Pathological risk prediction method and device, equipment, storage medium and program product

Abstract

The invention relates to a pathological risk prediction method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring to-be-predicted data and a needle biopsy result of a target user; determining a target image prediction model according to the puncture biopsy result; performing risk level feature construction on to-be-predicted image data of the to-be-predicted data according to the target image prediction model to obtain an image level probability vector; performing overall feature construction according to the image level probability vector and the to-be-predicted data to obtain an overall risk feature vector; and according to a pre-trained target pathology risk prediction model, performing pathology grading prediction on the overall risk feature vector to obtain a pathology grading prediction result. By adopting the method, the pathological risk level can be accurately predicted in combination with the multi-modal data.

Description

Technical Field

[0001] The present application relates to the technical field of pathology assessment, and in particular to a pathology risk prediction method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] With the development of pathological assessment technology, pathological risk prediction is needed to assist clinicians in assessing pathological risk during clinical pathological assessment. Subsequent treatment plans can be further determined based on the assessment results. However, in the existing technology, assessment technology for breast cancer pathology usually only uses image data or pathological staining sections as data sources for pathological assessment, and is used to classify breast cancer-related subtypes or assess related lymph node metastases. These assessment methods cannot accurately assist in preoperative pathological grade judgment, so the preoperative biopsy pathological grade may be underestimated, resulting in the problem of secondary surgery for patients. Summary of the invention

[0003] Based on this, it is necessary to provide a pathology risk prediction method, device, computer equipment, computer-readable storage medium and computer program product that can accurately predict the pathology risk level by combining multimodal data to address the above technical problems.

[0004] In a first aspect, the present application provides a pathology risk prediction method, comprising:

[0005] Obtain the target user's predicted data and biopsy results;

[0006] Determining a target image prediction model according to the puncture biopsy result;

[0007] Constructing risk level features of the image data to be predicted according to the target image prediction model to obtain an image level probability vector;

[0008] Performing overall feature construction according to the image grade probability vector and the data to be predicted to obtain an overall risk feature vector;

[0009] The pathology grade prediction is performed on the overall risk feature vector according to the pre-trained target pathology risk prediction model to obtain a pathology grade prediction result.

[0010] In one embodiment, determining a target image prediction model according to the biopsy result includes:

[0011] If the puncture biopsy result is the first preset result, the binary classification target prediction model is set as the target image prediction model;

[0012] If the puncture biopsy result is the second preset result, the three-category target prediction model is set as the target image prediction model.

[0013] In one embodiment, the overall feature construction is performed according to the image level probability vector and the data to be predicted to obtain the overall risk feature vector, including:

[0014] Performing text feature extraction on the to-be-predicted user data of the to-be-predicted data to obtain a text level feature vector;

[0015] The text level feature vector and the image level probability vector are concatenated to obtain the overall risk feature vector.

[0016] In one of the embodiments, constructing risk level features of the image data to be predicted according to the target image prediction model to obtain an image level probability vector includes:

[0017] Performing risk level assessment on the image data to be predicted according to the target image prediction model to obtain an assessed risk level probability;

[0018] A feature vector is constructed for the assessed risk level probability according to the type of the target image prediction model to obtain the image level probability vector.

[0019] In one embodiment, the method further comprises:

[0020] Acquire data to be trained and training biopsy results corresponding to the data to be trained;

[0021] Determining an original image prediction model according to the training biopsy results;

[0022] Performing a risk level probability assessment on the training image data of the training data according to the original image prediction model to obtain a training level probability vector;

[0023] Performing overall feature construction according to the training level probability vector and the data to be trained to obtain an overall training feature vector;

[0024] The original pathology risk prediction model is trained according to the overall training feature vector and the validation set data to obtain the target pathology risk prediction model.

[0025] In one embodiment, after performing pathological grade prediction on the overall risk feature vector according to the pre-trained target pathological risk prediction model to obtain a pathological grade prediction result, the method further includes:

[0026] Obtaining a thermal activation map of the target user;

[0027] Performing cluster analysis on the thermal activation map to obtain closed connected domains;

[0028] The closed connected area is highlighted to obtain a target prompt area.

