Prognostic effect prediction method and system based on multiple features

Abstract

Disclosed in the present invention are a prognostic effect prediction method and system based on multiple features. The method comprises: acquiring a tissue cell image of a target patient; performing feature calculation on the tissue cell image, so as to determine type parameters corresponding to a plurality of preset cell feature types; on the basis of a parameter relationship between the type parameters corresponding to at least two of the cell feature types, determining at least one corresponding prognostic effect prediction model; and on the basis of the prognostic effect prediction model and the type parameters corresponding to the cell feature types, calculating a prognostic risk score corresponding to the target patient. It can be seen that the present invention can make full use of the relationship between cell features to accurately select a prognostic effect prediction model, so as to achieve more accurate prognostic effect prediction, thereby improving the prediction efficiency and accuracy of cancer prognostic effect prediction, and providing an accurate data basis for disease treatment or research.

Description

Multi-feature-based prognosis effect prediction method and system TECHNICAL FIELD

[0001] The present application relates to the technical field of data processing, in particular to a multi-feature-based prognosis effect prediction method and system. BACKGROUND

[0002] Cancer is a major disease that seriously threatens human life and health. For cancer tumor patients, targeted therapy for specific driver gene mutations and the emerging immunotherapy for the regulation of the body's immune system in recent years have good effects, and effective prediction of the prognosis effect of treatment has become an important technical problem of concern. With the development of image processing technology and the popularization of medical auxiliary imaging technology, more and more institutions have introduced image recognition technology to extract and analyze cell features in biological tissue images to realize prediction of the prognosis effect of patients. However, the existing prognosis effect prediction technology generally only considers the feature parameters corresponding to the pre-screened features to determine the risk without further screening corresponding algorithm models according to the relationship between different cell features to make more accurate parameter prediction, so the efficiency and accuracy of the prognosis effect prediction are both lacking. Therefore, the existing technology has defects and needs to be solved. SUMMARY

[0003] The technical problem to be solved by the present application is to provide a multi-feature-based prognosis effect prediction method and system that can make full use of the relationship between cell features to accurately screen out a prognosis effect prediction model to realize more accurate prognosis effect prediction and improve the prediction efficiency and accuracy of cancer prognosis effect prediction, thereby providing accurate data basis for disease treatment or research.

[0004] To solve the above technical problems, the present application discloses a multi-feature-based prognosis effect prediction method in the first aspect, which comprises:

[0005] Obtaining a tissue cell image of a target patient;

[0006] Performing feature calculation on the tissue cell image to determine type parameters corresponding to a plurality of preset cell feature types;

[0007] Determining at least one corresponding prognosis effect prediction model according to the parameter relationship between the type parameters corresponding to at least two cell feature types;

[0008] Calculating a prognosis risk score corresponding to the target patient according to the prognosis effect prediction model and the type parameters corresponding to the cell feature types.

[0009] As an optional implementation, in the first aspect of the present application, the cell feature type is a preset distribution feature of cells of a preset cell type in a region of a preset region parameter type; the preset cell type includes at least one of CD3+ cells, PanCK+ cells, CD8+ cells and CD3+ / CD8- cells; the preset region parameter type includes at least one of an intra-tumor region, a tumor infiltration region, a tumor infiltration front region, a tumor preset outer expansion range region and a tumor preset inner contraction range region; and the preset distribution feature includes at least one of a density feature, a pattern feature, a quantity proportion feature, a co-localization feature and an abundance feature.

[0010] As an optional implementation, in the first aspect of the present application, the determination of the at least one corresponding prognosis effect prediction model according to the parameter relationship between the type parameters corresponding to the at least two cell feature types includes:

[0011] determining the preset cell type corresponding to the cell feature type corresponding to each type parameter corresponding to the tissue cell image, to obtain a plurality of image cell types corresponding to the tissue cell image;

[0012] determining the image cell types corresponding to at least two type parameters, to obtain multi-feature cell types;

[0013] for each multi-feature cell type, grouping all type parameters corresponding to the multi-feature cell type to obtain a same-type parameter combination;

[0014] inputting each same-type parameter combination into a corresponding fitting algorithm model to obtain a corresponding fitting parameter relationship;

[0015] determining the fitting parameter relationships corresponding to all multi-feature cell types as parameter relationship data corresponding to the tissue cell image;

[0016] determining at least one prognosis effect prediction model from a plurality of candidate prediction models according to the parameter relationship data.

[0017] As an optional implementation, in the first aspect of the present application, the preset distribution characteristics of the cell feature types corresponding to the type parameters in the same type parameter combination are the same or the preset area parameter types are the same; when the preset distribution characteristics of the cell feature types corresponding to the type parameters in the same type parameter combination are the same, the fitting algorithm model is a polynomial fitting model corresponding to the preset distribution characteristics of the same type parameter combination, and the polynomial fitting model is a density polynomial fitting model, a graph parameter polynomial fitting model, a quantity proportion polynomial fitting model, a colocalization polynomial fitting model or an abundance polynomial fitting model; when the preset area parameter types of the cell feature types corresponding to the type parameters in the same type parameter combination are the same, the fitting algorithm model is a trained vector feature extraction neural network, and the vector feature extraction neural network is obtained by only retaining the feature extraction convolutional layer of a preset convolutional neural network after training the convolutional neural network until convergence by using a training data set including a plurality of training cell feature type parameters and corresponding prognosis evaluation labels.

[0018] As an optional implementation, in the first aspect of the present application, the determining of at least one prognosis effect prediction model from a plurality of candidate prediction models according to the parameter relationship data comprises:

[0019] For each candidate prediction model, a plurality of data feature relationships of model training data corresponding to the candidate prediction model are obtained.

[0020] The relationship similarity between each data feature relationship and each fitting parameter relationship of the parameter relationship data is calculated.

[0021] The average value of all the relationship similarities corresponding to all the data feature relationships is calculated to obtain the model matching degree corresponding to the candidate prediction model.

[0022] The candidate prediction model with a model matching degree higher than a preset matching degree threshold is determined as a prognosis effect prediction model.

[0023] As an optional implementation, in the first aspect of the present application, the candidate prediction model is obtained by the following steps:

[0024] A plurality of patient training data including at least two type parameters of cell feature types and corresponding prognosis evaluation data are obtained.

