform any method of the present invention.

[0072] Those skilled in the art will appreciate that the modules or steps of the present invention described above can be implemented using a general-purpose computer device. Alternatively, they can be implemented using program code executable by a computing device, which can then be stored in a storage device and executed by the computing device. Alternatively, they can be fabricated into separate integrated circuit modules, or multiple modules or steps can be fabricated into a single integrated circuit module for implementation. The present invention is not limited to any specific combination of hardware and software.

[0073] Although the above describes the specific embodiments of the present invention in conjunction with the accompanying drawings, it is not intended to limit the scope of protection of the present invention. Those skilled in the art should understand that various modifications or variations that can be made by those skilled in the art on the basis of the technical solution of the present invention without any creative work are still within the scope of protection of the present invention.

Claims

1. The embryonic development stage classification method based on residual perception and wavelet coding is characterized by: include: Acquiring an embryo image, and preprocessing the embryo image to obtain a preprocessed embryo image; Performing a block operation on the preprocessed embryo image to obtain block-wise image data, inputting the block-wise image data into the residual perception attention module to extract the first feature map of the embryo image; Inputting the first feature map into a wavelet position encoding module to extract a second feature map containing frequency domain information; Adding the first feature map and the second feature map together and inputting the result into a trained deep neural network to perform global feature extraction and encoding to obtain a final feature map; The final feature map is input into a classifier for developmental stage classification, and the classification result of the embryonic development stage is output.

2. The embryonic development stage classification method based on residual perception and wavelet coding as claimed in claim 1, characterized in that: The process of preprocessing the embryo image to obtain a preprocessed embryo image includes: Perform scaling operations on the embryo image to reduce the pixel value; Data augmentation is performed on the scaled embryo image, and random rotation is used to obtain the scaled and enhanced embryo image.

3. The embryonic development stage classification method based on residual perception and wavelet coding as claimed in claim 1, characterized in that: The pre-processed embryo image is divided into blocks to obtain block image data, and the block image data is input into the residual perception attention module to extract the first feature map of the embryo image, including: Downsampling the preprocessed embryo image to obtain a downsampled feature map; The downsampled feature maps are input into the channel attention branch and spatial attention branch of the residual perception attention module respectively to obtain channel attention features and spatial attention features; The channel attention feature and the spatial attention feature are multiplied and fused with the block-wise image data through a jump connection to obtain a first feature map of the embryo image.

4. The embryonic development stage classification method based on residual perception and wavelet coding as claimed in claim 3, characterized in that: The process of obtaining spatial attention features using the spatial attention branch includes: The downsampled feature map is subjected to mean operation and reshape operation in sequence to obtain the reshaped feature map; Normalize and standardize each sample of the reshaped feature map to obtain a feature map with zero mean and unit variance; Multiply the feature map with zero mean and unit variance by the learnable weight parameter, add the bias term, and obtain the spatial attention mapping coefficient through the Sigmoid function; The spatial attention mapping coefficient is multiplied element-by-element with the block-by-block image data to obtain the spatial attention feature.

5. The embryonic development stage classification method based on residual perception and wavelet coding as claimed in claim 1, characterized in that: Inputting the first feature map into a wavelet position encoding module to extract a second feature map containing frequency domain information includes: The first feature map is processed by the frequency information extraction branch of the wavelet position coding module to obtain a first branch feature map that integrates frequency domain information; the first feature map is processed by the feature enhancement convolution branch to obtain a second branch feature map that enhances detail features; the first feature map is processed by the detail preservation convolution branch to obtain a third branch feature map that retains the original detail information; Fusing the first branch feature map and the second branch feature map by element-by-element multiplication to obtain a fused feature map; The fused feature map is fused with the third branch feature map by element-by-element addition to obtain a second feature map containing frequency domain information.

6. The embryonic development stage classification method based on residual perception and wavelet coding as claimed in claim 1, characterized in that: The first feature map and the second feature map are added and input into the trained deep neural network to perform global feature extraction and encoding to obtain the final feature map, including: The trained deep neural network uses a Transformer neural network to normalize the result of adding the first feature map and the second feature map; The normalized result is input into the multi-head self-attention mechanism, the attention score is calculated, and the output of the multi-head self-attention mechanism is obtained. The output of the multi-head self-attention mechanism is a weighted feature representation that integrates the attention information of different attention heads; The weighted feature representation is nonlinearly transformed and modeled using a feedforward neural network to obtain the final feature map.

7. The embryonic development stage classification method based on residual perception and wavelet coding as claimed in claim 1, characterized in that: The final feature map is input into a classifier for developmental stage classification, and the classification result of the embryonic developmental stage is output, including: The final feature map is linearly transformed through the fully connected layer, the features are mapped to the output space, and the output layer is used to output the classification results of the embryonic development stage.

8. Embryonic development stage classification system based on residual perception and wavelet coding, characterized by: include: The image preprocessing module is configured to: acquire an embryo image, preprocess the embryo image, and obtain a preprocessed embryo image; The feature map extraction module is configured to: perform a block operation on the preprocessed embryo image to obtain block image data, input the block image data into the residual perception attention module to extract a first feature map of the embryo image; input the first feature map into the wavelet position coding module to extract a second feature map containing frequency domain information; A global feature extraction module is configured to: add the first feature map and the second feature map and input the result into a trained deep neural network to perform global feature extraction and encoding to obtain a final feature map; The stage classification module is configured to: input the final feature map into a classifier to perform developmental stage classification, and output the classification result of the embryonic development stage.

9. A computer-readable storage medium having a program stored thereon, characterized in that: When the program is executed by a processor, the steps of the embryonic development stage classification method based on residual perception and wavelet coding are implemented.

10. An electronic device comprising a memory, a processor, and a program stored in the memory and executable on the processor, wherein: When the processor executes the program, the steps of the embryonic development stage classification method based on residual perception and wavelet coding are implemented as described in any one of claims 1 to 7.

Images