Cervical cell image analysis auxiliary method and system based on AI

Abstract

The invention relates to the technical field of medical image processing, in particular to an AI-based cervical cell image analysis auxiliary method and system, and the method comprises the steps: carrying out the multi-modal data collection, including cervical cell image data, clinical data and patient gene data, and carrying out the preprocessing of the collected cervical cell image data; constructing a cervix uteri cell image generative adversarial network to perform image enhancement and feature extraction on the cervix uteri cell image; fusing the extracted features with clinical data and patient gene data, and training a cervical cell classification model by using the fused feature data; and inputting a real cervical cell image into the trained cervical cell classification model, and outputting whether the cervical cells are abnormal or not and an analysis result of the abnormal type and degree. According to the method, multi-modal data acquisition is combined with an image enhancement and classification method based on a deep learning technology, so that the efficiency of cervical cell image analysis is improved, and the workload of doctors is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of medical image processing, and particularly relates to an AI-based cervical cell image analysis assistance method and system. Background Art

[0002] Cervical cancer is one of the common malignant tumors among women globally, seriously threatening women's health and lives. Early and accurate diagnosis is crucial for improving the survival rate of cervical cancer patients. Currently, cervical cell image analysis is one of the important means for cervical cancer screening. By observing and analyzing cervical cell smear images, it is judged whether the cells are abnormal. However, traditional cervical cell image analysis methods have many limitations.

[0003] On the one hand, manual film reading depends on the experience and professional level of pathologists, is highly subjective, and has a large workload, making it easy to have misdiagnosis and missed diagnosis. On the other hand, although some existing automatic analysis methods based on machine learning have improved the diagnosis efficiency to a certain extent, relying only on single cell image data, it is difficult to comprehensively capture the characteristic information of cells, resulting in limited diagnostic accuracy. In addition, due to the complexity and diversity of cervical cell images, the cell images of different patients show great differences, making the generalization ability of the model insufficient. Summary of the Invention

[0004] In view of the above existing technical deficiencies, the present invention provides an AI-based cervical cell image analysis assistance method and system to improve the accuracy and reliability of cervical cell image analysis.

[0005] The present invention is achieved through the following technical solutions:

[0006] There is provided an AI-based cervical cell image analysis assistance method, the method comprising the following steps:

[0007] Step S10: Perform multi-modal data collection, including cervical cell image data, clinical data, and patient gene data, and preprocess the collected cervical cell image data;

[0008] Step S20: Construct a generative adversarial network suitable for cervical cell images. Generate cervical cell images through the generator in the generative adversarial network, and input the cervical cell images generated by the generator and the collected real cervical cell images into a pre-trained convolutional neural network for feature extraction;

[0009] Step S30: Fuse the extracted features with the collected clinical data and patient gene data. Divide the fused feature data into a training set and a test set. The training set is used to train the cervical cell classification model, and the test set is used to evaluate the cervical cell classification model and optimize the model according to the evaluation results to obtain an optimized cervical cell classification model;

[0010] Step S40: Perform data collection, preprocessing and fusion on the real cervical cell image to be measured as in Step S10, and input it into the optimized cervical cell classification model. The model outputs whether the cervical cells are abnormal and the analysis results of the abnormal type and degree.

[0011] Preferably, in Step S10, multi-modal data collection is performed, including cervical cell image data, clinical data and patient gene data, and the collected cervical cell image data is preprocessed. The steps include:

[0012] Cervical cell image data collection: Use a high-resolution microscope to collect cervical cell images to ensure that the images are clear and can accurately display the morphological and structural features of cervical cells. At the same time, record the relevant parameters when collecting cervical cell images, including magnification and exposure time, etc.;

[0013] Clinical information collection: Collect the clinical information of the patient providing the cervical cells, including age, medical history and symptoms, etc. These information help to understand the patient's situation from different angles and provide more comprehensive background information for subsequent analysis;

[0014] Gene data collection: Collect the relevant gene data of the patient providing the cervical cells, including the gene expression levels closely related to the occurrence of cervical cancer, etc. Gene data can reflect the intrinsic molecular characteristics of cells. Combined with cell image data, it can more deeply reveal the abnormal conditions of cells;

[0015] Cervical cell image preprocessing: Perform image preprocessing on the collected cervical cell images, including image denoising, contrast enhancement and normalization, etc., to improve the image quality for subsequent analysis, and perform standardization processing on clinical information and gene data, and extract clinical information feature vectors and gene data feature vectors to enable effective fusion with cervical cell images.

[0016] Preferably, in Step S20, a generative adversarial network applicable to cervical cell images is constructed, including a generator and a discriminator. The steps for constructing the generator include:

[0017] Input layer fuses cervical cell features: While the input layer of the generator takes in a traditional random noise vector, it also incorporates multi-modal data closely related to the biological characteristics of cervical cells. For example, it extracts the density information of cells (number of cells per square millimeter) and statistical features of cell size (average diameter, size variance) from cervical cell smear samples. These features are encoded into vectors and concatenated with the random noise as the input to the generator, making the process of generating images by the generator more in line with the actual feature distribution of cervical cells;

[0018] Deconvolution layer with attention mechanism introduced: An attention mechanism is introduced in the deconvolution layer. In cervical cell images, different regions such as the cell nucleus and cytoplasm have different importance. Through the attention mechanism, the generator can focus more on generating images of key regions;