[0025] The patient training data set is divided into training part data and verification part data.

[0026] According to the training part data and the verification part data, a preset prediction model is trained and verified until convergence, so as to obtain the candidate prediction model; the prediction model is a polynomial weighted model, a neural network model or a random forest model;

[0027] Part data in the verification part data whose prediction accuracy is greater than a preset accuracy threshold during verification is determined as preferred data;

[0028] The type parameters in the preferred data are input into the fitting algorithm model, so as to obtain a plurality of data feature relationships corresponding to the candidate prediction model.

[0029] As an optional implementation, in the first aspect of the present application, the plurality of preset cell characteristic types are screened by the following steps:

[0030] A plurality of patient data are obtained; the patient data include patient information and type parameters of a plurality of candidate cell characteristic types and patient prognosis evaluation labels;

[0031] According to a Lasso-cox regression model and a preset evaluation function, an evaluation function value corresponding to each candidate cell characteristic type is calculated according to the plurality of patient data; the evaluation function is a partial likelihood deviation or a C-index;

[0032] According to the evaluation function value, a plurality of target cell characteristic types are screened from all the candidate cell characteristic types.

[0033] As an optional implementation, in the first aspect of the present application, the calculation of the prognosis risk score corresponding to the target patient according to the prognosis effect prediction model and the type parameters corresponding to the cell characteristic type comprises:

[0034] For each prognosis effect prediction model, an associated parameter relationship is screened from all the fitting parameter relationships; the average value of the relationship similarity between the associated parameter relationship and all the data feature relationships corresponding to the prognosis effect prediction model is greater than a preset parameter threshold;

[0035] The type parameters corresponding to the cell characteristic type in all the associated parameter relationships are input into the prognosis effect prediction model, so as to obtain an output model risk score;

[0036] a weighted sum average of model risk scores of all the prognosis effect prediction models is calculated to obtain a prognosis risk score corresponding to the target patient; wherein, a weight corresponding to each model risk score of the prognosis effect prediction models comprises a first weight and a second weight; the first weight is proportional to the model matching degree corresponding to the corresponding prognosis effect prediction model; and the second weight is proportional to the type number of the cell feature types in the parameter relationship between all the associated parameters in the input data corresponding to the corresponding prognosis effect prediction model.

[0037] The second aspect of the embodiments of the present application discloses a prognosis effect prediction system based on multiple features, which comprises:

[0038] a obtaining module, configured to obtain a tissue cell image of a target patient;

[0039] a first determining module, configured to determine type parameters corresponding to a plurality of preset cell feature types by performing feature calculation on the tissue cell image;

[0040] a second determining module, configured to determine at least one corresponding prognosis effect prediction model according to a parameter relationship between the type parameters corresponding to at least two cell feature types;

[0041] a calculating module, configured to calculate a prognosis risk score corresponding to the target patient according to the prognosis effect prediction model and the type parameters corresponding to the cell feature types.

[0042] As an optional implementation, in the second aspect of the present application, the cell feature types are preset distribution features of preset cell types of cells in a preset region parameter type region; the preset cell types comprise at least one of CD3+ cells, PanCK+ cells, CD8+ cells and CD3+ / CD8- cells; the preset region parameter type comprises at least one of an intratumoral region, a tumor infiltrating region, a tumor infiltrating front edge region, a tumor preset outward expansion range region and a tumor preset inward contraction range region; and the preset distribution features comprise at least one of a density feature, a pattern feature, a quantity proportion feature, a co-localization feature and an abundance feature.

[0043] As an optional implementation, in the second aspect of the present application, the second determining module determines at least one corresponding prognosis effect prediction model according to a parameter relationship between the type parameters corresponding to at least two cell feature types, and the specific manner comprises:

[0044] determining the preset cell types corresponding to the cell feature types corresponding to each type parameter of the tissue cell image to obtain a plurality of image cell types corresponding to the tissue cell image;

[0045] determining the image cell types corresponding to at least two of the type parameters, to obtain multi-feature cell types;

[0046] grouping all the type parameters corresponding to each of the multi-feature cell types to obtain same-type parameter combinations;

[0047] inputting each of the same-type parameter combinations into a corresponding fitting algorithm model to obtain a corresponding fitting parameter relationship;

[0048] determining the fitting parameter relationships corresponding to all the multi-feature cell types as the parameter relationship data corresponding to the tissue cell image;

[0049] determining at least one prognostic effect prediction model from a plurality of candidate prediction models according to the parameter relationship data.

[0050] As an optional implementation, in the second aspect of the present application, the type parameters in the same-type parameter combination correspond to the same preset distribution characteristics of the cell feature types or the same preset region parameter types; when the type parameters in the same-type parameter combination correspond to the same preset distribution characteristics of the cell feature types, the fitting algorithm model is a polynomial fitting model corresponding to the preset distribution characteristics of the same-type parameter combination, and the polynomial fitting model is a density polynomial fitting model, a graph parameter polynomial fitting model, a quantity proportion polynomial fitting model, a colocalization polynomial fitting model, or an abundance polynomial fitting model; when the type parameters in the same-type parameter combination correspond to the same preset region parameter types of the cell feature types, the fitting algorithm model is a trained vector feature extraction neural network, and the vector feature extraction neural network is obtained by only retaining the feature extraction convolutional layer of a preset convolutional neural network after training the convolutional neural network until convergence by using a training data set including a plurality of training cell feature type parameters and corresponding prognostic evaluation labels.

[0051] As an optional implementation, in the second aspect of the present application, the specific manner in which the second determining module determines at least one prognostic effect prediction model from a plurality of candidate prediction models according to the parameter relationship data includes:

[0052] for each of the candidate prediction models, obtaining a plurality of data feature relationships of model training data corresponding to the candidate prediction model;

[0053] calculating the relationship similarity between each of the data feature relationships and each of the fitting parameter relationships of the parameter relationship data;

[0054] calculating an average value of all the relationship similarities corresponding to all the data feature relationships to obtain a model matching degree corresponding to the candidate prediction model;

[0055] determining the candidate prediction model with the model matching degree higher than the preset matching degree threshold as the prognosis effect prediction model.