[0029] In a second aspect, the present application also provides a pathology risk prediction device, comprising:

[0030] An acquisition module is used to obtain the target user's predicted data and biopsy results;

[0031] A determination module, used to determine a target image prediction model according to the puncture biopsy result;

[0032] An evaluation module, configured to construct risk level features of the image data to be predicted according to the target image prediction model to obtain an image level probability vector;

[0033] A feature module, used for constructing an overall feature according to the image grade probability vector and the data to be predicted to obtain an overall risk feature vector;

[0034] The grading prediction module is used to perform pathology grading prediction on the overall risk feature vector according to the pre-trained target pathology risk prediction model to obtain a pathology grading prediction result.

[0035] In a third aspect, the present application further provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:

[0036] Obtain the target user's predicted data and biopsy results;

[0037] Determining a target image prediction model according to the puncture biopsy result;

[0038] Constructing risk level features of the image data to be predicted according to the target image prediction model to obtain an image level probability vector;

[0039] Performing overall feature construction according to the image grade probability vector and the data to be predicted to obtain an overall risk feature vector;

[0040] The pathology grade prediction is performed on the overall risk feature vector according to the pre-trained target pathology risk prediction model to obtain a pathology grade prediction result.

[0041] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the following steps are implemented:

[0042] Obtain the target user's predicted data and biopsy results;

[0043] Determining a target image prediction model according to the puncture biopsy result;

[0044] Constructing risk level features of the image data to be predicted according to the target image prediction model to obtain an image level probability vector;

[0045] Performing overall feature construction according to the image grade probability vector and the data to be predicted to obtain an overall risk feature vector;

[0046] The pathology grade prediction is performed on the overall risk feature vector according to the pre-trained target pathology risk prediction model to obtain a pathology grade prediction result.

[0047] In a fifth aspect, the present application further provides a computer program product, including a computer program, which implements the following steps when executed by a processor:

[0048] Obtain the target user's predicted data and biopsy results;

[0049] Determining a target image prediction model according to the puncture biopsy result;

[0050] Constructing risk level features of the image data to be predicted according to the target image prediction model to obtain an image level probability vector;

[0051] Performing overall feature construction according to the image grade probability vector and the data to be predicted to obtain an overall risk feature vector;

[0052] The pathology grade prediction is performed on the overall risk feature vector according to the pre-trained target pathology risk prediction model to obtain a pathology grade prediction result.

[0053] The above-mentioned pathological risk prediction method, device, computer equipment, computer-readable storage medium and computer program product obtain the target user's data to be predicted and the puncture biopsy results; determine the target image prediction model according to the puncture biopsy results; construct the risk level feature of the image data to be predicted according to the target image prediction model to obtain the image level probability vector; construct the overall feature according to the image level probability vector and the data to be predicted to obtain the overall risk feature vector; predict the pathological grade of the overall risk feature vector according to the pre-trained target pathological risk prediction model to obtain the pathological grade prediction result. Therefore, by determining the target image prediction model according to the puncture biopsy results, that is, using different target image prediction models to construct the risk level feature of the image data to be predicted, and then combining the data to be predicted to construct the overall feature, an overall risk feature vector that can better reflect the actual overall pathological situation of the target user is obtained, and then the pathological grade prediction of the overall risk feature vector is performed through the pre-trained target image prediction model, thereby achieving a more accurate pathological grade prediction, which can further avoid the problem of underestimation of the pathological grade of preoperative biopsy. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0055] FIG1 is a diagram of an application environment of a pathology risk prediction method in one embodiment;

[0056] FIG2 is a schematic diagram of a process flow of a pathology risk prediction method in one embodiment;

[0057] FIG3 is a schematic diagram of a process of predicting pathological risk in one embodiment;

[0058] FIG4 is a schematic diagram showing the effect of a pathological region prompting step of a pathological risk prediction method in one embodiment;

[0059] FIG5 is a schematic diagram of a process of a pathology risk prediction method in one embodiment;

[0060] FIG6 is a structural block diagram of a pathology risk prediction device in one embodiment;

[0061] FIG. 7 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0062] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0063] The pathological risk prediction method provided in the embodiment of the present application can be applied in the application environment shown in FIG. 1. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. The server 104 obtains the target user's to-be-predicted data and the puncture biopsy results; determines the target image prediction model according to the puncture biopsy results; constructs the risk level feature of the to-be-predicted image data according to the target image prediction model to obtain the image level probability vector; constructs the overall feature according to the image level probability vector and the to-be-predicted data to obtain the overall risk feature vector; and predicts the pathological grade of the overall risk feature vector according to the pre-trained target pathological risk prediction model to obtain the pathological grade prediction result. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, projection devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The head mounted device may be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server 104 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0064] In an exemplary embodiment, as shown in FIG2 , a pathology risk prediction method is provided, which is described by taking the method applied to the server 104 in FIG1 as an example, and includes the following steps 202 to 210. Among them:

[0065] Step 202: Obtain the target user's predicted data and biopsy results.

