Group intelligence-based ploidy analysis quality control method and system

Abstract

The invention relates to the technical field of bioinformatics, in particular to a ploidy analysis quality control method and system based on swarm intelligence, and the method comprises the steps: obtaining ploidy image data of a sample group and a corresponding reference template; executing group feature topological mapping based on the ploidy image data and the reference template; performing group theory based feature enhancement according to the group feature topological mapping result; on the basis of the feature enhancement result, constructing a dynamic weighting Laplacian matrix; according to the dynamic weighting Laplacian matrix, spectrum embedding and group consistency optimization are executed; based on the spectrum embedding and group consistency optimization result, performing ploidy quality evaluation; and the ploidy quality evaluation result and the optimized group characteristics are output, so that the overall information of the sample group can be fully utilized, the analysis process can be adaptively optimized, and the analysis accuracy, reliability and efficiency are effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of bioinformatics, and more specifically, to a ploidy analysis quality control method and system based on swarm intelligence. Background Art

[0002] Ploidy analysis plays a vital role in modern biological research and applications, especially in crop breeding, genetic disease research and other fields. With the rapid development of high-throughput sequencing technology and image analysis technology, the amount and complexity of ploidy analysis data are increasing. However, existing ploidy analysis methods often have problems such as insufficient accuracy and poor reliability when facing large-scale, high-dimensional data.

[0003] Ploidy analysis is one of the most important technologies in cytogenetic research and clinical diagnosis, and is widely used in tumor detection, genetic disease analysis, plant breeding and other fields. However, existing ploidy analysis methods still have many limitations in terms of accuracy and quality control, especially in large-scale sample analysis, and are weak in processing data noise, outliers, and overall optimization. In response to these problems, research in related fields has gradually emerged, but a unified and efficient technical solution has not yet been formed.

[0004] For example, Shao Shaowei and Luo Qiaoming published an article in the Journal of Diagnostic Pathology titled “Technology and Quality Control of DNA Ploidy Analysis of Cervical Exfoliated Cells” (2014, Vol. 21, No. 3, pp. 184-185), which mainly discussed the technology and quality control measures in DNA ploidy analysis of cervical exfoliated cells. However, the focus of this study was mainly on the preparation and staining process, and the quality control methods in data analysis were not explored in depth. In addition, this study lacked consideration of the synergy between different samples, making it difficult to generalize and apply in a complex data context.

[0005] In addition, Guo Jianwen et al. proposed a standardized strategy for quantitative analysis of tumor cell DNA ploidy using flow cytometry in the paper "Standardization and Control of Quantitative Analysis of Tumor Cell DNA Ploidy by Flow Cytometry" (2011, page 225) published in the "Collection of Papers from the Second China Modern Medical Research Methods and Interdisciplinary Innovation Seminar". Although a relatively complete solution is provided in terms of standardization of sample preparation and staining, there is less discussion on the quality control of the data processing process, especially when dealing with large sample sizes or high-noise data, there is a lack of effective algorithm support, which cannot fully meet the needs of practical applications.

[0006] In addition, Wang Jinlong et al. discussed the application of artificial intelligence in ploidy analysis in the article "Application of Artificial Intelligence Cell DNA Ploid Quantification Analysis in Cervical Cancer Screening" published in the Electronic Journal of Modern Medicine and Health Research (2021, Vol. 5, No. 3, p. 4042). Although this study demonstrated the advantages of artificial intelligence technology in specific diagnostic scenarios, it mainly focused on the diagnostic performance of the model without conducting an in-depth discussion on the systematic nature of quality control, especially in terms of multi-sample collaborative analysis.

[0007] In summary, the limitations of existing technologies are mainly reflected in the following aspects: First, quality control measures are mostly concentrated in the sample preparation stage, while there is little discussion on noise elimination, anomaly detection and feature optimization in the data analysis and processing process; second, most existing methods are oriented to specific application scenarios and lack generalization and adaptability; third, the collaborative information of the sample group is not fully utilized, and it is difficult to improve the robustness and accuracy of the analysis results through group intelligence methods. These problems limit the application of ploidy analysis technology in a wider range of fields and also affect the stability and reliability of the analysis results.

[0008] In addition, traditional ploidy analysis methods mainly rely on single sample analysis, which cannot fully utilize the overall information of the sample population, resulting in analysis results that are easily affected by individual differences and measurement errors. Although some researchers have tried to introduce statistical methods to improve the quality of analysis, these methods can usually only handle linear relationships and are difficult to capture the complex nonlinear characteristics of ploidy data.

