Cancer classification method and device based on quantum heuristic evolutionary algorithm

Abstract

The invention provides a cancer classification method and device based on a quantum heuristic evolutionary algorithm, and the method comprises the steps: obtaining a gene feature training set, a cancer category training set, a gene feature test set, and a cancer category test set, according to the gene feature training set and the cancer category training set, determining an initial corresponding relation between the gene features and the cancer categories; according to the gene feature test set, determining a gene feature sub-test set meeting a preset accuracy evaluation condition through a quantum heuristic evolutionary algorithm; iterating the initial corresponding relation according to the gene feature sub-test set and the cancer category test set to determine a target corresponding relation; and obtaining a current gene feature sample, and determining a current cancer category according to the current gene feature sample and the target corresponding relationship. According to the method, gene features are screened through the quantum heuristic evolutionary algorithm, and the cancer classification accuracy is improved.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a cancer classification method and device based on a quantum-inspired evolutionary algorithm. Background Art

[0002] Feature selection and classification methods are key steps and uses in machine learning and statistical analysis. Their purpose is to support more efficient decision-making and analysis by screening out the most meaningful features for the model and removing irrelevant, noisy, and redundant features to reduce the dimensionality of the feature space, improve the performance of model classification, reduce computational costs, and enhance the interpretability of results.

[0003] Feature selection and classification can be applied to cancer classification in the medical field, relying on subjective interpretation of clinical and histopathological information, with the goal of classifying tumors into currently recognized categories based on their tissue of origin.

[0004] However, clinical information may be incomplete or misleading, gene expression profile feature samples may contain a large amount of noise features, and existing artificial intelligence models cannot accurately identify genes that can distinguish cancer types. Summary of the Invention

[0005] In view of the above problems, the present application is proposed to provide a method and apparatus for cancer classification based on a quantum-inspired evolutionary algorithm that overcomes or at least partially solves the above problems, including: A cancer classification method based on a quantum-inspired evolutionary algorithm, the method comprising: Obtaining a gene signature training set, a cancer category training set, a gene signature test set, and a cancer category test set, and determining an initial correspondence between gene signatures and cancer categories based on the gene signature training set and the cancer category training set; Determining a gene feature subtest set that meets a preset accuracy evaluation condition based on the gene feature test set by using a quantum-inspired evolutionary algorithm; Iteratively determining a target correspondence relationship based on the gene signature subtest set and the cancer category test set on the initial correspondence relationship; A current gene signature sample is obtained, and a current cancer category is determined based on a correspondence between the current gene signature sample and the target.

[0006] Furthermore, the step of determining a gene feature sub-test set that meets a preset accuracy evaluation condition by using a quantum-inspired evolutionary algorithm based on the gene feature test set includes: Determining a plurality of corresponding qubit individuals based on the genetic feature test set; Selecting an optimal quantum population based on the plurality of individual qubits; A gene feature subtest set that meets a preset accuracy evaluation condition is determined based on the optimal quantum population.

[0007] Furthermore, the step of selecting the optimal quantum population based on the plurality of individual qubits includes: According to the plurality of individual qubits Update the plurality of individual qubits ; Where t represents the number of iterations; According to the plurality of individual qubits Select multiple optimal quantum populations ; According to the multiple optimal quantum populations Selecting the optimal quantum population .

[0008] Furthermore, the plurality of qubits are individually Update the plurality of individual qubits The steps include: According to the plurality of individual qubits Through the quantum revolving door Update the plurality of individual qubits ; Among them, the individual qubits Expressed as:

[0009] and Respectively represent the probability weight values ​​of different gene features in the gene feature test set; Among them, the quantum revolving gate Expressed as:

[0010] is the rotation angle value of the quantum rotating gate, Used to update individual qubits The probability weight value of .

[0011] Furthermore, the plurality of qubits are individually Select multiple optimal quantum populations The steps include: According to the plurality of individual qubits Select n individual qubits; Determine multiple optimal quantum populations based on the combination of n quantum bit individuals ; in,

[0012] n is the size of the quantum population.

[0013] Furthermore, the method is used to determine the correspondence between gene features and cancer categories using a CNN model; the method also includes: Determining the classification accuracy of the CNN model based on the gene signature subtest set; Determining a fitness function corresponding to the gene feature subtest set according to the classification accuracy, the number of the gene feature subtest sets, and the preset accuracy evaluation condition; A gene feature subtest set that satisfies an evaluation condition of the fitness function is determined based on the gene feature test set.

[0014] Furthermore, the step of iteratively determining the target correspondence relationship based on the gene signature subtest set and the cancer category test set comprises: Determining, in the cancer category test set according to the gene signature subtest set, a cancer category subtest set corresponding to the gene signature subtest set; The initial correspondence is iteratively determined based on the gene signature sub-test set and the cancer category sub-test set to determine a target correspondence.