[0066] The target user can be a patient who has been diagnosed with breast cancer or a patient who is at risk of breast cancer. The data to be predicted includes image data to be predicted and user data to be predicted. The image data to be predicted can include breast ultrasound images and breast mammography images. The user data to be predicted can include basic information of the patient and immunohistochemistry report results. The basic information includes age, height, and weight. The immunohistochemistry report results include the positive and negative results of estrogen receptors, progesterone receptors, human epidermal growth factor receptors, and proliferation index.

[0067] Among them, estrogen receptor refers to ER, progesterone receptor is PR (progesterone receptor), and human epidermal growth factor receptor refers to HER2 (human epidermal growth factor receptor 2). The proliferation index refers to the Ki67 biomarker. Ki67 is an indicator related to cell proliferation. The higher the positive rate, the faster the tumor grows, the worse the tissue differentiation, and the more sensitive it is to chemotherapy.

[0068] In some embodiments, the target user's predicted data and the target user's puncture biopsy results can be directly obtained from the patient database, or the target user's predicted data and puncture biopsy results can be obtained in real time through device detection, but is not limited thereto.

[0069] In some embodiments, after the data to be predicted is obtained, the images to be predicted in the data to be predicted are preprocessed, and each image to be predicted is converted into a grayscale image, and its size is adjusted to 1*448*448.

[0070] Step 204: determine a target image prediction model according to the biopsy result.

[0071] Among them, the puncture biopsy results can be divided into two types of results, namely low-grade ductal carcinoma in situ and intermediate- and high-grade ductal carcinoma in situ.

[0072] In some embodiments, the target image prediction model is determined by the biopsy result, and a target image prediction model that matches the type of the biopsy result can be selected, thereby improving the accuracy of risk level assessment using the target image prediction model in subsequent steps.

[0073] In some embodiments, the target image prediction model is determined according to the puncture biopsy result, including: if the puncture biopsy result is a first preset result, a two-category target prediction model is set as the target image prediction model; if the puncture biopsy result is a second preset result, a three-category target prediction model is set as the target image prediction model.

[0074] Among them, the first preset result can be medium or high grade ductal carcinoma in situ, the second preset result can be low grade ductal carcinoma in situ, the two-category target prediction model is a pre-trained two-category neural network model, and the three-category target prediction model is a pre-trained three-category neural network model.

[0075] In some embodiments, if the puncture biopsy result is the first preset result, that is, if the puncture biopsy result is a medium-to-high-grade ductal carcinoma in situ, the binary target prediction model is set as the target image prediction model. It should be noted that when the puncture biopsy result is a medium-to-high-grade ductal carcinoma in situ, the patient's subsequent risk assessment level process only needs to evaluate the probability of medium-to-high grades and upgrading to invasive cancer. Therefore, using a binary target prediction model as a target image prediction model does not require the evaluation of low-level grade probabilities, thereby reducing the low-level probability evaluation training in the pre-training stage of the target image prediction model, thereby saving computer resource consumption, and the training speed of the target image prediction model can be more efficient.

[0076] In some embodiments, if the puncture biopsy result is the second preset result, that is, if the puncture biopsy result is low-grade ductal carcinoma in situ, the three-category target prediction model is set as the target image prediction model. It should be noted that when the puncture biopsy result is low-grade ductal carcinoma in situ, the patient's subsequent risk assessment level process needs to be evaluated for low-grade, medium-high-grade and the probability of infiltration. Therefore, the use of the three-category target prediction model can more accurately evaluate the three situations.

[0077] Step 206 , construct risk level features for the image data to be predicted according to the target image prediction model to obtain an image level probability vector.

[0078] The image level probability vector is a probability vector used to represent that the image data to be predicted is assessed as having different risk levels.