[0019] Output layer simulates cervical cell texture: A special combination of activation functions is used in the output layer to make the generated cervical cell images closer to the texture of real cervical cell images. The Tanh function is used to limit the output value within the range of -1 to 1, and a texture enhancement layer is added. This texture enhancement layer extracts the texture features of the image through convolution operations on the output image and fuses them with the original output image to enhance the texture details of the generated image and simulate the delicate texture features in cervical cell images;

[0020] Generator loss function combines cervical cell feature matching: The loss function of the generator includes the traditional cross-entropy loss and cell feature matching loss. Calculate the differences in key cell features between the generated cervical cell images and real cervical images, such as the size, shape, and staining intensity of the cell nucleus. The mean squared error loss can be used to measure these feature differences, and it is weighted and summed with the cross-entropy loss as the total loss of the generator, which can make the images generated by the generator closer to real images in terms of key cell features.

[0021] Preferably, the step S20 constructs a generative adversarial network applicable to cervical cell images, including a generator and a discriminator. The steps of constructing the discriminator include:

[0022] Multi-scale feature extraction input layer: The input to the discriminator includes the cervical cell image itself received by the discriminator and feature maps of the cervical cell image at different scales. The features of the cervical cell image have different manifestations at different scales. Through multi-scale feature extraction, the discriminator can understand the features of the image more comprehensively. By using convolutional kernels of different sizes to process the input cervical cell image in parallel, feature maps of the cervical cell at different scales are obtained, and these cervical cell feature maps are concatenated together as the input to the discriminator;

[0023] Feature pooling layer based on cervical cell morphology: Add a feature pooling layer based on cervical cell morphology between convolutional layers. Cervical cells have unique morphological features, such as round, oval, etc. This pooling layer performs pooling operations on the feature map according to the morphological information of cervical cells, retaining important features related to the morphology of cervical cells;

[0024] Output layer combined with clinical knowledge constraints: The output of the discriminator judges true or false based on the features of the image itself and is constrained by clinical knowledge. By introducing a clinical knowledge module in the output layer of the discriminator, the clinical knowledge module contains the proportion information of normal and abnormal cervical cells. The discrimination result is adjusted according to this proportion information to improve the accuracy of the discriminator;

[0025] Discriminator loss function combined with cell distribution information: The loss function of the discriminator combines the distribution information of cervical cells in the image to distinguish the true and false of cervical cell images. In real cervical cell images, the distribution of cells has certain rules, such as the distance between cells, the degree of aggregation, etc. The difference in cell distribution between the generated image and the real image can be calculated, and the Wasserstein distance is used to measure the similarity of cervical cell distribution and incorporated into the loss function of the discriminator.

[0026] Preferably, in step S30, the extracted features are fused with the collected clinical data and patient gene data, and the splicing fusion method is adopted. The steps include:

[0027] Splicing fusion: Sequentially splice the feature vectors of cervical cell images extracted by the pre-trained convolutional neural network, the preprocessed clinical information feature vectors, and the gene data feature vectors together to form a new fused feature vector;

[0028] Normalization processing: Perform normalization processing on the fused feature vector obtained after splicing to balance the contributions of different modality data and ensure that each feature has the same importance in subsequent classification or analysis tasks.

[0029] Preferably, the steps of training the cervical cell classification model in step S30 include:

[0030] Data division: Divide the fused feature data into a training set and a test set according to a ratio of 7:3;

[0031] Construction and Training of Cervical Cell Classification Model: The cervical cell classification model is constructed based on support vector machines and random forests. The kernel function parameters and model parameters are adjusted to make the model applicable to the classification calculation of cervical cells. According to the complexity of the cervical cell image features, the radial basis function is selected as the kernel function. The radial basis function can handle non-linearly separable data and is effective in capturing the complex relationships between cervical cell features. The penalty factor C and the kernel coefficient γ are initialized, and the grid search method is used to dynamically adjust the penalty factor C and the kernel coefficient γ. By evaluating the model performance under different parameter combinations on the test set, the optimal parameter combination is selected. The penalty factor C controls the balance between the training error and the model complexity of the cervical cell classification model. First, a moderate initial value is set, such as C = 1. The kernel coefficient γ affects the calculation result of the radial basis function and can be initialized as γ = 0.1. The random forest includes decision trees. The appropriate number of decision trees is selected according to the actual data, such as 100. Too few decision trees may lead to underfitting of the model, while too many will increase the computational cost and may cause overfitting. According to the model training results, the random search method is used to dynamically adjust the number of decision trees and other random forest parameters, such as the size of the feature subset. When splitting the nodes of each decision tree, a certain number of features are randomly selected from the fused feature data as candidate features. For the fused features of cervical cell images, the appropriate size of the feature subset can be selected according to the number of features and experience, such as selecting the square root of the total number of features as the size of the feature subset.

[0032] Evaluation and Optimization of Cervical Cell Classification Model: The performance of the cervical cell classification model is comprehensively evaluated using accuracy, recall rate, and F1 value. Different weights are assigned to the three evaluation indicators, and the evaluation score of the cervical cell classification model is obtained through weighted calculation. When the evaluation score is greater than the set qualified value, the optimal classification model is obtained. After determining the model version, it is saved and put into actual use.