[0015] A cancer classification device based on a quantum-inspired evolutionary algorithm, comprising: an initial correspondence determination module, configured to obtain a gene feature training set, a cancer category training set, a gene feature test set, and a cancer category test set, and determine an initial correspondence between gene features and cancer categories based on the gene feature training set and the cancer category training set; A gene feature subtest set determination module is used to determine a gene feature subtest set that meets a preset accuracy evaluation condition based on the gene feature test set through a quantum-inspired evolutionary algorithm; a target correspondence determination module, configured to iteratively determine a target correspondence for the initial correspondence based on the gene feature subtest set and the cancer category test set; The current cancer category determination module is used to obtain a current gene feature sample and determine the current cancer category based on the corresponding relationship between the current gene feature sample and the target.

[0016] A computer device comprises a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the method described in any embodiment of the present application.

[0017] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in any embodiment of the present application is implemented.

[0018] This application has the following advantages: In an embodiment of the present application, in view of the fact that existing artificial intelligence models cannot accurately identify genes that can distinguish types of cancer, the present application provides a solution for training artificial intelligence models by screening features through a quantum-inspired evolutionary algorithm, specifically: obtaining a gene feature training set, a cancer category training set, a gene feature test set, and a cancer category test set, and determining the initial correspondence between gene features and cancer categories based on the gene feature training set and the cancer category training set; determining a gene feature subtest set that meets preset accuracy evaluation conditions based on the gene feature test set through a quantum-inspired evolutionary algorithm; iterating the initial correspondence to determine the target correspondence based on the gene feature subtest set and the cancer category test set; obtaining a current gene feature sample, and determining the current cancer category based on the current gene feature sample and the target correspondence. The present application improves the accuracy of cancer classification by screening gene features through a quantum-inspired evolutionary algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for the description of the present application. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0020] Figure 1 This is a flowchart of a cancer classification method based on a quantum-inspired evolutionary algorithm provided in one embodiment of the present application; Figure 2 This is a flowchart of feature selection using a quantum-inspired evolutionary algorithm in one embodiment of the present application; Figure 3 This is a flowchart of model training using a quantum fusion model feature selection method in one embodiment of the present application; Figure 4 This is a comparison table of the accuracy and number of feature selections between the quantum fusion model feature selection method and the traditional genetic algorithm in one embodiment of the present application; Figure 5 This is a comparison table of the average selection percentage of each gene feature between the quantum fusion model feature selection method and the traditional genetic algorithm in one embodiment of the present application; Figure 6 This is a structural block diagram of a cancer classification device based on a quantum-inspired evolutionary algorithm provided in one embodiment of the present application; Figure 7 This is a structural block diagram of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0021] To make the objectives, features, and advantages of this application more readily apparent, the present application is further described below in conjunction with the accompanying drawings and specific embodiments. It is apparent that the embodiments described are only a portion of the embodiments of this application, not all of them. All other embodiments derived by persons of ordinary skill in the art based on the embodiments in this application without inventive effort are also within the scope of protection of this application.

[0022] By analyzing the prior art, the inventors found that the morphological range of cancer is very wide, and many tumors are atypical or lack morphological features that are helpful for differential diagnosis. Direct classification based on tumor morphology may lead to inaccurate cancer classification results, prompting the requirement for mandatory second opinions in surgical pathology cases, which may hinder patient care, as well as lead to increased costs and confusion in clinical trial results.

[0023] In recent years, tumor gene expression profiling based on deoxyribonucleic acid (DNA) microarrays has been used for cancer identification and classification. However, because characteristic molecular markers for most solid tumors have yet to be identified, research has been limited to a few cancer types, and comparisons between different datasets are complex. The feasibility of gene expression signature-based cancer identification and classification for common tumors using a single reference database remains unexplored. Commonly used methods, such as genetic algorithms, have poor scalability when dealing with high-dimensional feature data. Furthermore, because genetic algorithms aim to find the feature combination that achieves the best results during model training, doing so requires extensive computing resources and is often unable to perform global spatial searches, preventing them from effectively and comprehensively identifying valid feature variables.

[0024] Furthermore, due to the significant noise inherent in gene microarray data, accurately identifying genes that distinguish tumor types and reducing classification errors is a challenging problem. Furthermore, the volume of genetically encoded data across tumor types is enormous. Previous studies have found that genetic algorithms (GAs) have slow feature selection convergence, poor local search capabilities, and are difficult to implement for large-scale training samples, requiring significant time.

[0025] As a result, the complexity of feature selection increases exponentially as the size of a dataset and the number of features increase. Therefore, comprehensive feature selection requires exhaustive searches that are both time-consuming and costly. Effective feature selection and classification have become a crucial issue in machine learning. Whether in applications involving vast amounts of data across diverse domains or in prediction tasks requiring high precision, feature selection and classification are essential.