[0009] In recent years, machine learning methods, especially deep learning technology, have been introduced into the field of ploidy analysis, which has improved the accuracy of analysis to a certain extent. However, these methods still have some significant problems. First, they often require a large amount of labeled data for training, while in practical applications, high-quality labeled data is often difficult to obtain. Second, these methods are usually black box models, lack interpretability, and are difficult to effectively combine with biological knowledge. In addition, they perform poorly when dealing with abnormal samples and newly emerging ploidy types, and lack the necessary robustness and adaptability.

[0010] Another issue worth noting is that existing methods have obvious deficiencies in quality control of ploidy analysis. Most methods lack effective mechanisms to evaluate the reliability of analysis results, and are unable to automatically identify and process abnormal samples. As a result, in practical applications, researchers often need to spend a lot of time on manual verification, which greatly reduces analysis efficiency. Summary of the invention

[0011] In view of the above problems, new ploidy analysis methods are urgently needed, which can make full use of sample population information, improve the accuracy and reliability of analysis, and have good interpretability and adaptability. In particular, such methods should be able to effectively process large-scale, high-dimensional ploidy data, automatically perform quality control, and be able to flexibly cope with various complex analysis scenarios.

[0012] The present invention provides a ploidy analysis quality control method based on swarm intelligence, comprising:

[0013] The acquisition steps include:

[0014] Obtaining ploidy image data of a sample population and a corresponding reference template;

[0015] Processing steps include:

[0016] Based on the ploidy image data and the reference template, performing population feature topology mapping;

[0017] According to the group feature topology mapping result, group theory based feature enhancement is performed;

[0018] Based on the feature enhancement result, constructing a dynamic weighted Laplacian matrix;

[0019] performing spectral embedding and group consistency optimization according to the dynamic weighted Laplacian matrix;

[0020] Based on the spectrum embedding and population consistency optimization results, ploidy quality assessment is performed;

[0021] Output steps include:

[0022] Output ploidy quality assessment results and optimized population characteristics.

[0023] Preferably, the group feature topology mapping specifically includes:

[0024] Construct the characteristic topological space T = {X i |i∈{1, 2, ..., n}}, where X i is the feature vector of the i-th sample, and n is the total number of samples;

[0025] Define topology mapping function in Y i is the feature vector after mapping, W is the weight matrix, b is the bias vector, and m is the feature dimension after mapping;

[0026] Calculate the group consistency matrix C, where C ij =exp(-||Y i -Y j ||2 / (2σ 2 )),σ is the Gaussian kernel parameter, which represents the parameter used to control the influence of distance in similarity calculation.

[0027] Preferably, the group theory based feature enhancement specifically includes:

[0028] Define the characteristic transformation group G = {g k |k∈{1,2,…,K}},g k is the kth transformation operation, K is the total number of transformation operations;

[0029] Apply the group operation to each sample and get Z i is the enhanced feature vector.

[0030] Preferably, the dynamic weighted Laplacian matrix construction specifically includes:

[0031] Calculate the adjacency matrix A, where A ij =exp(-||Z i -Z j || 2 / (2τ 2 ))·C ij ,τ is the temperature parameter, which represents the parameter used to adjust the size of elements in the adjacency matrix;

[0032] Calculate the degree matrix D, where

[0033] Construct the Laplacian matrix L=DA.

[0034] Preferably, the spectrum embedding and group consistency optimization specifically include:

[0035] Solve the optimization problem min U Tr(U T LU) T U = I, where U is the embedding matrix, I is the identity matrix, and Tr represents the trace of the matrix;

[0036] Define group consistency optimization goals Where α is the equilibrium parameter, ∥·∥ F represents the Frobenius norm, and Z is the output matrix of feature enhancement.

[0037] Preferably, the ploidy quality assessment specifically includes:

[0038] Define ploidy quality assessment function in U i is the i-th row of U, Uavg is the average value of U, σ q is a quality assessment parameter, which indicates a parameter used to adjust the degree of influence of distance in the quality assessment function;

[0039] Calculate the feedback function to update the weight matrix W, Where η is the learning rate, is the gradient of Q(U) with respect to W.

[0040] Preferably, it also includes an iterative optimization step:

[0041] Based on the ploidy quality assessment result, adjusting the parameters of the population characteristic topological mapping;

[0042] The processing steps are repeatedly performed until a preset convergence condition is met or a maximum number of iterations is reached.