[0033] Preferably, the analysis results of whether the cervical cells are abnormal and the abnormal type and degree output by the model in step S40 include:

[0034] Determination of Whether Cervical Cells are Abnormal: The cervical cell classification model determines whether the cervical cells are normal or abnormal based on the input real cervical cell image to be tested. For example, the cervical cell classification model comprehensively analyzes the features of the input cervical cell image to be tested and the fused clinical and genetic data features. When its feature pattern highly matches the feature pattern of normal cervical cells in the training set, the model outputs that the cervical cells are normal; when it highly matches the feature pattern of abnormal cervical cells, the model outputs that the cervical cells are abnormal.

[0035] Analysis of abnormal cervical cell results: When the model outputs abnormal cervical cells, the types of abnormal cervical cells are further subdivided, and the degree of abnormal cervical cells is evaluated through a series of quantitative indicators, including the size of the cell nucleus, the degree of irregularity of the cell nucleus shape, the density and distribution uniformity of chromatin, and the specific value of the nuclear-cytoplasmic ratio. Based on the above quantitative indicators, the model divides the degree of abnormal cervical cells into different levels.

[0036] In addition, to achieve the above object, the present invention also provides an AI-based auxiliary system for cervical cell image analysis, and the AI-based auxiliary system for cervical cell image analysis includes:

[0037] Cervical cell multimodal data acquisition module: used for multimodal data acquisition, including cervical cell image data, clinical data, and patient gene data, and preprocessing the acquired cervical cell image data;

[0038] Cervical cell image generation and feature extraction module: used to construct a generative adversarial network suitable for cervical cell images, generate cervical cell images through the generator in the generative adversarial network, and input the cervical cell images generated by the generator and the acquired real cervical cell images into a pre-trained convolutional neural network for feature extraction;

[0039] Cervical cell feature fusion and cervical cell classification model training module: used to fuse the extracted features with the acquired clinical data and patient gene data, divide the fused feature data into a training set and a test set, the training set is used to train the cervical cell classification model, the test set is used to evaluate the cervical cell classification model, and optimize the model according to the evaluation results to obtain an optimized cervical cell classification model;

[0040] Cervical cell abnormality judgment module: used to perform data acquisition, preprocessing, and fusion on the real cervical cell image to be measured, input it into the optimized cervical cell classification model, and the model outputs whether the cervical cells are abnormal and the analysis results of the abnormal type and degree.

[0041] In addition, to achieve the above object, the present invention also provides an AI-based auxiliary device for cervical cell image analysis, and the device includes: a memory, a processor, and programs such as an AI-based auxiliary algorithm for cervical cell image analysis stored on the memory and executable on the processor, and the programs such as the AI-based auxiliary algorithm for cervical cell image analysis are used to implement the steps of an AI-based auxiliary method for cervical cell image analysis as described above.

[0042] In addition, to achieve the above object, the present invention also provides a computer program product, which includes programs such as an AI-based auxiliary algorithm for cervical cell image analysis. When the programs such as the AI-based auxiliary algorithm for cervical cell image analysis are executed by a processor, they implement an AI-based auxiliary method for cervical cell image analysis as described above.

[0043] The advantages and effects of the present invention are as follows:

[0044] An AI-based auxiliary method and system for cervical cell image analysis proposed by the present invention make full use of the complementary information of cervical cell images, clinical information, and gene data through multi-modal data collection, can understand the state of cervical cells more comprehensively and deeply, improve the diagnostic accuracy of cervical cell abnormalities, and reduce the occurrence of misdiagnosis and missed diagnosis. In addition, by combining generative adversarial networks to enhance the features of cervical cell image data, the cervical cell image dataset is expanded, the data diversity is increased, enabling the model to learn richer cell features, improving the generalization ability of the model, and better coping with the complex cell images of different patients. At the same time, it provides complete, objective, and accurate analysis results for doctors, assists doctors in diagnosing cervical cell images, reduces the workload of doctors, improves the diagnostic efficiency, and provides strong support for the early diagnosis and treatment of cervical cancer. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0046] Figure 1 It is a flowchart of an AI-based auxiliary method for cervical cell image analysis of the present invention.

[0047] Figure 2 It is a schematic structural diagram of an AI-based auxiliary system for cervical cell image analysis of the present invention.

[0048] Figure 3 It is a schematic block diagram of the structure of an electronic device for an AI-based auxiliary cervical cell image analysis of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0049] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0050] As Figure 1 shown, in an embodiment of the present invention, an AI-based auxiliary method for cervical cell image analysis includes the following steps:

[0051] Step S10: Perform multi-modal data collection, including cervical cell image data, clinical data, and patient gene data, and preprocess the collected cervical cell image data.