[0026] Based on the above systematic analysis of existing technologies, one of the core technical concepts of this application is to address the aforementioned issues by employing a gene feature selection and cancer classification method based on the fusion of a convolutional neural network model (CNN) and a quantum-inspired evolutionary algorithm (QIEA). This method enables more efficient selection of gene features for cancer identification and classification. QIEA is an evolutionary algorithm inspired by the principles of quantum mechanics (such as quantum superposition and quantum entanglement) and runs on a classical computer rather than a quantum computer. It enhances the search capabilities of traditional evolutionary algorithms by simulating quantum behavior (such as quantum bit encoding and quantum gate operations).

[0027] Reference Figure 1 , shows a cancer classification method based on a quantum-inspired evolutionary algorithm provided by an embodiment of the present application, the method comprising: S110, obtaining a gene feature training set, a cancer category training set, a gene feature test set, and a cancer category test set, and determining an initial correspondence between gene features and cancer categories based on the gene feature training set and the cancer category training set; S120, determining a gene feature subtest set that meets a preset accuracy evaluation condition based on the gene feature test set by using a quantum-inspired evolutionary algorithm; S130, iterating the initial correspondence relationship to determine a target correspondence relationship based on the gene feature subtest set and the cancer category test set; S140: Obtain a current gene feature sample, and determine a current cancer category based on the corresponding relationship between the current gene feature sample and the target.

[0028] In an embodiment of the present application, in view of the fact that existing artificial intelligence models cannot accurately identify genes that can distinguish types of cancer, the present application provides a solution for training artificial intelligence models by screening features through a quantum-inspired evolutionary algorithm, specifically: obtaining a gene feature training set, a cancer category training set, a gene feature test set, and a cancer category test set, and determining the initial correspondence between gene features and cancer categories based on the gene feature training set and the cancer category training set; determining a gene feature subtest set that meets preset accuracy evaluation conditions based on the gene feature test set through a quantum-inspired evolutionary algorithm; iterating the initial correspondence to determine the target correspondence based on the gene feature subtest set and the cancer category test set; obtaining a current gene feature sample, and determining the current cancer category based on the current gene feature sample and the target correspondence. The present application improves the accuracy of cancer classification by screening gene features through a quantum-inspired evolutionary algorithm.

[0029] Next, a cancer classification method based on a quantum-inspired evolutionary algorithm in this exemplary embodiment will be further described.

[0030] As described in step S110, a gene feature training set, a cancer category training set, a gene feature test set, and a cancer category test set are obtained, and an initial correspondence between gene features and cancer categories is determined based on the gene feature training set and the cancer category training set.

[0031] It should be noted that the gene signature training set, cancer category training set, gene signature test set, and cancer category test set can be generated from a high-dimensional sample set using a k-fold cross-validation approach. The obtained sample set can be represented by a microarray expression matrix, where the gene signatures are rows and the corresponding tumor samples are columns. The correspondence between the gene signatures and the cancer categories can be represented using a convolutional neural network (CNN) model.

[0032] As an example, a sample set can be obtained from 218 gene expression profiles, which can include 16,063 gene features and corresponding 14 common human cancer categories, covering 14 common cancer categories: breast adenocarcinoma, prostate adenocarcinoma, lung adenocarcinoma, colorectal adenocarcinoma, lymphoma, bladder transitional cell carcinoma, melanoma, uterine adenocarcinoma, leukemia, renal cell carcinoma, pancreatic cancer, ovarian adnexal cancer, pleural mesothelioma and brain glioblastoma.

[0033] As described in step S120, a gene feature sub-test set that meets a preset accuracy evaluation condition is determined based on the gene feature test set through a quantum-inspired evolutionary algorithm.

[0034] It should be noted that given the typically large number of features and complex data in gene microarrays, quantum-inspired evolutionary algorithms can identify the feature dimensions with the greatest discriminatory power to simplify the feature space and obtain the optimal feature subset, ensuring optimal performance of the entire solution. The preset accuracy assessment condition can be a predefined feature subset convergence criterion, which can be represented by the classification accuracy of the target domain.

[0035] As described in step S130 , the initial correspondence is iteratively determined based on the gene feature subtest set and the cancer category test set to determine a target correspondence.

[0036] It should be noted that this application selects gene features that are strongly correlated with the target cancer category in a high-dimensional gene feature matrix space to train the artificial intelligence model, thereby determining and iterating the correspondence between gene features and cancer categories.

[0037] As described in step S140 , a current gene feature sample is obtained, and a current cancer category is determined based on the corresponding relationship between the current gene feature sample and the target.

[0038] It should be noted that cancer can be classified based on the current gene feature samples through the iterated convolutional neural network model.

[0039] Compared with the feature selection process of traditional genetic algorithms, the gene features with high overlap in the early stage in this application are not easily eliminated during the iteration process and are not prone to falling into local extreme values. It can more effectively find the optimal gene feature subset in the high-dimensional gene data space, reduce training complexity, and achieve better classification accuracy.