[0056] As an optional implementation form, in the second aspect of the present application, the candidate prediction model is obtained by the following steps:

[0057] obtaining a plurality of patient training data including type parameters of at least two cell feature types and corresponding prognosis evaluation data;

[0058] dividing the patient training data set into training part data and verification part data;

[0059] training and verifying a preset prediction model until convergence according to the training part data and the verification part data to obtain the candidate prediction model; the prediction model is a polynomial weighted model, a neural network model or a random forest model;

[0060] determining part data in the verification part data with a prediction accuracy greater than a preset accuracy threshold during verification as preferred data;

[0061] inputting the type parameters in the preferred data into the fitting algorithm model to obtain a plurality of data feature relationships corresponding to the candidate prediction model.

[0062] As an optional implementation form, in the second aspect of the present application, the plurality of preset cell feature types are obtained by the following steps:

[0063] obtaining a plurality of patient data; the patient data includes patient information and type parameters of a plurality of candidate cell feature types and patient prognosis evaluation labels;

[0064] calculating an evaluation function value corresponding to each candidate cell feature type according to the plurality of patient data according to a Lasso-cox regression model and a preset evaluation function; the evaluation function is a partial likelihood deviation or a C-index;

[0065] selecting a plurality of target cell feature types from all the candidate cell feature types according to the evaluation function value.

[0066] As an optional implementation form, in the second aspect of the present application, the specific manner in which the calculation module calculates the prognosis risk score corresponding to the target patient according to the prognosis effect prediction model and the type parameters corresponding to the cell feature types comprises:

[0067] For each of the prognosis effect prediction models, a relevant parameter relationship is selected from all the fitted parameter relationships; the average of the relationship similarities between the relevant parameter relationship and all the data feature relationships corresponding to the prognosis effect prediction model is greater than a preset parameter threshold value;

[0068] The type parameters corresponding to the cell feature types in all the relevant parameter relationships are input into the prognosis effect prediction model to obtain an output model risk score;

[0069] A weighted sum average of the model risk scores of all the prognosis effect prediction models is calculated to obtain a prognosis risk score corresponding to the target patient; wherein the weight corresponding to the model risk score of each of the prognosis effect prediction models comprises a first weight and a second weight; the first weight is proportional to the model matching degree corresponding to the corresponding prognosis effect prediction model; and the second weight is proportional to the type number of the cell feature types in all the relevant parameter relationships in the input data corresponding to the corresponding prognosis effect prediction model.

[0070] The third aspect of the present application discloses another prognosis effect prediction system based on multiple features, which comprises:

[0071] A memory storing executable program codes;

[0072] A processor coupled with the memory;

[0073] The processor invokes the executable program codes stored in the memory to execute part or all of the steps of the prognosis effect prediction method based on multiple features disclosed in the first aspect of the present application.

[0074] The fourth aspect of the present application discloses a computer storage medium storing computer instructions, which are invoked to execute part or all of the steps of the prognosis effect prediction method based on multiple features disclosed in the first aspect of the present application.

[0075] Compared with the prior art, the embodiments of the present application have the following beneficial effects:

[0076] The application can first determine type parameters corresponding to a plurality of cell feature types based on a plurality of preset cell feature types for feature calculation of tissue cell images, then determine at least one corresponding prognosis effect prediction model through parameter relationships between the type parameters corresponding to the cell feature types, and finally calculate a prognosis risk score corresponding to a patient based on the type parameters corresponding to the cell feature types according to the prognosis effect prediction model, so as to accurately screen out the prognosis effect prediction model by making full use of the relationships between the cell features, realize more accurate prognosis effect prediction, improve the prediction efficiency and accuracy of cancer prognosis effect prediction, and provide accurate data basis for disease treatment or research. BRIEF DESCRIPTION OF DRAWINGS

[0077] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0078] Fig. 1 is a flow diagram of a prognosis effect prediction method based on multiple features disclosed by an embodiment of the present application.

[0079] Fig. 2 is a structural diagram of a prognosis effect prediction system based on multiple features disclosed by an embodiment of the present application.

[0080] Fig. 3 is a structural diagram of another prognosis effect prediction system based on multiple features disclosed by an embodiment of the present application. Embodiment of the present application

[0081] In order to make the person skilled in the art better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely in the following with reference to the drawings of the embodiments of the present application. Obviously, the described embodiments are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0082] The terms "first", "second", etc. in the specification and claims of the present application and the above drawings are used to distinguish different objects, not to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, device, product or equipment including a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, product or equipment.

[0083] Reference to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are they necessarily mutually exclusive of one another. It is expressly understood that any of the embodiments described herein can be combined with any of the other embodiments unless specifically noted otherwise.

[0084] The application discloses a multi-feature-based prognosis effect prediction method and system, which can first determine type parameters corresponding to a plurality of cell feature types based on feature calculation of tissue cell images according to a plurality of preset cell feature types, then determine at least one corresponding prognosis effect prediction model based on parameter relationships between the type parameters corresponding to the cell feature types, and finally calculate a prognosis risk score corresponding to a patient based on the prognosis effect prediction model according to the type parameters corresponding to the cell feature types, so that the relationship between cell features can be fully utilized to accurately screen out the prognosis effect prediction model to realize more accurate prognosis effect prediction, improve the prediction efficiency and accuracy of cancer prognosis effect prediction, and provide accurate data basis for disease treatment or research. The following will be described in detail.

[0085] Embodiment one

[0086] Please refer to FIG. 1, which is a flowchart of a multi-feature-based prognosis effect prediction method disclosed by the embodiment of the application. The multi-feature-based prognosis effect prediction method described in FIG. 1 can be applied to a data processing system / data processing device / data processing server (wherein the server includes a local processing server or a cloud processing server). As shown in FIG. 1, the multi-feature-based prognosis effect prediction method can include the following operations:

[0087] 101, obtaining a tissue cell image of a target patient.

[0088] 102, performing feature calculation on the tissue cell image to determine type parameters corresponding to a plurality of preset cell feature types.

[0089] Specifically, the cell recognition model can be used to recognize the tissue cell image to obtain cell regions and cell types in the image, and then the type parameters corresponding to each cell feature type can be calculated based on the cell regions and cell types.