[0052] Specifically, in step S10, multi-modal data collection is performed, including cervical cell image data, clinical data, and patient gene data, and preprocessing of the collected cervical cell image data is carried out. The steps include:

[0053] Cervical cell image data collection: Use a high-resolution microscope to collect cervical cell images, ensuring that the images are clear and can accurately display features such as the morphology and structure of cervical cells. At the same time, record relevant parameters when collecting cervical cell images, including magnification and exposure time, etc. For example, 1000 cervical cell smear samples are collected from a certain hospital, and images are collected using a microscope with a resolution of 400 times to ensure that the cell morphology is clearly shown in the images;

[0054] Clinical information collection: Collect the clinical information of the patients providing cervical cells, including age, medical history, and symptoms, etc. These information helps to understand the patients from different perspectives and provide more comprehensive background information for subsequent analysis. For example, record clinical information such as the age of the patient, whether there is a gynecological history, and whether they are infected with human papillomavirus (HPV);

[0055] Gene data collection: Collect relevant gene data of the patients providing cervical cells, including the gene expression levels closely related to the occurrence of cervical cancer, etc. Gene data can reflect the internal molecular characteristics of cells. Combined with cell image data, it can more deeply reveal the abnormal conditions of cells. For example, through gene detection technology, obtain the expression level data of 10 key genes related to cervical cancer in patients;

[0056] Cervical cell image preprocessing: Perform image preprocessing on the collected cervical cell images, including operations such as image denoising, contrast enhancement, and normalization, to improve the image quality for subsequent analysis. Also, standardize the clinical information and gene data, and extract the clinical information feature vector and gene data feature vector to enable effective fusion with the cervical cell images. For example, perform Gaussian denoising on the cell images, use the histogram equalization method to enhance the contrast, and normalize the images to the [0,1] interval. Encode the clinical information and standardize the gene data so that its mean is 0 and the standard deviation is 1;

[0057] Step S20: Construct a generative adversarial network suitable for cervical cell images. Use the generator in the generative adversarial network to generate cervical cell images, and input the cervical cell images generated by the generator and the collected real cervical cell images into a pre-trained convolutional neural network for feature extraction.

[0058] Specifically, constructing a generative adversarial network suitable for cervical cell images in step S20 includes a generator and a discriminator. The steps for constructing the generator include:

[0059] Input layer fusing cervical cell features: The input layer of the generator, while inputting a traditional random noise vector, also incorporates multi-modal data closely related to the biological characteristics of cervical cells. For example, extract the density information of cells in the cervical cell smear sample (number of cells per square millimeter), statistical features of cell size (average diameter, size variance), etc. Encode these features into vectors and splice them with the random noise as the input of the generator, making the process of the generator generating pictures more conform to the actual feature distribution of cervical cells;

[0060] Deconvolution layer introducing attention mechanism: Introduce an attention mechanism in the deconvolution layer. In cervical cell images, different regions such as the cell nucleus and cytoplasm have different importance. Through the attention mechanism, the generator pays more attention to the image generation of key regions. For example, after a certain deconvolution operation, calculate the attention weights at each position and perform weighted processing on the feature map, so that the features of key regions such as the cell nucleus are generated more accurately. A convolutional-based attention module, such as a channel attention module (SE-Block), can be used to weight the channel dimension of the feature map and highlight important feature channels;

[0061] Output layer simulates cervical cell texture: At the output layer, a special combination of activation functions is used to make the generated cervical cell image closer to the texture of the real cervical cell image. The Tanh function is used to limit the output value within the range of -1 to 1, and a texture enhancement layer is added. This texture enhancement layer extracts the texture features of the image through convolution operations on the output image and fuses them with the original output image to enhance the texture details of the generated image and simulate the delicate texture features in the cervical cell image;

[0062] Generator loss function combines cervical cell feature matching: The loss function of the generator includes the traditional cross-entropy loss and the cell feature matching loss. Calculate the differences between the generated cervical cell image and the real cervical image in terms of key cell features, such as the size, shape, and staining intensity of the cell nucleus. The mean squared error loss can be used to measure these feature differences, and it is weighted and summed with the cross-entropy loss as the total loss of the generator, which can make the image generated by the generator closer to the real image in terms of key cell features.

[0063] Specifically, in step S20, a generative adversarial network suitable for cervical cell images is constructed, including a generator and a discriminator. The steps for constructing the discriminator include:

[0064] Multi-scale feature extraction input layer: The input of the discriminator includes the cervical cell image itself received by the discriminator and the feature maps of the cervical cell image at different scales. The features of the cervical cell image have different manifestations at different scales. For example, at a small scale, the microscopic structure of the cell, such as the distribution of chromatin, can be captured; at a large scale, the overall morphology and distribution of the cell can be observed. Through multi-scale feature extraction, the discriminator can understand the features of the image more comprehensively. By using convolution kernels of different sizes to process the input cervical cell image in parallel, feature maps of different scales of the cervical cell are obtained, and these cervical cell feature maps are concatenated together as the input of the discriminator;

[0065] Feature pooling layer based on cervical cell morphology: A feature pooling layer based on cervical cell morphology is added between the convolutional layers. Cervical cells have unique morphological features, such as round and oval shapes. This pooling layer performs pooling operations on the feature map according to the morphological information of the cervical cell, retaining the important features related to the cervical cell morphology. For example, a shape-based pooling kernel can be used to pool the areas that conform to the cell morphology, enhancing the sensitivity of the discriminator to cell morphological features;

[0066] Output layer combined with clinical knowledge constraint: The output of the discriminator judges the authenticity based on the features of the image itself and is constrained by combining clinical knowledge. For example, in real cervical cell images, the ratio of normal cervical cells to abnormal cervical cells is relatively stable within a certain range. By introducing a clinical knowledge module in the output layer of the discriminator, the clinical knowledge module contains the ratio information of normal cervical cells and abnormal cervical cells, and the discrimination result is adjusted according to this ratio information to improve the accuracy of the discriminator;

[0067] Discriminator loss function combined with cell distribution information: The loss function of the discriminator combines the distribution information of cervical cells in the image to distinguish the authenticity of cervical cell images. In real cervical cell images, the distribution of cells has certain rules, such as the distance between cells, the degree of aggregation, etc. The difference in cell distribution between the generated image and the real image can be calculated, and the Wasserstein distance is used to measure the similarity of cervical cell distribution and incorporated into the loss function of the discriminator. So that the discriminator can better judge the authenticity of the image.