[0040] Reference Figure 2 In one embodiment of the present application, the specific process of "determining a gene feature sub-test set that meets the preset accuracy evaluation conditions through a quantum-inspired evolutionary algorithm based on the gene feature test set" in step S120 can be further explained in combination with the following description.

[0041] As described in the following steps, a plurality of corresponding qubit individuals are determined based on the genetic feature test set; Selecting an optimal quantum population based on the plurality of individual qubits; A gene feature subtest set that meets a preset accuracy evaluation condition is determined based on the optimal quantum population.

[0042] It should be noted that, compared to traditional genetic algorithms that usually use binary values ​​to represent individuals in a population, the quantum-inspired evolutionary algorithm of the present application uses a matrix composed of multiple qubits to represent qubit individuals; each quantum population corresponds to multiple groups of genetic features in the genetic feature test set, each qubit individual corresponds to a group of genetic features in the genetic feature test set, and each qubit corresponds to a different genetic feature in the genetic feature test set; The qubit individual includes multiple qubits, each of which represents the probability of a genetic feature being selected. By using different qubits to represent multiple states of a linear combination of probabilities, each qubit individual can be sampled with equal probability in the global search space, thereby representing more feature vector dimensions, and then using a smaller number of individuals to find the optimal genetic feature subset that meets the preset accuracy assessment conditions. At the same time, it can also avoid the easy elimination of certain genetic features in the early stage, and can retain as many features as possible that are highly redundant but indeed relevant, thereby ensuring the population diversity of the feature screening results.

[0043] Reference Figure 2 In one embodiment of the present application, the specific process of "selecting the optimal quantum population based on the multiple individual quantum bits" can be further explained in combination with the following description.

[0044] As described in the following steps, based on the plurality of qubit individuals Update the plurality of individual qubits ; Where t represents the number of iterations; According to the plurality of individual qubits Select multiple optimal quantum populations ; According to the multiple optimal quantum populations Selecting the optimal quantum population .

[0045] It should be noted that individual qubits It can be expressed as:

[0046] and They represent the probability weight values ​​of different gene features in the gene feature test set, which are 0 or 1. That is, the probability weight corresponding to quantum bit i can be expressed as ; where i is 1, 2, ..., m, and j is 1, 2, ..., n; where m represents the number of genetic features, n is the size of the quantum population, and t represents the number of iterations of the quantum bit individual.

[0047] Since different genetic features are defined as multiple binary vectors in individual qubits, each qubit corresponds to the probability of a genetic feature being selected. By using linear superposition probability combinations, a wider range of spatial regions can be evaluated for search, rather than the single-point search of the genetic algorithm. This more effectively guides the optimal convergence area of ​​the search space, that is, when selecting a feature, the probability is used to determine whether to retain or discard the feature. Each individual qubit can represent more feature vector dimensions with probabilistic sampling in the global search space. Therefore, this application only requires fewer individual definitions to find the optimal feature subset. At the same time, this application selects multiple better quantum populations first. , and then select the optimal quantum population The secondary feature selection ensures that the features are selected globally to avoid falling into local optimality.

[0048] However, using traditional genetic algorithms for feature selection requires defining the number of individuals in advance, placing all features in a multidimensional space, and using the pre-set number of individuals to find the optimal subset. When the number of variables is large, a relatively large number of individuals needs to be set. In addition, during the iteration process, the early features are prone to high overlap and are eliminated prematurely, which can easily lead to falling into local extreme values.

[0049] Reference Figure 2 In one embodiment of the present application, the following description can be combined to further illustrate "according to the plurality of qubit individuals Update the plurality of individual qubits "The specific process.

[0050] As described in the following steps, based on the plurality of qubit individuals Through the quantum turnstile Update the plurality of individual qubits ; Among them, the individual qubits Expressed as:

[0051] and Respectively represent the probability weight values ​​of different gene features in the gene feature test set; Among them, the quantum revolving gate Expressed as:

[0052] is the rotation angle value of the quantum rotating gate, Used to update individual qubits The probability weight value of .

[0053] It should be noted that is the rotation angle value of the quantum rotating gate, Used to update individual qubits The probability weight value of , its amplitude is towards 0 or 1 state, in each iteration, The value of is determined by a predetermined lookup table (Kuk Hyun Han, Jong-Hwan Kim et al), and its rotation angle value is used to update the individual qubits The probability weight value of the selected gene feature is recalculated, and the probability of the selected gene feature is recalculated, and whether the feature can be retained or removed is determined, and whether there are other new features that can be selected into the latest feature subset. The larger the amplitude of the rotation adjustment, the smaller the quantum bit individual The larger the update, the larger the feature subset needs to be. This means that the feature subset needs to have a relatively large feature combination change and can point to a better solution. In the quantum-inspired evolutionary algorithm of this application, the next generation of quantum bit individuals is generated through a quantum rotating gate (Q-gate). These individuals point to better feature subset solutions, enhance search dynamics, improve the classification accuracy of the CNN model, and guide the search towards a gene expression spectrum space that can produce better results, thereby improving the convergence to a single state.