[0079] In some embodiments, risk level features are constructed for the image data to be predicted according to the target image prediction model to obtain an image level probability vector, including: evaluating the risk level of the image data to be predicted according to the target image prediction model to obtain an evaluated risk level probability; and constructing a feature vector for the evaluated risk level probability according to the type of the target image prediction model to obtain an image level probability vector.

[0080] The evaluated risk level probability is used to indicate the probability that the image data to be predicted is evaluated as having different risk levels.

[0081] In some embodiments, the risk level of the image data to be predicted in the prediction data is evaluated according to the target image prediction model, and the probability of the image data to be predicted belonging to each risk level can be accurately evaluated, thereby obtaining the estimated risk level probability. For example, the probability of the estimated risk level obtained by evaluating a piece of image data to be predicted can be that the probability that the target user is a medium-to-high-level patient is 20%, and the probability that the target user is a patient with infiltration is 80%.

[0082] In some embodiments, a feature vector of the assessed risk level probability is constructed according to the type of the target image prediction model. If the type is a binary classification neural network model, the assessed risk level probability is constructed into an image level probability vector with a length of 3 by filling in 0 bits. If the type is a three-classification neural network model, the assessed risk level probability is directly constructed into an image level probability vector with a length of 3.

[0083] It should be noted that since the binary target prediction model can only generate a probability vector of length 2, and the three-category target prediction model generates a probability vector of length 3, in order to ensure that the length of the vector generated by the target image prediction model is consistent, so as to facilitate the subsequent steps and improve the system processing efficiency, the probability vector generated by the binary target prediction model will be padded with 0 bits, filling the 0 bit of the probability vector with 0, representing the low-level probability of 0, forming a probability vector of length 3.

[0084] Step 208, constructing an overall feature based on the image level probability vector and the data to be predicted to obtain an overall risk feature vector.

[0085] The overall risk feature vector is an overall feature vector representing all the risk features related to the pathology in the data to be predicted.

[0086] In some embodiments, an overall feature is constructed based on the image level probability vector and the data to be predicted to obtain an overall risk feature vector, including: extracting text features from the user data to be predicted to obtain a text level feature vector; and concatenating the text level feature vector and the image level probability vector to obtain an overall risk feature vector.

[0087] In some embodiments, text extraction is performed on the user data to be predicted to obtain patient information and immunohistochemistry results, and the patient information and immunohistochemistry results are one-hot encoded and embedded to form a text-level feature vector with a length of 7, and the image-level probability vector and the text-level feature vector are concatenated to form an overall risk feature vector with a length of 10.

[0088] Among them, unique hot encoding embedding refers to One-hot encoding embedding.

[0089] Step 210 , predicting the pathology grade of the overall risk feature vector according to the pre-trained target pathology risk prediction model, and obtaining a pathology grade prediction result.

[0090] Among them, the target pathology risk prediction model is a disease risk assessment model pre-trained by machine learning methods, which is used to predict the possibility of ductal carcinoma in situ pathological grading in patients.

[0091] In some embodiments, the overall risk feature vector is input into the target pathology risk prediction model, and the target pathology risk prediction model performs pathology grading prediction on the overall risk feature vector to obtain pathology grading prediction results including various pathology levels predicted and corresponding probabilities.

[0092] Among them, the pathological grading prediction result can be used as a reference scoring indicator in other pathological grading evaluation systems. It should be noted that the pathological grading prediction result is only an intermediate result to assist doctors in judging and analyzing the patient's pathology.

[0093] In some embodiments, the method also includes: a model training step, including: obtaining the data to be trained and the training biopsy results corresponding to the data to be trained; determining the original image prediction model based on the training biopsy results; performing a risk level probability assessment on the image data to be trained based on the original image prediction model to obtain a training level probability vector; performing overall feature construction based on the training level probability vector and the data to be trained to obtain an overall training feature vector; performing model training on the original pathology risk prediction model based on the overall training feature vector and the validation set data to obtain a target pathology risk prediction model.

[0094] Among them, the data to be trained are training users used for model training. The training users are patients who have been diagnosed with breast cancer or patients who are at risk of breast cancer. The data to be trained include image data to be trained and user data to be trained. Among them, the image data to be trained may include breast ultrasound images and breast mammography images, and the user data to be trained may include basic information of the training users and immunohistochemistry report results.

[0095] Among them, the training biopsy results can be divided into two types of results, namely low-grade ductal carcinoma in situ and intermediate- and high-grade ductal carcinoma in situ.