[0068] Step S30: Fuse the extracted features with the collected clinical data and patient gene data, divide the fused feature data into a training set and a test set. The training set is used to train the cervical cell classification model, and the test set is used to evaluate the cervical cell classification model, and the model is optimized according to the evaluation results to obtain an optimized cervical cell classification model.

[0069] Specifically, in step S30, the extracted features are fused with the collected clinical data and patient gene data, and the splicing fusion method is adopted. The steps include:

[0070] Splicing fusion: The feature vector of the cervical cell image extracted by the pre-trained convolutional neural network, the preprocessed clinical information feature vector and the gene data feature vector are spliced together in sequence to form a new fused feature vector. For example, the length of the cervical cell image feature vector is 512, the length of the clinical information feature vector is 10, and the length of the gene data feature vector is 20, then the length of the spliced feature vector is 542;

[0071] Normalization processing: Normalize the fused feature vector obtained after splicing to balance the contributions of different modality data and ensure that each feature has the same importance in subsequent classification or analysis tasks.

[0072] Specifically, the steps of training the cervical cell classification model in step S30 include:

[0073] Data division: Divide the fused feature data into a training set and a test set according to a ratio of 7:3;

[0074] Construction and Training of Cervical Cell Classification Model: The cervical cell classification model is constructed based on support vector machine and random forest. The kernel function parameters and model parameters are adjusted to make the model applicable to the classification calculation of cervical cells. According to the complexity of the cervical cell image features, the radial basis function is selected as the kernel function. The radial basis function can handle non-linearly separable data and is effective in capturing the complex relationships between cervical cell features. The penalty factor C and kernel coefficient γ are initialized, and the grid search method is used to dynamically adjust the penalty factor C and kernel coefficient γ. By evaluating the model performance under different parameter combinations on the test set, the optimal parameter combination is selected. The penalty factor C controls the balance between the training error and model complexity of the cervical cell classification model. First, a moderate initial value is set, such as C = 1. The kernel coefficient γ affects the calculation result of the radial basis function and can be initialized as γ = 0.1. The random forest includes decision trees. The appropriate number of decision trees is selected according to the actual data, such as 100. Too few decision trees may lead to underfitting of the model, while too many will increase the calculation cost and may cause overfitting. According to the model training results, the random search method is used to dynamically adjust the number of decision trees and other random forest parameters, such as the size of the feature subset. When splitting the nodes of each decision tree, a certain number of features are randomly selected from the fused feature data as candidate features. For the fused features of cervical cell images, the appropriate size of the feature subset can be selected according to the number of features and experience, such as selecting the square root of the total number of features as the size of the feature subset.

[0075] Evaluation and Optimization of Cervical Cell Classification Model: The performance of the cervical cell classification model is comprehensively evaluated using accuracy, recall rate, and F1 value. Different weights are assigned to the three evaluation indicators. For example, for the cervical cell classification task, the recall rate (especially the recall rate for abnormal cells) is more important because misdiagnosing abnormal cells may lead to serious consequences. Therefore, the weight of the recall rate is set to 50%, the weight of the accuracy is 25%, and the weight of the F1 value is 25%. The evaluation score of the cervical cell classification model is obtained through weighted calculation. When the evaluation score is greater than the set qualified value, the optimal classification model is obtained. After determining the model version, it is saved and put into actual use.

[0076] Step S40: The real cervical cell image to be tested is subjected to data acquisition, preprocessing, and fusion in Step S10, and then input into the optimized cervical cell classification model. The model outputs whether the cervical cells are abnormal and the analysis results of the abnormal type and degree.

[0077] Specifically, the analysis results of whether the cervical cells are abnormal and the abnormal type and degree output by the model in Step S40 include:

[0078] Determination of whether cervical cells are abnormal: The cervical cell classification model determines whether cervical cells are normal or abnormal based on the input real cervical cell image to be tested. For example, the cervical cell classification model comprehensively analyzes the characteristics of the input cervical cell image to be tested and the fused clinical and genetic data characteristics. When its characteristic pattern highly matches the characteristic pattern of normal cervical cells in the training set, the model outputs that the cervical cells are normal; when it highly matches the characteristic pattern of abnormal cervical cells, the model outputs that the cervical cells are abnormal;