[0054] Compared to other genetic feature subsets, qubit individuals with higher fitness values ​​have a greater weighted value. Therefore, the optimal qubit individuals can be used as a high-weight constant to control the amplitude of the quantum rotary gate's rotation. Conversely, if the qubit individual weight decreases, the corresponding rotation adjustment update amplitude will also be reduced, avoiding the problem of excessive search and falling into local minimums. This allows the quantum rotary gate to be dynamically redefined at each evolutionary iteration, allowing the genetic feature subsets generated with each iteration to gradually point to a more optimal solution. As a result, qubit individuals with higher fitness values ​​provide more options and exploration capabilities for finding the optimal feature subset.

[0055] However, genetic computations require a large number of individuals to represent the feature space by setting up a predetermined random population and individuals each time. Furthermore, only the top features with a probability exceeding a user-defined probability are selected for the next evolutionary iteration. The remaining candidate features are further randomly selected when needed, but when the number of features is very large, this can easily lead to local search, failing to fully select strongly correlated gene features in the high-dimensional gene expression data space.

[0056] Reference Figure 2 In one embodiment of the present application, the following description can be combined to further illustrate "according to the plurality of qubit individuals Select multiple optimal quantum populations "The specific process.

[0057] As described in the following steps, based on the plurality of qubit individuals Select n individual qubits; Determine multiple optimal quantum populations based on the combination of n quantum bit individuals ; in,

[0058] n is the size of the quantum population.

[0059] It should be noted that multiple qubit individuals that meet the preset accuracy evaluation conditions will be selected and combined as a better quantum population. , thus multiple different better quantum populations can be combined , and then select the optimal quantum population , defined as the top layer of the quantum-inspired evolutionary algorithm, each generation will be stored, and each iteration can be evaluated using a fitness function to measure whether it is improved over the previous iteration.

[0060] Reference Figure 3In one embodiment of the present application, the method is used to determine the correspondence between gene features and cancer categories using a CNN model; the method further includes: Determine the classification accuracy of the CNN model based on the gene signature subtest set as described in the following steps; Determining a fitness function corresponding to the gene feature subtest set according to the classification accuracy, the number of the gene feature subtest sets, and the preset accuracy evaluation condition; A gene feature subtest set that satisfies an evaluation condition of the fitness function is determined based on the gene feature test set.

[0061] It should be noted that the quantum fusion model feature selection method is used for model training to optimize feature subset selection, and the cancer classification accuracy of the corresponding model is calculated based on different feature subsets. In the feature subset selection process, the fitness function of the quantum fusion model and the feature representation of the quantum bit vector group are used for evolutionary iteration to find a better feature combination until the optimal feature subset is selected.

[0062] As an example, the classification accuracy of the CNN model uses the True Positive (TP) and False Positive (FP) rates to evaluate the performance of this method. Accuracy, also known as precision, is used to measure the frequency of correctly predicted cancer classifications in the samples of this application. In the cancer classification scenario, TP represents the number of correctly predicted cancer categories, and FP represents the number of incorrectly predicted cancer categories. Therefore, accuracy can be expressed as TP / (TP + FP). The CNN model's classification accuracy and the number of selected gene features are used to construct a fitness function. Each feature subset and its corresponding classification result are evaluated using the fitness function equation. Based on the fitness function strategy of Huang Cheng-Lung et al., high fitness values ​​are determined by both high classification accuracy and the number of features. Furthermore, to avoid the denominator reaching zero, the fitness function can be expressed as:

[0063] in, is the classification accuracy; is the weight of classification accuracy; represents a genetic characteristic; P is a preset constant to prevent the denominator from reaching zero, usually set to 1 to 10; Can be pre-set to 75% to 100%; is the feature weight, which can be expressed as:

[0064] for The value of , “1” means that the gene feature is selected, and “0” means that the gene feature is not selected; The fitness function represents the representativeness of each feature subset for the sample in the high-dimensional gene expression data space. It is used to assess the quality of the results produced during the evolution process. After multiple iterations, when the termination criteria are met or the evolutionary process reaches 50 iterations, the standard average is used as the final result. The operation ends when the feature subset with the best fitness value is obtained; otherwise, the next round of evolutionary iterations of the gene feature subset continues.

[0065] In one embodiment of the present application, the specific process of "iteratively determining the target correspondence relationship based on the gene feature sub-test set and the cancer category test set for the initial correspondence relationship" in step S130 can be further explained in combination with the following description.

[0066] As described in the following steps, determining a cancer category subtest set corresponding to the gene feature subtest set in the cancer category test set based on the gene feature subtest set; The initial correspondence is iteratively determined based on the gene signature sub-test set and the cancer category sub-test set to determine a target correspondence.

[0067] It should be noted that the gene feature subtest set and the cancer category subtest set correspond to each other. The CNN model is trained using the gene feature subtest set and the cancer category subtest set, thereby iterating the correspondence between gene features and cancer categories to obtain the target correspondence.