[0043] Preferably, the group feature topology mapping further includes:

[0044] Use deep learning models to preprocess ploidy images, including image denoising, enhancement, and standardization;

[0045] Based on the preprocessed image, extract the initial feature vector X i .

[0046] Preferably, the method further includes an anomaly detection step:

[0047] Based on the ploidy quality assessment result, identifying abnormal samples;

[0048] Identified abnormal samples are analyzed and processed individually;

[0049] Update the population feature topological space to eliminate the influence of abnormal samples.

[0050] The ploidy analysis quality control system based on swarm intelligence includes:

[0051] A data acquisition module, used to acquire ploidy image data of a sample population and a corresponding reference template;

[0052] A population feature topology mapping module, used to perform population feature topology mapping based on the ploidy image data and the reference template;

[0053] A feature enhancement module, used for performing group theory based feature enhancement according to the group feature topology mapping result;

[0054] A Laplacian matrix construction module, used for constructing a dynamic weighted Laplacian matrix based on the feature enhancement result;

[0055] A spectrum embedding optimization module, for performing spectrum embedding and group consistency optimization according to the dynamic weighted Laplacian matrix;

[0056] A quality assessment module, used for performing ploidy quality assessment based on the spectrum embedding and population consistency optimization results;

[0057] Output module, used to output ploidy quality assessment results and optimized population characteristics;

[0058] An iterative optimization module, used to adjust the parameters of the population feature topology mapping based on the ploidy quality assessment result, and control the system to repeatedly execute the processing steps;

[0059] The anomaly detection module is used to identify and process abnormal samples based on the ploidy quality assessment results.

[0060] The method of the present invention has the following significant beneficial effects:

[0061] The ploidy analysis quality control method based on swarm intelligence proposed in this paper constructs a new ploidy analysis framework by innovatively combining advanced concepts such as swarm intelligence, topology and group theory. This framework can not only make full use of the overall information of the sample population, but also adaptively optimize the analysis process, effectively improving the accuracy, reliability and efficiency of the analysis.

[0062] Specifically, the present invention effectively solves the problem that traditional methods are difficult to capture complex nonlinear features through population feature topological mapping and group theory based feature enhancement. The construction of the dynamic weighted Laplacian matrix and spectral embedding optimization cleverly integrates population structure information into the analysis process, significantly improving the robustness and adaptability of the method. It is particularly worth mentioning that the ploidy quality assessment mechanism introduced in the present invention can not only automatically evaluate the reliability of the analysis results, but also continuously optimize the analysis process through a feedback mechanism, which largely solves the shortcomings of existing methods in quality control.

[0063] The method of the present invention exhibits significant advantages in many aspects. First, it greatly improves the accuracy of ploidy analysis, especially when dealing with complex and highly heterogeneous sample groups. Secondly, the method has good interpretability, and each step of the analysis process has clear mathematical and biological meanings, which helps researchers to have a deeper understanding of ploidy characteristics. Furthermore, the method of the present invention exhibits excellent robustness and adaptability, and can effectively handle abnormal samples and newly emerging ploidy types, which is of great significance for complex situations often encountered in practical applications.

[0064] In addition, the method of the present invention also has remarkable performance in terms of computational efficiency. Although a complex mathematical model is introduced, the method can efficiently process large-scale data through ingenious algorithm design. In particular, there is a synergistic effect between the various steps of the method, such as group feature topological mapping provides a good basis for subsequent feature enhancement, and the dynamic weighted Laplacian matrix creates favorable conditions for spectrum embedding optimization. This complementary and superposition effect between steps not only improves the overall analysis quality, but also offsets the time overhead caused by complex calculations to a certain extent.

[0065] By introducing the concept of swarm intelligence, the present invention utilizes the synergy between sample groups to construct a feature topological space to achieve feature enhancement and dynamic optimization of ploidy data. At the same time, through the construction of a dynamic weighted Laplacian matrix and the introduction of spectral embedding technology, the present invention can improve the ability to distinguish data while reducing the impact of noise. In addition, the present invention designs a comprehensive set of quality assessment functions, combines feedback mechanisms to dynamically adjust analysis parameters, and thus achieves closed-loop optimization in the context of multi-sample big data.

[0066] The present invention not only solves the problems of high noise and low accuracy in ploidy analysis through the complementarity, superposition and synergy of feature mapping, feature enhancement and spectrum embedding optimization, but also significantly improves the consistency and stability of the analysis results. Finally, the present invention realizes the whole process quality control in the ploidy analysis process, which is efficient, robust and widely applicable, and provides a new technical path and implementation scheme for this field.