[0079] Analysis of abnormal cervical cell results: When the model outputs that the cervical cells are abnormal, the abnormal types of cervical cells are further subdivided, and a series of quantitative indicators are used to evaluate the degree of abnormal cervical cells, including the size of the cell nucleus, the degree of irregularity of the cell nucleus shape, the density and uniform distribution of chromatin, and the specific value of the nuclear-cytoplasmic ratio, etc. For example, for the size of the cell nucleus, the model calculates the degree of deviation from the average size of the normal cell nucleus and outputs it in percentage form, such as the size of the cell nucleus exceeding 50% of the normal average; for the degree of irregularity of the shape, it is quantitatively evaluated by calculating the ratio of the perimeter of the cell nucleus to the perimeter of a circle with the same area. The more this ratio deviates from 1, the more irregular the shape; the chromatin density can be quantified by statistically analyzing the gray value of the chromatin region in the image. The higher the degree of abnormality, the higher the value of the chromatin density and the more uneven the distribution; based on the above quantitative indicators, the model divides the degree of abnormal cervical cells into different levels, such as mild abnormality, moderate abnormality, and severe abnormality. Mild abnormality is manifested as only slight deviation of indicators such as the size and shape of the cell nucleus from the normal range, and the chromatin change is not obvious; moderate abnormality is manifested as a more significant degree of deviation of each indicator, and the cell morphology shows a certain degree of change; severe abnormality is manifested as serious changes in the cell nucleus, chromatin, and the overall cell morphology, with a great difference from normal cells. At this time, the abnormal degree level output by the model is severe. This degree classification result can help doctors more intuitively understand the severity of abnormal cervical cells, so as to formulate corresponding diagnosis and treatment plans.

[0080] In addition, as Figure 2 shown, in an embodiment of the present invention, an AI-based cervical cell image analysis assistance system is proposed. The AI-based cervical cell image analysis assistance system includes:

[0081] Cervical cell multimodal data acquisition module: used for multimodal data acquisition, including cervical cell image data, clinical data, and patient gene data, and preprocessing the acquired cervical cell image data;

[0082] Cervical cell image generation and feature extraction module: It is used to construct a generative adversarial network applicable to cervical cell images. The generator in the generative adversarial network is used to generate cervical cell images. The generated cervical cell images by the generator and the collected real cervical cell images are input into a pre-trained convolutional neural network for feature extraction;

[0083] Cervical cell feature fusion and cervical cell classification model training module: It is used to fuse the extracted features with the collected clinical data and patient gene data. The fused feature data is divided into a training set and a test set. The training set is used to train the cervical cell classification model, and the test set is used to evaluate the cervical cell classification model and optimize the model according to the evaluation results to obtain an optimized cervical cell classification model;

[0084] Cervical cell abnormality judgment module: It is used to perform data collection and preprocessing in step S10 on the real cervical cell images to be measured and fuse them, and then input them into the optimized cervical cell classification model. The model outputs whether the cervical cells are abnormal and the analysis results of the abnormal type and degree.

[0085] An AI-based cervical cell image analysis assistance system provided by the present application adopts an AI-based cervical cell image analysis assistance method in the above embodiment, which can solve the technical problems of low efficiency and low accuracy of traditional cervical cell analysis methods. Compared with the prior art, the beneficial effects of the AI-based cervical cell image analysis assistance system provided by the present application are the same as those of the AI-based cervical cell image analysis assistance method provided by the above embodiment, and other technical features in the AI-based cervical cell image analysis assistance system are the same as the features disclosed in the method of the above embodiment, which will not be elaborated here.

[0086] The present application provides an AI-based cervical cell image analysis assistance device. The AI-based cervical cell image analysis assistance device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute an AI-based cervical cell image analysis assistance method in Embodiment 1 above.

[0087] As Figure 3As shown, in an embodiment of the present invention, a schematic structural diagram of an AI-based cervical cell image analysis assistance device suitable for implementing the embodiments of the present application is shown. An AI-based cervical cell image analysis assistance device in the embodiments of the present application may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Portable Application Descriptions), PMPs (Portable Media Players), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 3 The shown AI-based cervical cell image analysis assistance device is merely an example and should not impose any limitations on the functions and usage scope of the embodiments of the present application.

[0088] Figure 3 The shown AI-based cervical cell image analysis assistance device may include a processing system 1001 (such as a central processing unit, a graphics processing unit, etc.), which may perform various appropriate actions and processes according to the program stored in the read-only memory (ROM: Read Only Memory) 1002 or the program loaded from the storage system 1003 into the random access memory (RAM: Random Access Memory) 1004. In the RAM 1004, various programs and data required for the operation of the AI-based cervical cell image analysis assistance device are also stored. The processing system 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. The input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems may be connected to the I / O interface 1006: an input system 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output system 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage system 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication system 1009. The communication system 1009 may allow the AI-based cervical cell image analysis assistance device to communicate with other devices wirelessly or wiredly to exchange data. Although the figure shows an AI-based cervical cell image analysis assistance device with various systems, it should be understood that it is not required to implement or have all the shown systems. Instead, more or fewer systems may be implemented or had.

[0089] In particular, according to the embodiments disclosed in the present application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, the embodiments disclosed in the present application include a computer program product that includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from a network through a communication system, or installed from a storage system 1003, or installed from a ROM 1002. When the computer program is executed by a processing system 1001, the above functions defined in the methods of the embodiments disclosed in the present application are executed.