[0068] Reference Figure 3 In a specific embodiment of the present application, in order to compare the effect difference between the feature selection method based on the quantum fusion model of the present application and the feature selection method of the traditional genetic algorithm, the present application uses a gene feature sample set obtained from 218 gene expression profiles to classify cancer. The sample set can include 16,063 gene features and corresponding 14 common human cancer categories. The sample set can be divided into a 60% training set and a 40% test set, and the feasibility of cancer classification is demonstrated by comparing the test set. The key parameter settings of the CNN model: the softmax layer is set to 0.01, and the learning rate is set to 0.01.

[0069] Reference Figure 4 , shows a comparison table of the accuracy and number of feature selections of this embodiment and the traditional genetic algorithm. It can be seen that the average accuracy of the feature selection method of the quantum fusion model is better than that of the traditional genetic algorithm, and the number of selected features is also greater, which helps to improve the accuracy of model classification. Specifically, Figure 4As shown in the figure, the average accuracy of the quantum fusion model feature selection method for classifying 14 cancer types is 89.61%, while the average accuracy of the genetic algorithm feature selection method is lower, at 84.89%. This also indicates that the prediction classification is driven by the intrinsic gene expression pattern of cancer, and many cancers retain the identity of their tissue of origin during metastasis evolution. In addition, as Figure 4 As shown, based on the same 16,063 input gene information, the quantum fusion model feature selection method of the present application selected 118 gene features, while the genetic algorithm feature selection method selected fewer gene features, namely 95.

[0070] Reference Figure 5 , shows the average selection percentage of each gene feature in this embodiment and the traditional genetic algorithm. The solid line represents the feature selection percentage using the quantum fusion model method, and the dotted line represents the feature selection percentage used by the genetic algorithm. The higher the selection percentage, the more frequently the gene feature is selected, that is, it has a higher correlation with the cancer category. Figure 5 The largest differences in feature selection percentages between the two methods can also be seen. For example, while both selection methods can select gene microarray features 25-28, including prostate-specific antigen (prostate cancer), carcinoembryonic antigen (colon cancer), CD20 (lymphoma), and S100 (melanoma), the present application also discovered some gene microarray features not identified by the genetic algorithm feature selection method, including tissue-specific genes. For example, colorectal adenocarcinoma-specific marker features 32-34, including intestinal-specific transcription factors, cytoskeleton and adhesion molecules, signaling molecules, and membrane-bound tumor markers, as well as features 36-38, including prostate-specific antigen, carcinoembryonic antigen, and estrogen receptor, were all selected at a higher rate using the present application. This demonstrates that the present application is able to mine and select a greater number of gene microarray features. As the number of selected features increases, the CNN model used for classification yields more accurate cancer classification predictions. Overall, the present application is able to select more effective gene features for training the CNN model in complex, high-dimensional encodings of gene expression patterns, leading to more accurate cancer classification.

[0071] The above is a description of the method embodiment of the present application. As for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple. For relevant details, please refer to the partial description of the method embodiment.

[0072] Reference Figure 6 , shows a cancer classification device based on a quantum-inspired evolutionary algorithm provided by an embodiment of the present application, the device comprising: An initial correspondence determination module 610 is configured to obtain a gene signature training set, a cancer category training set, a gene signature test set, and a cancer category test set, and determine an initial correspondence between gene signatures and cancer categories based on the gene signature training set and the cancer category training set; A gene feature subtest set determination module 620 is configured to determine a gene feature subtest set that satisfies a preset accuracy evaluation condition based on the gene feature test set by using a quantum-inspired evolutionary algorithm; a target correspondence determination module 630, configured to iteratively determine a target correspondence for the initial correspondence based on the gene signature subtest set and the cancer category test set; The current cancer category determination module 640 is configured to obtain a current gene feature sample and determine a current cancer category based on the corresponding relationship between the current gene feature sample and the target.

[0073] In one embodiment of the present application, the gene signature subtest set determination module 620 includes: A qubit individual determination submodule, configured to determine a plurality of corresponding qubit individuals based on the genetic feature test set; An optimal quantum population selection submodule, configured to select an optimal quantum population based on the plurality of individual qubits; The gene feature sub-test set determination submodule is used to determine the gene feature sub-test set that meets the preset accuracy evaluation conditions based on the optimal quantum population.

[0074] In one embodiment of the present application, the optimal quantum population selection submodule includes: A qubit individual updating unit is configured to update the qubit individual Update the plurality of individual qubits ; Where t represents the number of iterations; A better quantum population selection unit is used to select the optimal quantum population based on the multiple quantum bit individuals Select multiple optimal quantum populations ; The optimal quantum population selection unit is used to select the optimal quantum population according to the multiple optimal quantum populations. Selecting the optimal quantum population .