[0096] In some embodiments, the data to be trained and the training biopsy results corresponding to the data to be trained may be obtained from a patient database, but is not limited thereto.

[0097] In some embodiments, different types of original image prediction models are determined according to the type of training biopsy results. If the training biopsy result is a first preset result, the two-category original prediction model is set as the original image prediction model; if the training biopsy result is a second preset result, the three-category original prediction model is set as the original image prediction model.

[0098] In some embodiments, a risk level probability assessment is performed on the image data to be trained according to the original image prediction model to obtain a training level probability vector, and an overall feature is constructed based on the training level probability vector and the data to be trained to obtain an overall training feature vector, and then the overall training feature vector is input into the original pathology risk prediction model, and the validation set data is used as a label, and the original pathology risk prediction model is trained through a machine learning method to obtain a target pathology risk prediction model.

[0099] Among them, the original pathological risk prediction model is a risk assessment model pre-constructed according to the overall training feature vector, the validation set data refers to the postoperative pathological results of the training users, and the machine learning method can adopt linear regression and decision tree algorithms.

[0100] In the above pathological risk prediction method, the target user's data to be predicted and the biopsy results are obtained; the target image prediction model is determined according to the biopsy results; the risk level feature of the image data to be predicted is constructed according to the target image prediction model to obtain the image level probability vector; the overall feature is constructed according to the image level probability vector and the data to be predicted to obtain the overall risk feature vector; the pathological grade prediction is performed on the overall risk feature vector according to the pre-trained target pathological risk prediction model to obtain the pathological grade prediction result. Therefore, by determining the target image prediction model according to the biopsy results, that is, using different target image prediction models to construct the risk level feature of the image data to be predicted, and then combining the data to be predicted to perform overall feature construction, an overall risk feature vector that can better reflect the actual overall pathological situation of the target user is obtained, and then the pathological grade prediction is performed on the overall risk feature vector through the pre-trained target image prediction model, thereby achieving a more accurate pathological grade prediction, which can further avoid the problem of underestimation of the pathological grade of preoperative biopsy.

[0101] In an exemplary embodiment, after the pathological grade prediction is performed, it is still necessary to pay attention to whether the pathological grade prediction result is consistent with the puncture pathological result. If they are inconsistent, it means that the target user may have missed tissue parts, and the pathological area of ​​the target user needs to be re-judged. Therefore, this solution resets the pathological area prompting step as shown in Figure 3, and the pathological area prompting step includes steps 302 to 306. Among them:

[0102] Step 302: Obtain a thermal activation map of the target user.

[0103] Among them, the thermal activation map refers to the thermal activation map of the malignant characteristic area of ​​breast nodules under the ultrasound mode of the target user. The highlighted area of ​​the thermal activation map indicates that there is a high probability of pathological upgrade.

[0104] In some embodiments, the thermal activation map may be obtained by acquiring a thermal activation map stored in a database, or by acquiring a thermal activation map by ultrasonic detection, but is not limited thereto.

[0105] In some embodiments, if the pathological grading prediction result is inconsistent with the puncture pathological result, the thermal activation map of the target user is obtained. It should be noted that a major reason for misdiagnosis of pathological grading in clinical practice is that the tissue sampled by the puncture biopsy is not the highest grade of malignancy. When the pathological grading prediction result is inconsistent with the puncture pathological result, we obtain the thermal activation map of the target user for information feedback to indicate the tissue part that may be missed, thereby further improving the accuracy of pathological grading.

[0106] Step 304: Perform cluster analysis on the thermal activation map to obtain closed connected domains.

[0107] Among them, the closed connected domain refers to the closed connected domain that is most relevant to the upgrade of pathological grade.

[0108] In some embodiments, as shown in FIG. 4 , based on the thermal activation map, a pixel clustering method is used to perform cluster analysis on the ultrasound image region of the thermal activation map to find the closed connected domain that is most relevant to the upgrade of the pathology grade.

[0109] Step 306: highlight the closed connected area to obtain a target prompt area.

[0110] In some embodiments, the closed interconnected area is circled and highlighted to obtain a target prompt area, which is then displayed.

[0111] In this embodiment, by obtaining the thermal activation map of the target user, clustering analysis is performed on the thermal activation map to obtain a closed connected domain, and the closed connected area is highlighted to obtain a target prompt area. Accurate and timely feedback can be given on misdiagnosis of pathological grading to prompt tissue parts that may be missed, thereby further improving the accuracy of pathological grading.