[0090] 103, determining at least one corresponding prognosis effect prediction model according to parameter relationships between type parameters corresponding to at least two cell feature types.

[0091] 104, calculating a prognosis risk score corresponding to the target patient according to the prognosis effect prediction model and the type parameters corresponding to the cell feature types.

[0092] It can be seen that the above-mentioned embodiments of the application can first determine the type parameters corresponding to a plurality of cell feature types based on a plurality of preset cell feature types for feature calculation of tissue cell images, and then determine at least one corresponding prognosis effect prediction model through the parameter relationship between the type parameters corresponding to the cell feature types, so as to finally calculate the prognosis risk score corresponding to the patient based on the prognosis effect prediction model according to the type parameters corresponding to the cell feature types, thereby being able to make full use of the relationship between the cell features to accurately screen out the prognosis effect prediction model to realize more accurate prognosis effect prediction, improve the prediction efficiency and accuracy of cancer prognosis effect prediction, and provide accurate data basis for disease treatment or research.

[0093] As an optional embodiment, in the above-mentioned steps, the cell feature type is a preset distribution feature of a preset cell type in a preset region parameter type region; the preset cell type includes at least one of CD3+ cells, PanCK+ cells, CD8+ cells and CD3+ / CD8- cells; the preset region parameter type includes at least one of an intratumoral region, a tumor infiltrating region, a tumor infiltrating front region, a tumor preset outward expansion range region and a tumor preset inward contraction range region; and the preset distribution feature includes at least one of a density feature, a pattern feature, a quantity proportion feature, a co-localization feature and an abundance feature.

[0094] Specifically, some examples of the cell feature type can be a plurality of features of CD8+ cells, such as the mean, variance, maximum-minimum value ratio and dispersion of the Voronoi polygon area of CD8+ cells, the mean, variance, maximum-minimum value ratio and dispersion of the number of cells in the tumor outward expansion 1000um region, and the co-localization feature or abundance feature between different types of cells, such as dividing the selected region into fixed-size square regions, calculating the number of first-type cells N1 and the number of second-type cells N2 in each square region, then calculating the first-type cell proportion P1=N1 / (N1+N2) and the second-type cell proportion P2=N2 / (N1+N2) in each square region, and calculating the abundance T and the co-localization feature M according to the two-type cell proportion values of each tile, wherein the calculation formula of the co-localization feature M is:

[0095] ;

[0096] The calculation formula of the abundance T is:

[0097] ;

[0098] Wherein, the superscript i represents the i-th square region, and N represents the number of square regions cut out from the selected region.

[0099] It can be seen that through the above optional embodiments, the content of the cell feature type is limited to realize accurate representation of cell features, and subsequent realization of more accurate prognosis effect prediction, auxiliary realization of full use of the relationship between cell features to accurately screen out a prognosis effect prediction model to realize more accurate prognosis effect prediction, improvement of the prediction efficiency and accuracy of cancer prognosis effect prediction, and provision of accurate data basis for disease treatment or research.

[0100] As an optional embodiment, in the above step, determining the at least one corresponding prognosis effect prediction model according to the parameter relationship between the type parameters corresponding to the at least two cell feature types comprises:

[0101] determining a preset cell type corresponding to a cell feature type corresponding to each type parameter corresponding to the tissue cell image, to obtain a plurality of image cell types corresponding to the tissue cell image;

[0102] determining at least an image cell type corresponding to two type parameters, to obtain a multi-feature cell type;

[0103] For each multi-feature cell type, grouping all type parameters corresponding to the multi-feature cell type to obtain a same-type parameter combination;

[0104] inputting each same-type parameter combination into a corresponding fitting algorithm model to obtain a corresponding fitting parameter relationship;

[0105] determining the fitting parameter relationship corresponding to all multi-feature cell types as the parameter relationship data corresponding to the tissue cell image;

[0106] determining at least one prognosis effect prediction model from the plurality of candidate prediction models according to the parameter relationship data.

[0107] It can be seen that through the above optional embodiments, feature screening can be performed based on patient data and a Lasso-cox regression model to determine a plurality of cell feature types with higher correlation, subsequent realization of more accurate prognosis effect prediction, auxiliary realization of full use of the relationship between cell features to accurately screen out a prognosis effect prediction model to realize more accurate prognosis effect prediction, improvement of the prediction efficiency and accuracy of cancer prognosis effect prediction, and provision of accurate data basis for disease treatment or research.

[0108] As an optional embodiment, in the above step, the preset distribution characteristics of the cell characteristic types corresponding to the type parameters in the same type parameter combination are the same or the preset region parameter types are the same; when the preset distribution characteristics of the cell characteristic types corresponding to the type parameters in the same type parameter combination are the same, the fitting algorithm model is a polynomial fitting model corresponding to the preset distribution characteristics corresponding to the same type parameter combination, and the polynomial fitting model is a density polynomial fitting model, a graph parameter polynomial fitting model, a quantity proportion polynomial fitting model, a colocalization polynomial fitting model or an abundance polynomial fitting model; when the preset region parameter types of the cell characteristic types corresponding to the type parameters in the same type parameter combination are the same, the fitting algorithm model is a trained vector feature extraction neural network, and the vector feature extraction neural network is obtained by only retaining the feature extraction convolutional layer of the preset convolutional neural network after training the convolutional neural network until convergence through a training data set including a plurality of training cell characteristic type parameters and corresponding prognosis evaluation labels.

[0109] It can be seen that, through the above optional embodiments, the details of the cell characteristic relationship between the type parameters and the corresponding parameter relationship fitting algorithm model are determined, so as to accurately analyze the parameter relationship between the type parameters, and subsequently realize accurate risk prediction, thereby assisting in accurately screening the prognosis effect prediction model by fully utilizing the relationship between the cell characteristics to realize more accurate prognosis effect prediction, improving the prediction efficiency and accuracy of cancer prognosis effect prediction, and providing accurate data basis for disease treatment or research.

[0110] As an optional embodiment, in the above step, the at least one prognosis effect prediction model is determined from the plurality of candidate prediction models according to the parameter relationship data, including:

[0111] For each candidate prediction model, a plurality of data characteristic relationships of model training data corresponding to the candidate prediction model are obtained.