[0090] An AI-based cervical cell image analysis assistance device provided by the present application adopts an AI-based cervical cell image analysis assistance method in the above embodiments, and can solve the technical problems of low efficiency and low accuracy of traditional cervical cell analysis methods. Compared with the prior art, the beneficial effects of the AI-based cervical cell image analysis assistance device provided by the present application are the same as those of the AI-based cervical cell image analysis assistance method provided by the above embodiments, and other technical features in the AI-based cervical cell image analysis assistance device are the same as the features disclosed in the method of the previous embodiment, which will not be elaborated here.

[0091] Each part disclosed in the present application can be implemented by hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in a suitable manner in any one or more embodiments or examples.

[0092] The present application also provides a computer program product, including a computer program, and when the computer program is executed by a processor, the steps of an AI-based cervical cell image analysis assistance method as described above are implemented.

[0093] The computer program product provided by the present application can solve the technical problems of low efficiency and low accuracy of traditional cervical cell analysis methods. Compared with the prior art, the beneficial effects of the computer program product provided by the present application are the same as those of the AI-based cervical cell image analysis assistance method provided by the above embodiments, which will not be elaborated here.

[0094] Obviously, those skilled in the art can make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these modifications and variations.

Claims

1. An AI-based auxiliary method for cervical cell image analysis, characterized in that, The method includes the following steps: Step S10: Perform multi-modal data collection, including cervical cell image data, clinical data, and patient gene data, and preprocess the collected cervical cell image data; Step S20: Construct a generative adversarial network applicable to cervical cell images. Generate cervical cell images through the generator in the generative adversarial network, and input the cervical cell images generated by the generator and the collected cervical cell images into a pre-trained convolutional neural network for feature extraction; Step S30: Integrate the extracted features with the collected clinical data and patient gene data, divide the integrated feature data into a training set and a test set. The training set is used to train a cervical cell classification model, and the test set is used to evaluate the cervical cell classification model, and optimize the model according to the evaluation results to obtain an optimized cervical cell classification model; Step S40: Perform data collection, preprocessing, and integration on the real cervical cell images to be measured, input them into the optimized cervical cell classification model, and the model outputs whether the cervical cells are abnormal and the analysis results of the abnormal type and degree.

2. The auxiliary method for cervical cell image analysis based on AI according to claim 1, wherein In step S10, multi-modal data collection is performed, including cervical cell image data, clinical data, and patient gene data, and preprocessing of the collected cervical cell image data is carried out. The steps include: Cervical cell image data collection: Use a microscope to collect cervical cell images, and record relevant parameters when collecting cervical cell images, including magnification and exposure time; Clinical information collection: Collect the clinical information of the patient providing the cervical cells, including age, medical history, and symptoms; Gene data collection: Collect relevant gene data of the patient providing the cervical cells, including the gene expression levels closely related to the occurrence of cervical cancer; Cervical cell image preprocessing: Perform image preprocessing on the collected cervical cell images, including image denoising, contrast enhancement, and normalization operations, and perform standardization processing on the clinical information and gene data, extract clinical information feature vectors and gene data feature vectors, and make them effectively integrated with the cervical cell images.

3. The auxiliary method for cervical cell image analysis based on AI according to claim 1, characterized in that In step S20, a generative adversarial network applicable to cervical cell images is constructed, including a generator and a discriminator. The steps for constructing the generator include: Input layer fusing cervical cell features: The input layer of the generator, while inputting a traditional random noise vector, also integrates multi-modal data closely related to the biological characteristics of cervical cells, including extracting the density information of cells and the statistical features of cell sizes in cervical cells, encoding these features into vectors and splicing them with the random noise as the input of the generator; Deconvolution layer introducing an attention mechanism: Introduce an attention mechanism in the deconvolution layer. In cervical cell images, the importance of the nucleus and cytoplasm regions is different. Through the attention mechanism, the generator pays more attention to the image generation in the key regions; The output layer simulates the texture of cervical cells: A special combination of activation functions is used in the output layer to make the generated cervical cell images closer to the texture of real cervical cell images. The Tanh function is used to limit the output values within the range of -1 to 1, and a texture enhancement layer is added. This texture enhancement layer extracts the texture features of the image through convolution operations on the output image and fuses them with the generated image of the output layer to enhance the texture details of the generated image and simulate the texture features in cervical cell images; The generator loss function combines cervical cell feature matching: The loss function of the generator includes the traditional cross-entropy loss and the cell feature matching loss. The differences between the generated cervical cell images and the real cervical images in the key cell features are calculated, and the mean squared error loss is used to measure these feature differences. These are weighted and summed with the cross-entropy loss as the total loss of the generator.

4. An AI-based auxiliary method for cervical cell image analysis according to claim 1, wherein, In step S20, a generative adversarial network applicable to cervical cell images is constructed, including a generator and a discriminator. The steps for constructing the discriminator include: Multi-scale feature extraction input layer: The input of the discriminator includes the cervical cell image itself received by the discriminator and the feature maps of the cervical cell image at different scales. The features of the cervical cell image have different manifestations at different scales. By using convolutional kernels of different sizes to process the input cervical cell image in parallel, feature maps of the cervical cell image at different scales are obtained, and these feature maps of the cervical cell image are concatenated together as the input of the discriminator; Feature pooling layer based on cervical cell morphology: A feature pooling layer based on cervical cell morphology is added between convolutional layers. This pooling layer performs pooling operations on the feature maps according to the morphological information of cervical cells, retaining the features related to the morphology of cervical cells; Output layer combined with clinical knowledge constraints: The output of the discriminator judges the authenticity based on the features of the image itself and is combined with clinical knowledge for constraints. By introducing a clinical knowledge module in the output layer of the discriminator, the clinical knowledge module contains the proportion information of normal and abnormal cervical cells, and the discrimination result is adjusted according to this proportion information; The discriminator loss function combines cell distribution information: The loss function of the discriminator combines the distribution information of cervical cells in the image to distinguish the authenticity of cervical cell images. In real cervical cell images, the distribution of cells has certain rules. By calculating the differences in cell distribution between the generated image and the real image, the Wasserstein distance is used to measure the similarity of cervical cell distribution and incorporated into the loss function of the discriminator.