[0112] In order to better understand the solution of the present application, an explanation is given here in conjunction with FIG. 5 , as follows:

[0113] In some embodiments, as shown in FIG5 , the image data to be predicted (including ultrasound images and molybdenum target images as shown in FIG5 ) and the puncture knot results in the data to be predicted are first input into a selected target image prediction model (a neural network as shown in FIG5 ), and then the image grade probability vector corresponding to each image is evaluated (a risk grade probability matrix as shown in FIG5 ), and then the overall feature is constructed together with the user data to be predicted in the data to be predicted (the patient information and immune refinement information as shown in FIG5 ) to obtain an overall risk feature vector, and finally the target pathology risk prediction model (a linear regression model or a decision tree model as shown in FIG5 ) is input for pathology grade prediction to obtain a pathology grade prediction result.

[0114] It should be understood that, although the various steps in the flowcharts involved in the above-mentioned embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps does not have a strict order restriction, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-mentioned embodiments can include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0115] Based on the same inventive concept, the embodiment of the present application also provides a pathological risk prediction device for implementing the pathological risk prediction method involved above. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above method, so the specific limitations in one or more pathological risk prediction device embodiments provided below can refer to the limitations of the pathological risk prediction method above, and will not be repeated here.

[0116] In an exemplary embodiment, as shown in FIG6 , a pathology risk prediction device is provided, comprising: an acquisition module 601, a determination module 602, a first feature module 603, a second feature module 604 and a grading prediction module 605, wherein:

[0117] An acquisition module 601 is used to acquire the target user's to-be-predicted data and the biopsy result;

[0118] A determination module 602 is used to determine a target image prediction model according to the puncture biopsy result;

[0119] The first feature module 603 is used to construct risk level features of the image data to be predicted according to the target image prediction model to obtain an image level probability vector;

[0120] The second feature module 604 is used to construct an overall feature according to the image level probability vector and the data to be predicted to obtain an overall risk feature vector;

[0121] The grading prediction module 605 is used to perform pathology grading prediction on the overall risk feature vector according to the pre-trained target pathology risk prediction model to obtain a pathology grading prediction result.

[0122] In some embodiments, the determination module 602 is also used to set the two-category target prediction model as the target image prediction model if the puncture biopsy result is the first preset result; if the puncture biopsy result is the second preset result, set the three-category target prediction model as the target image prediction model.

[0123] In some embodiments, the second feature module 604 is further used to extract text features from the user data to be predicted to obtain a text level feature vector; and to concatenate the text level feature vector and the image level probability vector to obtain an overall risk feature vector.

[0124] In some embodiments, the first feature module 603 is also used to evaluate the risk level of the image data to be predicted according to the target image prediction model to obtain the evaluated risk level probability; construct a feature vector of the evaluated risk level probability according to the type of the target image prediction model to obtain an image level probability vector.

[0125] In some embodiments, the device also includes: a model training module, used to obtain the data to be trained and the training biopsy results corresponding to the data to be trained; determine the original image prediction model according to the training biopsy results; perform risk level probability evaluation on the training image data to be trained according to the original image prediction model to obtain a training level probability vector; perform overall feature construction based on the training level probability vector and the data to be trained to obtain an overall training feature vector; perform model training on the original pathology risk prediction model based on the overall training feature vector and the validation set data to obtain a target pathology risk prediction model.

[0126] In some embodiments, after performing pathology grading prediction on the overall risk feature vector according to the pre-trained target pathology risk prediction model to obtain the pathology grading prediction result, the device also includes: a prompt module, used to obtain a thermal activation map of the target user; perform cluster analysis on the thermal activation map to obtain a closed connected domain; and highlight the closed connected area to obtain a target prompt area.

[0127] In the above-mentioned pathological risk prediction device, the target user's data to be predicted and the puncture biopsy results are obtained; the target image prediction model is determined according to the puncture biopsy results; the risk level feature of the image data to be predicted is constructed according to the target image prediction model to obtain an image level probability vector; the overall feature is constructed according to the image level probability vector and the data to be predicted to obtain an overall risk feature vector; the pathological grade prediction is performed on the overall risk feature vector according to the pre-trained target pathological risk prediction model to obtain a pathological grade prediction result. Therefore, by determining the target image prediction model according to the puncture biopsy results, that is, using different target image prediction models to construct the risk level feature of the image data to be predicted, and then combining the data to be predicted to perform overall feature construction, an overall risk feature vector that can better reflect the actual overall pathological situation of the target user is obtained, and then the pathological grade prediction is performed on the overall risk feature vector through the pre-trained target image prediction model, thereby achieving a more accurate pathological grade prediction, which can further avoid the problem of underestimation of the pathological grade of preoperative biopsy.