[0112] The relationship similarity between each data characteristic relationship and each fitting parameter relationship of the parameter relationship data is calculated.

[0113] The average value of all relationship similarities corresponding to all data characteristic relationships is calculated to obtain a model matching degree corresponding to the candidate prediction model.

[0114] The candidate prediction model with a model matching degree higher than a preset matching degree threshold is determined as the prognosis effect prediction model.

[0115] It can be seen that, through the above optional embodiments, the model matching degree can be determined based on the similarity between the data feature relationship of the model training data of the candidate prediction model and the parameter relationship of the current patient's tissue cell image, so as to screen out a suitable prognosis effect prediction model, and then realize more accurate prognosis effect prediction, thereby assisting in realizing accurate screening of a prognosis effect prediction model by fully utilizing the relationship between cell characteristics to realize more accurate prognosis effect prediction, improving the prediction efficiency and accuracy of cancer prognosis effect prediction, and providing accurate data basis for disease treatment or research.

[0116] As an optional embodiment, in the above step, the candidate prediction model is trained by the following steps:

[0117] Obtain a plurality of patient training data including type parameters of at least two cell characteristic types and corresponding prognosis evaluation data;

[0118] Divide the patient training data set into training part data and verification part data;

[0119] According to the training part data and the verification part data, the preset prediction model is trained and verified until convergence, and a candidate prediction model is obtained; optionally, the prediction model is a polynomial weighted model, a neural network model or a random forest model;

[0120] The part of data in the verification part data whose prediction accuracy during verification is greater than a preset accuracy threshold is determined as preferred data;

[0121] The type parameters in the preferred data are input into a fitting algorithm model to obtain a plurality of data feature relationships corresponding to the candidate prediction model.

[0122] It can be seen that, through the above optional embodiments, the model matching degree can be determined by training the preset model based on the patient training data and fitting and calculating the cell feature relationship in the data set with higher verification effect, so as to realize model screening and accurate prognosis effect prediction, thereby assisting in realizing accurate screening of a prognosis effect prediction model by fully using the relationship between cell characteristics to realize more accurate prognosis effect prediction, improving the prediction efficiency and accuracy for cancer prognosis effect prediction, and providing accurate data basis for disease treatment or research.

[0123] As an optional embodiment, in the above step, the plurality of preset cell characteristic types are screened by the following steps:

[0124] Obtain a plurality of patient data; optionally, the patient data includes patient information and type parameters of a plurality of candidate cell characteristic types and patient prognosis evaluation labels;

[0125] According to the Lasso-cox regression model and the preset evaluation function, the evaluation function value corresponding to each candidate cell feature type is calculated according to multiple patient data; optionally, the evaluation function is partial likelihood deviation or C-index.

[0126] According to the evaluation function value, multiple target cell feature types are screened from all candidate cell feature types.

[0127] In a specific embodiment, a feature screening model is implemented, multiple features are included, and the Lasso-cox regression model is used for feature screening and model training.

[0128] First, the features are binarized for preprocessing, and the median of each feature value in the training set is obtained. If the feature value is greater than or equal to the median, it is set to 1, and if it is less than the median, it is set to 0. The samples of the validation and test sets are also binarized based on the median value of the features in the training set.

[0129] Based on two models for feature screening, one is the DFS model, which uses the Lasso-cox regression model with 10-fold cross-validation on the training set for feature screening, and the evaluation function selected is the partial likelihood deviation value. The features corresponding to the lambda value with the smallest partial likelihood deviation value are selected. The other is the OS model, which uses the Lasso-cox regression model with 10-fold cross-validation on the training set for feature screening, and the evaluation function selected is C-index. The features corresponding to the lambda value at the maximum C-index value are selected.

[0130] As can be seen, through the above optional embodiments, feature screening can be performed based on patient data and the Lasso-cox regression model to determine multiple cell feature types with higher relevance, and subsequent more accurate prognosis effect prediction can be achieved. It helps to accurately screen out the prognosis effect prediction model by fully utilizing the relationship between cell features to achieve more accurate prognosis effect prediction, improve the prediction efficiency and accuracy of cancer prognosis effect prediction, and provide accurate data basis for disease treatment or research.

[0131] As an optional embodiment, in the above steps, the prognosis risk score corresponding to the target patient is calculated according to the prognosis effect prediction model and the type parameter corresponding to the cell feature type, which includes:

[0132] For each prognosis effect prediction model, the associated parameter relationship is screened from all fitted parameter relationships; optionally, the average value of the relationship similarity between the associated parameter relationship and all data feature relationships corresponding to the prognosis effect prediction model is greater than a preset parameter threshold;

[0133] Input the type parameters corresponding to the cell feature types in all the associated parameter relationships into the prognosis effect prediction model to obtain an output model risk score;

[0134] Calculate a weighted sum average of the model risk scores of all the prognosis effect prediction models to obtain a prognosis risk score corresponding to the target patient. Optionally, the weight corresponding to each prognosis effect prediction model's model risk score includes a first weight and a second weight; the first weight is directly proportional to the model matching degree corresponding to the corresponding prognosis effect prediction model; and the second weight is directly proportional to the type quantity of the cell feature types in all the associated parameter relationships in the input data corresponding to the corresponding prognosis effect prediction model.

[0135] As can be seen, through the above optional embodiments, the type parameters input into each prognosis effect prediction model can be screened based on the relationship similarity to achieve model scoring, and the accurate weighted calculation of the scoring based on the weight rules related to the model matching degree and the feature type quantity can be performed to achieve accurate screening of the prognosis effect prediction model based on the relationship between the cell features to achieve more accurate prognosis effect prediction, improve the prediction efficiency and accuracy of cancer prognosis effect prediction, and provide accurate data basis for disease treatment or research.

[0136] Embodiment Two

[0137] Please refer to FIG. 2, which is a structural schematic diagram of a multi-feature-based prognosis effect prediction system disclosed by an embodiment of the present application. The multi-feature-based prognosis effect prediction system described in FIG. 2 can be applied in a data processing system / data processing device / data processing server (wherein the server includes a local processing server or a cloud processing server). As shown in FIG. 2, the multi-feature-based prognosis effect prediction system can include:

[0138] The acquisition module 201 is configured to acquire a tissue cell image of a target patient.