5. The auxiliary method for cervical cell image analysis based on AI according to claim 1, wherein In step S30, the extracted features are fused with the collected clinical data and patient gene data, using the splicing fusion method. The steps include: Splicing fusion: The feature vectors of cervical cell images extracted by a pre-trained convolutional neural network, the pre-processed clinical information feature vectors, and the gene data feature vectors are concatenated in sequence to form a new fused feature vector; Normalization processing: The fused feature vector obtained after concatenation is subjected to normalization processing.

6. The auxiliary method for cervical cell image analysis based on AI according to claim 1, wherein The steps for training the cervical cell classification model in step S30 include: Data division: Divide the fused feature data into a training set and a test set according to a ratio of 7:

3. Construction and training of the cervical cell classification model: The cervical cell classification model is constructed based on support vector machines and random forests. Adjust the kernel function parameters and model parameters to make the model applicable to the classification calculation of cervical cells. Select the radial basis function as the kernel function, initialize the penalty factor C and the kernel coefficient γ, and use the grid search method to dynamically adjust the penalty factor C and the kernel coefficient γ. By evaluating the model performance under different parameter combinations on the test set, select the optimal parameter combination. The penalty factor C controls the balance between the training error and the model complexity of the cervical cell classification model, and set an initial value first; the kernel coefficient γ affects the calculation result of the radial basis function, and is initialized to γ = 0.1; the random forest includes decision trees, select an appropriate number of decision trees according to the actual data, and dynamically adjust the number of decision trees and other random forest parameters using the random search method according to the model training results. When splitting the nodes of each decision tree, randomly select a certain number of features from the fused feature data as candidate features. Evaluation and optimization of the cervical cell classification model: Use accuracy, recall rate, and F1 value to comprehensively evaluate the performance of the cervical cell classification model. The three evaluation indicators use different weights, and the evaluation score of the cervical cell classification model is obtained through weighted calculation. When the evaluation score is greater than the set qualified value, the optimal classification model is obtained.

7. A method for assisting in the analysis of cervical cell images based on AI according to claim 1, characterized in that, In step S40, the model outputs whether the cervical cells are abnormal and the analysis results of the abnormal type and degree, including: Determination of whether cervical cells are abnormal: The cervical cell classification model determines whether the cervical cells are normal or abnormal according to the input real cervical cell image to be measured. Analysis of the abnormal results of cervical cells: When the model outputs that the cervical cells are abnormal, further subdivide the abnormal types of cervical cells, and evaluate the degree of cervical cell abnormality through a series of quantitative indicators, including the size of the cell nucleus, the degree of irregularity of the cell nucleus shape, the density and distribution uniformity of chromatin, and the specific value of the nuclear-cytoplasmic ratio; based on the above quantitative indicators, the model divides the degree of cervical cell abnormality into different levels.

8. An AI-based auxiliary system for cervical cell image analysis, characterized in that, The described AI-based cervical cell image analysis auxiliary system includes: Cervical cell multi-modal data acquisition module: Used for multi-modal data acquisition, including cervical cell image data, clinical data, and patient gene data, and preprocess the collected cervical cell image data. Cervical cell image generation and feature extraction module: Used to construct a generative adversarial network applicable to cervical cell images, generate cervical cell images through the generator in the generative adversarial network, and input the cervical cell images generated by the generator and the collected real cervical cell images into a pre-trained convolutional neural network for feature extraction. Cervical cell feature fusion and cervical cell classification model training module: used to fuse the extracted features with the collected clinical data and patient gene data, divide the fused feature data into a training set and a test set, the training set is used to train the cervical cell classification model, the test set is used to evaluate the cervical cell classification model, and optimize the model according to the evaluation results to obtain an optimized cervical cell classification model; Cervical cell abnormality judgment module: used to perform data collection and preprocessing in step S10 on the real cervical cell image to be measured and perform fusion, input it into the optimized cervical cell classification model, and the model outputs whether the cervical cells are abnormal and the analysis results of the abnormal type and degree.

9. An AI-based auxiliary device for cervical cell image analysis, characterized in that, The AI-based cervical cell image analysis auxiliary device described above includes: A memory, a processor, and an AI-based cervical cell image analysis auxiliary program stored on the memory and executable on the processor. When the AI-based cervical cell image analysis auxiliary program is executed by the processor, it implements the AI-based cervical cell image analysis auxiliary method according to any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes an AI-based cervical cell image analysis auxiliary program. When the AI-based cervical cell image analysis auxiliary program is executed by the processor, it implements the AI-based cervical cell image analysis auxiliary method according to any one of claims 1 to 7.

Images