[0075] In one embodiment of the present application, the individual qubit update unit includes: A quantum rotation gate subunit is used to determine the quantum bit Through the quantum turnstile Update the plurality of individual qubits ; Among them, the individual qubits Expressed as:

[0076] and Respectively represent the probability weight values ​​of different gene features in the gene feature test set; Among them, the quantum revolving gate Expressed as:

[0077] is the rotation angle value of the quantum rotating gate, Used to update individual qubits The probability weight value of .

[0078] In one embodiment of the present application, the preferred quantum population selection unit includes: The qubit individual selection subunit is used to select the qubit individual according to the plurality of qubit individual Select n individual qubits; The qubit individual combination subunit is used to determine multiple optimal quantum populations based on the combination of n qubit individuals ; in,

[0079] n is the size of the quantum population.

[0080] In one embodiment of the present application, the apparatus is used to determine the correspondence between gene features and cancer categories using a CNN model; the apparatus further comprises: A classification accuracy determination module, configured to determine the classification accuracy of the CNN model based on the gene signature subtest set; a fitness function determination module, configured to determine a fitness function corresponding to the gene feature subtest set based on the classification accuracy, the number of the gene feature subtest sets, and the preset accuracy evaluation condition; The fitness function evaluation module is used to determine a gene feature sub-test set that meets the evaluation conditions of the fitness function based on the gene feature test set.

[0081] In one embodiment of the present application, the target correspondence determination module 630 includes: a cancer category subtest set determination submodule, configured to determine, in the cancer category test set based on the gene signature subtest set, a cancer category subtest set corresponding to the gene signature subtest set; The target correspondence determination submodule is configured to iteratively determine a target correspondence for the initial correspondence based on the gene feature subtest set and the cancer category subtest set.

[0082] Reference Figure 7 , shows a block diagram of a computer device provided in an embodiment of the present application. The computer device 12 is suitable for implementing the embodiments of the present invention and may specifically include the following: Computer device 12 is implemented as a general-purpose computing device. Components of computer device 12 may include, but are not limited to, one or more processing units 16, system memory 28, and a bus 18 connecting various system components (including system memory 28 and processing units 16). Computer device 12 may be a device connected to the bus.

[0083] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures. Examples of these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MCA) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0084] The computer device 12 typically includes a variety of computer system readable media. These media can be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0085] System memory 28 may include computer system readable media in the form of volatile memory, such as RAM 30 (random access memory) and / or cache 32. Computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 may be configured to read and write to non-removable, non-volatile magnetic media (commonly referred to as a "hard drive"). Although Figure 7 Although not shown, a magnetic disk drive for reading and writing to a removable non-volatile magnetic disk (e.g., a "floppy disk"), as well as an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 via one or more data media interfaces. System memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of various embodiments of the present invention.

[0086] A program / utility 40 having a set (at least one) of program modules 42 may be stored, for example, in system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data, each of which, or some combination thereof, may include an implementation of a network environment. Program modules 42 generally perform the functions and / or methodologies of the embodiments described herein.

[0087] The computer device 12 may also communicate with one or more external devices 14 (e.g., a keyboard, a pointing device, a display 24, etc.), one or more devices that enable a user to interact with the computer device 12, and / or any device that enables the computer device 12 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.). Such communication may be performed through an I / O interface 22 (input / output interface). Furthermore, the computer device 12 may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network (e.g., the Internet)) through a network adapter 20. Figure 7 As shown, the network adapter 20 communicates with the other modules of the computer device 12 via the bus 18. Figure 7 Not shown, other hardware and / or software modules may be used in conjunction with computer device 12, including but not limited to microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0088] The processing unit 16 executes programs stored in the system memory 28 to perform various functional applications and data processing, such as implementing the method provided by any embodiment of the present invention.

[0089] That is, when the program is executed by the processor, it implements: obtaining a gene feature training set, a cancer category training set, a gene feature test set, and a cancer category test set, and determining an initial correspondence between gene features and cancer categories based on the gene feature training set and the cancer category training set; Determining a gene feature sub-test set that meets a preset accuracy evaluation condition based on the gene feature test set by using a quantum-inspired evolutionary algorithm; Iteratively determining a target correspondence relationship based on the gene signature subtest set and the cancer category test set on the initial correspondence relationship; A current gene signature sample is obtained, and a current cancer category is determined based on a correspondence between the current gene signature sample and the target.

[0090] The computer device 12 is merely an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0091] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method provided in any embodiment of the present application.

[0092] That is, when the program is executed by the processor, it implements: obtaining a gene feature training set, a cancer category training set, a gene feature test set, and a cancer category test set, and determining an initial correspondence between gene features and cancer categories based on the gene feature training set and the cancer category training set; Determining a gene feature sub-test set that meets a preset accuracy evaluation condition based on the gene feature test set by using a quantum-inspired evolutionary algorithm; Iteratively determining a target correspondence relationship based on the gene signature subtest set and the cancer category test set on the initial correspondence relationship; A current gene signature sample is obtained, and a current cancer category is determined based on a correspondence between the current gene signature sample and the target.