[0139] The first determination module 202 is configured to perform feature calculation on the tissue cell image to determine type parameters corresponding to a plurality of preset cell feature types.

[0140] The second determination module 203 is configured to determine at least one corresponding prognosis effect prediction model according to the parameter relationship between the type parameters corresponding to at least two cell feature types.

[0141] The calculation module 204 is configured to calculate a prognosis risk score corresponding to the target patient according to the prognosis effect prediction model and the type parameters corresponding to the cell feature types.

[0142] It can be seen that the above embodiments of the application can first determine the type parameters corresponding to the multiple cell feature types based on the feature calculation of the tissue cell image based on the multiple preset cell feature types, and then determine at least one corresponding prognosis effect prediction model based on the parameter relationship between the type parameters corresponding to the cell feature types, so as to finally calculate the prognosis risk score corresponding to the patient based on the type parameters corresponding to the cell feature types based on the prognosis effect prediction model, thereby being capable of fully utilizing the relationship between the cell features to accurately screen the prognosis effect prediction model to realize more accurate prognosis effect prediction, improving the prediction efficiency and accuracy of the cancer prognosis effect prediction, and providing accurate data basis for disease treatment or research.

[0143] As an optional embodiment, the cell feature type is a preset distribution feature of cells of a preset cell type in a region of a preset region parameter type; the preset cell type includes at least one of CD3+ cells, PanCK+ cells, CD8+ cells, and CD3+ / CD8- cells; the preset region parameter type includes at least one of an intratumoral region, a tumor infiltrating region, a tumor infiltrating front region, a tumor preset outward expansion range region, and a tumor preset inward contraction range region; and the preset distribution feature includes at least one of a density feature, a pattern feature, a quantity proportion feature, a co-localization feature, and an abundance feature.

[0144] It can be seen that by the above optional embodiment, the content of the cell feature type is limited to realize accurate representation of the cell feature, and more accurate prognosis effect prediction is realized, thereby assisting in realizing full utilization of the relationship between the cell features to accurately screen the prognosis effect prediction model to realize more accurate prognosis prediction, improving the prediction efficiency and accuracy of the cancer prognosis effect prediction, and providing accurate data for disease treatment or research.

[0145] As an optional embodiment, the second determining module determines the specific manner of the at least one corresponding prognosis effect prediction model according to the parameter relationship between the type parameters corresponding to the at least two cell feature types, including:

[0146] determining the preset cell type corresponding to the cell feature type corresponding to each type parameter of the tissue cell image to obtain multiple image cell types corresponding to the tissue cell image;

[0147] determining at least an image cell type corresponding to two type parameters to obtain a multi-feature cell type;

[0148] for each multi-feature cell type, grouping all type parameters corresponding to the multi-feature cell type to obtain a same-type parameter combination;

[0149] inputting each same-type parameter combination into a corresponding fitting algorithm model to obtain a corresponding fitting parameter relationship;

[0150] The fitting parameter relationships corresponding to all multi-feature cell types are determined as the parameter relationship data corresponding to the tissue cell images;

[0151] Based on the parameter relationship data, at least one prognostic effect prediction model is determined from multiple candidate prediction models.

[0152] As can be seen, through the above optional embodiments, the model matching degree can be determined based on the similarity between the data feature relationship of the model training data of the candidate prediction model and the parameter relationship of the current patient's tissue cell image, so as to screen out a suitable prognostic effect prediction model, and subsequently achieve more accurate prognostic effect prediction. This helps to make full use of the relationship between cell features to accurately screen out the prognostic effect prediction model, thereby improving the prediction efficiency and accuracy of cancer prognostic effect prediction and providing an accurate data foundation for disease treatment or research.

[0153] As an optional embodiment, the preset distribution features of cell feature types corresponding to the type parameters in the same type parameter combination are the same, or the preset region parameter types are the same. When the preset distribution features of cell feature types corresponding to the type parameters in the same type parameter combination are the same, the fitting algorithm model is a polynomial fitting model corresponding to the preset distribution features of the same type parameter combination. The polynomial fitting model is a density polynomial fitting model, a graph parameter polynomial fitting model, a quantity proportion polynomial fitting model, a co-localization polynomial fitting model, or an abundance polynomial fitting model. When the preset region parameter types of cell feature types corresponding to the type parameters in the same type parameter combination are the same, the fitting algorithm model is a trained vector feature extraction neural network. The preset convolutional neural network is trained on a training dataset including multiple training cell feature type parameters and corresponding prognostic evaluation annotations until convergence. Only the feature extraction convolutional layer of the convolutional neural network is retained to obtain the vector feature extraction neural network.

[0154] As can be seen, the above optional embodiments clarify the details of the cell feature relationships between type parameters and the corresponding parameter relationship fitting algorithm model, so as to accurately analyze the parameter relationships between type parameters, and subsequently achieve accurate risk prediction. This helps to fully utilize the relationships between cell features to accurately screen out prognostic effect prediction models to achieve more accurate prognostic effect prediction, improve the prediction efficiency and accuracy of cancer prognostic effect prediction, and provide an accurate data foundation for disease treatment or research.

[0155] As an optional embodiment, the second determining module determines at least one prognostic effect prediction model from multiple candidate prediction models based on parameter relationship data in the following specific manner:

[0156] For each candidate prediction model, obtain the relationships between multiple data features in the model training data corresponding to that candidate prediction model;

[0157] Calculate the similarity between each data feature relationship and parameter relationship, and the similarity between each fitted parameter relationship of the data.

[0158] Calculate the average similarity of all relationships corresponding to all data feature relationships to obtain the model matching degree corresponding to the candidate prediction model;

[0159] Candidate prediction models with a model matching degree higher than the preset matching degree threshold are identified as prognostic effect prediction models.

[0160] As can be seen, through the above optional embodiments, feature screening can be performed based on patient data and Lasso-cox regression models to identify multiple cell feature types with higher correlation, thereby achieving more accurate prognostic effect prediction. This helps to fully utilize the relationship between cell features to accurately screen prognostic effect prediction models, thereby improving the prediction efficiency and accuracy of cancer prognostic effect prediction and providing an accurate data foundation for disease treatment or research.