[0128] Each module in the above-mentioned pathological risk prediction device can be implemented in whole or in part by software, hardware and a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0129] In an exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be shown in FIG7. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store the target user's to-be-predicted data and the puncture biopsy results. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a pathological risk prediction method is implemented.

[0130] Those skilled in the art will understand that the structure shown in FIG. 7 is merely a block diagram of a partial structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different arrangement of components.

[0131] In an exemplary embodiment, a computer device is provided. The computer device includes a memory and a processor. The memory stores a computer program. When the processor executes the computer program, the steps in the above method embodiments are implemented.

[0132] In some embodiments, a computer-readable storage medium is provided, which stores a computer program. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0133] In some embodiments, a computer program product is provided. The computer program product includes a computer program. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0134] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0135] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment method can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0136] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0137] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the present application. It should be noted that, for a person of ordinary skill in the art, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A pathological risk prediction method, characterized in that: The method comprises: Obtain the target user's predicted data and biopsy results; Determining a target image prediction model according to the puncture biopsy result; Constructing risk level features of the image data to be predicted according to the target image prediction model to obtain an image level probability vector; Performing overall feature construction according to the image grade probability vector and the data to be predicted to obtain an overall risk feature vector; The pathology grade prediction is performed on the overall risk feature vector according to the pre-trained target pathology risk prediction model to obtain a pathology grade prediction result.

2. The method according to claim 1, characterized in that Determining the target image prediction model according to the puncture biopsy result includes: If the puncture biopsy result is the first preset result, the binary classification target prediction model is set as the target image prediction model; If the puncture biopsy result is the second preset result, the three-category target prediction model is set as the target image prediction model.

3. The method according to claim 1, characterized in that The overall feature construction is performed according to the image level probability vector and the data to be predicted to obtain an overall risk feature vector, including: Performing text feature extraction on the to-be-predicted user data of the to-be-predicted data to obtain a text level feature vector; The text level feature vector and the image level probability vector are concatenated to obtain the overall risk feature vector.

4. The method according to claim 1, characterized in that: The step of constructing risk level features of the image data to be predicted according to the target image prediction model to obtain an image level probability vector includes: Performing risk level assessment on the image data to be predicted according to the target image prediction model to obtain an assessed risk level probability; A feature vector is constructed for the assessed risk level probability according to the type of the target image prediction model to obtain the image level probability vector.

5. The method according to claim 1, characterized in that: The method further comprises: Acquire data to be trained and training biopsy results corresponding to the data to be trained; Determining an original image prediction model according to the training biopsy results; Performing a risk level probability assessment on the training image data of the training data according to the original image prediction model to obtain a training level probability vector; Performing overall feature construction according to the training level probability vector and the data to be trained to obtain an overall training feature vector; The original pathology risk prediction model is trained according to the overall training feature vector and the validation set data to obtain the target pathology risk prediction model.

6. The method according to any one of claims 1 to 5, characterized in that: After performing pathological grade prediction on the overall risk feature vector according to the pre-trained target pathological risk prediction model to obtain a pathological grade prediction result, the method further includes: Obtaining a thermal activation map of the target user; Performing cluster analysis on the thermal activation map to obtain closed connected domains; The closed connected area is highlighted to obtain a target prompt area.

7. A pathology risk prediction device, characterized in that: The device comprises: An acquisition module is used to obtain the target user's predicted data and biopsy results; A determination module, used to determine a target image prediction model according to the puncture biopsy result; A first feature module is used to construct risk level features of the image data to be predicted according to the target image prediction model to obtain an image level probability vector; A second feature module is used to construct an overall feature according to the image grade probability vector and the data to be predicted to obtain an overall risk feature vector; The grading prediction module is used to perform pathology grading prediction on the overall risk feature vector according to the pre-trained target pathology risk prediction model to obtain a pathology grading prediction result.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images