[0093] Computer storage media can take the form of any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, RAM, read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, a portable CD-ROM, an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0094] A computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take a variety of forms, including, but not limited to, electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device.

[0095] Program code embodied on a computer-readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, radio frequency (RF), etc., or any suitable combination of the foregoing.

[0096] The computer program code for performing the operations of the present invention may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer via any type of network, including a LAN or WAN, or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0097] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they become aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the embodiments of the present invention.

[0098] Finally, it should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or elements inherent to such process, method, article, or terminal device. In the absence of further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article, or terminal device that includes the element.

[0099] The above is a detailed introduction to the cancer classification method and device based on the quantum-inspired evolutionary algorithm provided by the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method and core ideas of the present application. At the same time, for those skilled in the art, based on the ideas of the present application, there may be changes in the specific implementation methods and application scopes. In summary, the content of this specification should not be understood as limiting the present application.

Claims

1. A cancer classification method based on quantum-inspired evolutionary algorithm, characterized in that: The method comprises: Obtaining a gene signature training set, a cancer category training set, a gene signature test set, and a cancer category test set, and determining an initial correspondence between gene signatures and cancer categories based on the gene signature training set and the cancer category training set; Determining a gene feature sub-test set that meets a preset accuracy evaluation condition based on the gene feature test set by using a quantum-inspired evolutionary algorithm; Iteratively determining a target correspondence relationship based on the gene signature subtest set and the cancer category test set on the initial correspondence relationship; A current gene signature sample is obtained, and a current cancer category is determined based on a correspondence between the current gene signature sample and the target.

2. The method according to claim 1, characterized in that The step of determining a gene feature sub-test set that meets a preset accuracy evaluation condition based on the gene feature test set by using a quantum-inspired evolutionary algorithm comprises: Determining a plurality of corresponding qubit individuals based on the genetic feature test set; Selecting an optimal quantum population based on the plurality of individual qubits; A gene feature subtest set that meets a preset accuracy evaluation condition is determined based on the optimal quantum population.

3. The method according to claim 2, characterized in that The step of selecting the optimal quantum population based on the plurality of individual qubits comprises: According to the plurality of individual qubits Update the plurality of individual qubits ; Where t represents the number of iterations; According to the plurality of individual qubits Select multiple optimal quantum populations ; According to the multiple optimal quantum populations Selecting the optimal quantum population .

4. The method according to claim 3, characterized in that The plurality of qubit individuals Update the plurality of individual qubits The steps include: According to the plurality of individual qubits Through the quantum turnstile Update the plurality of individual qubits ; Among them, the individual qubits Expressed as: and Respectively represent the probability weight values ​​of different gene features in the gene feature test set; Among them, the quantum revolving gate Expressed as: is the rotation angle value of the quantum rotating gate, Used to update individual qubits The probability weight value of .

5. The method according to claim 3, characterized in that The plurality of qubit individuals Select multiple optimal quantum populations The steps include: According to the plurality of individual qubits Select n individual qubits; Determine multiple optimal quantum populations based on the combination of n quantum bit individuals ; in, n is the size of the quantum population.

6. The method according to claim 1, characterized in that The method is used to determine the correspondence between gene features and cancer categories through a CNN model; the method also includes: Determining the classification accuracy of the CNN model based on the gene signature subtest set; Determining a fitness function corresponding to the gene feature subtest set according to the classification accuracy, the number of the gene feature subtest sets, and the preset accuracy evaluation condition; A gene feature subtest set that satisfies an evaluation condition of the fitness function is determined based on the gene feature test set.

7. The method according to claim 1, characterized in that The step of iteratively determining the target correspondence relationship based on the gene feature subtest set and the cancer category test set comprises: Determining, in the cancer category test set according to the gene signature subtest set, a cancer category subtest set corresponding to the gene signature subtest set; The initial correspondence is iteratively determined based on the gene signature sub-test set and the cancer category sub-test set to determine a target correspondence.

8. A cancer classification device based on a quantum-inspired evolutionary algorithm, characterized in that: The device comprises: an initial correspondence determination module, configured to obtain a gene feature training set, a cancer category training set, a gene feature test set, and a cancer category test set, and determine an initial correspondence between gene features and cancer categories based on the gene feature training set and the cancer category training set; A gene feature subtest set determination module is used to determine a gene feature subtest set that meets a preset accuracy evaluation condition based on the gene feature test set through a quantum-inspired evolutionary algorithm; a target correspondence determination module, configured to iteratively determine a target correspondence for the initial correspondence based on the gene feature subtest set and the cancer category test set; The current cancer category determination module is used to obtain a current gene feature sample and determine the current cancer category based on the corresponding relationship between the current gene feature sample and the target.

9. A computer device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program implements the method according to any one of claims 1 to 7 when executed by the processor.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images