[0161] As an optional implementation, the candidate prediction model is trained through the following steps:

[0162] Acquire multiple patient training data, including type parameters with at least two cell feature types and corresponding prognostic assessment data;

[0163] The patient training dataset is divided into training data and validation data.

[0164] Based on the training data and validation data, the preset prediction model is trained and validated until convergence to obtain candidate prediction models; optionally, the prediction model is a multinomial weighted model, a neural network model or a random forest model.

[0165] The data in the validation dataset that has a prediction accuracy greater than a preset accuracy threshold during validation is selected as preferred data.

[0166] The type parameters from the selected data are input into the fitting algorithm model to obtain the relationships between multiple data features corresponding to the candidate prediction model.

[0167] As can be seen, through the above optional embodiments, a preset model can be trained based on patient training data and the cell feature relationships in the dataset with higher validation effect can be fitted and calculated to facilitate model screening and accurate prognostic effect prediction. This helps to fully utilize the relationships between cell features to accurately screen prognostic effect prediction models to achieve more accurate prognostic effect prediction, improve the prediction efficiency and accuracy of cancer prognostic effect prediction, and provide an accurate data foundation for disease treatment or research.

[0168] As an optional embodiment, multiple preset cell feature types are obtained through the following steps:

[0169] Acquire data from multiple patients; optionally, the patient data may include patient information, type parameters of multiple candidate cell feature types, and patient prognostic assessment annotations.

[0170] Based on the Lasso-cox regression model and the preset evaluation function, the evaluation function value corresponding to each candidate cell feature type is calculated based on multiple patient data; optionally, the evaluation function is partial likelihood deviation or C-index.

[0171] Based on the evaluation function value, multiple target cell feature types are selected from all candidate cell feature types.

[0172] As can be seen, through the above optional embodiments, feature screening can be performed based on patient data and Lasso-cox regression models to identify multiple cell feature types with higher correlation, thereby achieving more accurate prognostic effect prediction. This helps to fully utilize the relationship between cell features to accurately screen prognostic effect prediction models, thereby improving the prediction efficiency and accuracy of cancer prognostic effect prediction and providing an accurate data foundation for disease treatment or research.

[0173] As an optional embodiment, the calculation module calculates the prognostic risk score for the target patient based on the prognostic outcome prediction model and the type parameters corresponding to the cell feature types in a specific way, including:

[0174] For each prognosis prediction model, the associated parameter relationships are selected from all fitted parameter relationships; optionally, the average similarity between the associated parameter relationship and all data feature relationships corresponding to the prognosis prediction model is greater than a preset parameter threshold.

[0175] Input the type parameters corresponding to the cell feature types in all the correlation parameters into the prognostic effect prediction model to obtain the output model risk score;

[0176] Calculate the weighted average of the model risk scores of all prognostic effect prediction models to obtain the prognostic risk score corresponding to the target patient; optionally, the weights corresponding to the model risk score of each prognostic effect prediction model include a first weight and a second weight; the first weight is proportional to the model matching degree of the corresponding prognostic effect prediction model; the second weight is proportional to the number of cell feature types in all correlation parameter relationships in the input data of the corresponding prognostic effect prediction model.

[0177] As can be seen, through the above optional embodiments, the type parameters of the input corresponding to each prognostic effect prediction model can be screened based on the similarity of relationships to achieve model scoring. Then, the scoring is accurately weighted based on the weight rules related to the model matching degree and the number of feature types. This fully utilizes the relationships between cell features to accurately screen prognostic effect prediction models to achieve more accurate prognostic effect prediction, improve the prediction efficiency and accuracy of cancer prognostic effect prediction, and provide an accurate data foundation for disease treatment or research.

[0178] Example 3

[0179] Please refer to Figure 3, which illustrates another prognostic prediction system based on multiple features disclosed in this embodiment of the invention. The prognostic prediction system based on multiple features described in Figure 3 is applied in a data processing system / data processing equipment / data processing server (wherein, the server includes a local processing server or a cloud processing server). As shown in Figure 3, this prognostic prediction system based on multiple features may include:

[0180] Memory 301 storing executable program code;

[0181] Processor 302 coupled to memory 301;

[0182] The processor 302 calls the executable program code stored in the memory 301 to execute the steps of the prognostic effect prediction method based on multiple features described in Embodiment 1.

[0183] Example 4

[0184] This invention discloses a computer read storage medium that stores a computer program for electronic data interchange, wherein the computer program causes a computer to execute the steps of the multi-feature-based prognostic prediction method described in Embodiment 1.

[0185] Example 5

[0186] This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps of the multi-feature-based prognostic effect prediction method described in Embodiment 1.

[0187] The foregoing has described specific embodiments of this specification; other embodiments are within the scope of the appended claims. In some cases, the actions or steps described in the claims may be performed in a different order than those shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily have to follow the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0188] The systems, devices, modules, or units described in the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, a computer can be, for example, a personal computer, laptop computer, cellular phone, camera phone, smartphone, personal digital assistant, media player, navigation device, email device, game console, tablet computer, wearable device, or any combination of these devices.

[0189] For ease of description, the above devices are described in terms of function, divided into various units. Of course, in implementing this specification, the functions of each unit can be implemented in one or more software and / or hardware components.

[0190] Those skilled in the art will understand that the embodiments of this specification can be provided as methods, systems, or computer program products. Therefore, the embodiments of this specification can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the embodiments of this specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0191] This specification is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this specification. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more flowchart illustrations and / or one or more block diagrams.

[0192] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0193] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0194] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0195] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0196] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0197] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0198] This specification can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This specification can also be practiced in distributed computing environments, where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0199] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to interchangeably. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0200] Finally, it should be noted that the prognostic effect prediction method and system based on multiple features disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A prognostic prediction method based on multiple features, characterized in that, The method includes: Acquire tissue cell images of the target patient; Feature calculations are performed on the tissue cell image to determine type parameters corresponding to multiple preset cell feature types; Based on the parameter relationship between the type parameters corresponding to at least two of the cell feature types, at least one corresponding prognostic effect prediction model is determined; Based on the prognostic effect prediction model and the type parameters corresponding to the cell feature types, the prognostic risk score corresponding to the target patient is calculated.

2. The prognostic effect prediction method based on multiple features according to claim 1, characterized in that, The cell feature type is a preset distribution feature of cells of a preset cell type within a preset region parameter type region; the preset cell type includes at least one of CD3+ cells, PanCK+ cells, CD8+ cells, and CD3+ / CD8- cells; the preset region parameter type includes at least one of the following: tumor intratumoral region, tumor infiltration region, tumor infiltration frontal region, preset tumor expansion range region, and preset tumor contraction range region; the preset distribution feature includes at least one of the following: density feature, pattern feature, quantity proportion feature, colocalization feature, and abundance feature.

3. The prognostic effect prediction method based on multiple features according to claim 1, characterized in that, The step of determining at least one corresponding prognostic prediction model based on the parameter relationship between type parameters corresponding to at least two of the cell feature types includes: Determine the preset cell type corresponding to the cell feature type corresponding to each type parameter of the tissue cell image, so as to obtain multiple image cell types corresponding to the tissue cell image; The image cell types corresponding to at least two of the type parameters are determined to obtain multi-feature cell types; For each of the multi-feature cell types, all the type parameters corresponding to the multi-feature cell type are grouped to obtain the same type parameter combination; Each of the same type of parameter combinations is input into the corresponding fitting algorithm model to obtain the corresponding fitting parameter relationship; The fitting parameter relationships corresponding to all the multi-feature cell types are determined as the parameter relationship data corresponding to the tissue cell image; Based on the parameter relationship data, at least one prognostic effect prediction model is determined from multiple candidate prediction models.

4. The prognostic effect prediction method based on multiple features according to claim 3, characterized in that, In the same type parameter combination, the preset distribution features of the cell feature types corresponding to the type parameters are the same, or the preset region parameter types are the same; when the preset distribution features of the cell feature types corresponding to the type parameters in the same type parameter combination are the same, the fitting algorithm model is a polynomial fitting model corresponding to the preset distribution features of the same type parameter combination, and the polynomial fitting model is a density polynomial fitting model, a graph parameter polynomial fitting model, a quantity proportion polynomial fitting model, a co-localization polynomial fitting model, or an abundance polynomial fitting model; when the preset region parameter types of the cell feature types corresponding to the type parameters in the same type parameter combination are the same, the fitting algorithm model is a trained vector feature extraction neural network, which is trained on a preset convolutional neural network until convergence using a training dataset including multiple training cell feature type parameters and corresponding prognostic evaluation annotations, and only the feature extraction convolutional layer of the convolutional neural network is retained to obtain the vector feature extraction neural network.

5. The prognostic effect prediction method based on multiple features according to claim 3, characterized in that, The step of determining at least one prognostic effect prediction model from multiple candidate prediction models based on the parameter relationship data includes: For each candidate prediction model, obtain multiple data feature relationships of the model training data corresponding to that candidate prediction model; Calculate the similarity between each of the data feature relationships and each of the fitted parameter relationships of the parameter relationship data; Calculate the average of the similarities of all the relationships corresponding to all the data feature relationships to obtain the model matching degree corresponding to the candidate prediction model; Candidate prediction models whose model matching degree is higher than a preset matching degree threshold are identified as prognostic effect prediction models.

6. The prognostic effect prediction method based on multiple features according to claim 5, characterized in that, The candidate prediction model is trained through the following steps: Acquire multiple patient training data, including type parameters of at least two of the said cell feature types and corresponding prognostic assessment data; The patient training dataset is divided into training data and validation data. Based on the training data and the validation data, the preset prediction model is trained and validated until convergence to obtain the candidate prediction model; the prediction model is a multinomial weighted model, a neural network model, or a random forest model. The portion of the verification data whose prediction accuracy during verification is greater than a preset accuracy threshold is identified as preferred data. The type parameters in the preferred data are input into the fitting algorithm model to obtain multiple data feature relationships corresponding to the candidate prediction model.

7. The prognostic effect prediction method based on multiple features according to claim 1, characterized in that, The multiple preset cell feature types are obtained through the following steps: Acquire data from multiple patients; the patient data includes patient information, type parameters of multiple candidate cell feature types, and patient prognostic assessment annotations; Based on the Lasso-cox regression model and a preset evaluation function, the evaluation function value corresponding to each candidate cell feature type is calculated based on the multiple patient data; the evaluation function is partial likelihood deviation or C-index. Based on the evaluation function value, multiple target cell feature types are selected from all the candidate cell feature types.

8. The prognostic effect prediction method based on multiple features according to claim 5, characterized in that, The step of calculating the prognostic risk score for the target patient based on the prognostic outcome prediction model and the type parameters corresponding to the cell feature types includes: For each of the prognostic prediction models, the associated parameter relationships are selected from all the fitted parameter relationships; the average value of the relationship similarity between the associated parameter relationship and all the data feature relationships corresponding to the prognostic prediction model is greater than a preset parameter threshold. Input the type parameters corresponding to the cell feature types in all the aforementioned correlation parameter relationships into the prognostic effect prediction model to obtain the output model risk score; Calculate the weighted summation average of the model risk scores of all the prognostic effect prediction models to obtain the prognostic risk score corresponding to the target patient; wherein, the weight corresponding to the model risk score of each prognostic effect prediction model includes a first weight and a second weight; the first weight is proportional to the model matching degree corresponding to the corresponding prognostic effect prediction model; the second weight is proportional to the number of cell feature types in all the correlation parameter relationships in the input data corresponding to the corresponding prognostic effect prediction model.

9. A prognostic effect prediction system based on multiple features, characterized in that, The system includes: The acquisition module is used to acquire tissue cell images of the target patient; The first determining module is used to perform feature calculations on the tissue cell image and determine type parameters corresponding to multiple preset cell feature types. The second determining module is used to determine at least one corresponding prognostic effect prediction model based on the parameter relationship between the type parameters corresponding to at least two of the cell feature types. The calculation module is used to calculate the prognostic risk score corresponding to the target patient based on the prognostic effect prediction model and the type parameters corresponding to the cell feature type.

10. A prognostic effect prediction system based on multiple features, characterized in that, The system includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the prognostic effect prediction method based on multiple features as described in any one of claims 1-8.

Images