[0067] In general, the ploidy analysis quality control method and system based on swarm intelligence provided by the present invention has brought a qualitative leap in the field of ploidy analysis by innovatively solving multiple key problems in the prior art. It not only improves the accuracy and reliability of the analysis, but also greatly improves the automation level and efficiency of the analysis process. The application of this method is expected to have a far-reaching impact in many fields such as crop breeding and genetic disease research, and promote the rapid development of related scientific research and technological applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0069] Please refer to Figure 1 The present invention provides a ploidy analysis quality control method and system based on swarm intelligence. The method improves the accuracy and reliability of ploidy analysis through the idea of ​​swarm intelligence. The specific implementation methods of the present invention will be described in detail below.

[0070] The method of the present invention comprises an acquisition step, a processing step and an output step. In the acquisition step, the method acquires ploidy image data of a sample population and a corresponding reference template. These data can come from a plurality of different laboratories or research institutions, thereby forming a representative sample population.

[0071] In the processing step, the method first performs population feature topological mapping based on the acquired ploidy image data and the reference template. This step aims to map the original ploidy image data into a high-dimensional feature space to better capture the relationship between samples. Preferably, the present invention uses a deep learning model to pre-process the ploidy image, including image denoising, enhancement and standardization, to improve the accuracy of subsequent analysis.

[0072] Next, the method performs group theory based feature enhancement based on the result of the group feature topology mapping. The purpose of this step is to further enhance the expressiveness of the feature through the concept of group theory. In one embodiment of the present invention, feature enhancement can be achieved by applying a series of transformation operations, which form a transformation group, thereby enriching the expression of the feature.

[0073] Subsequently, based on the results of feature enhancement, the method constructs a dynamic weighted Laplacian matrix. This step takes into account the similarity and population structure between samples, providing important information for the subsequent optimization process. Preferably, the present invention adopts an adaptive weight calculation method so that the Laplacian matrix can better reflect the dynamic characteristics of the sample population.

[0074] After constructing the Laplacian matrix, the method performs spectral embedding and group consistency optimization. This step obtains a low-dimensional embedding representation by solving the optimization problem while maintaining group consistency between samples. In one embodiment of the present invention, the optimization objective function includes a spectral embedding term and a group consistency term, and by adjusting the balance parameter between these two terms, a better optimization result can be obtained.

[0075] Finally, the method performs ploidy quality assessment based on the results of spectrum embedding and population consistency optimization. This step quantitatively evaluates the quality of each sample by designing a suitable evaluation function. Preferably, the present invention adopts an evaluation method based on population statistical characteristics, so as to better reflect the relative quality of the sample in the population.

[0076] In the output step, the method outputs the ploidy quality assessment results and optimized population characteristics. These output results can provide important reference information for subsequent ploidy analysis.

[0077] Assume that ploidy image data from different sources are obtained from multiple laboratories. These images may come from different experimental conditions, equipment, or operators. There may be noise, background interference, and other problems in the original image data, and direct analysis may lead to inaccurate results.

[0078] The group feature topological mapping of the present invention specifically includes constructing a feature topological space, defining a topological mapping function, and calculating a group consistency matrix. When constructing the feature topological space, the method defines a set T:

[0079] T={X i |i∈{1,2,…,n}},

[0080] Among them, X i is the feature vector of the i-th sample, and n is the total number of samples. This feature topological space provides the basis for subsequent analysis.

[0081] In an actual ploidy analysis project, researchers collected 1,000 ploidy images from five different laboratories. Due to the different equipment and operating procedures in each laboratory, the quality of the images varied. By constructing a feature topological space, these images can be uniformly mapped to a high-dimensional feature space to eliminate the impact of equipment differences.

[0082] By mapping image data to a high-dimensional feature space, the system can capture more details and patterns, thereby improving the accuracy of subsequent analysis. For example, when processing chromosome images, the feature topology space can help the system more accurately identify the morphology and structure of chromosomes and reduce misjudgments. The high-dimensional feature space can filter out noise in the image, making subsequent analysis more stable and reliable. For example, when processing low-quality ploidy images, the feature topology space can effectively remove background interference and retain key chromosome information. The construction of the feature topology space provides a basis for processing large-scale ploidy image data and can efficiently process data from multiple laboratories. For example, in a global genome research project, the system can quickly process tens of thousands of ploidy images through the feature topology space to ensure the consistency and reliability of the analysis results.

[0083] In ploidy analysis, raw image data usually contains a lot of redundant information and noise. Directly using these data for analysis may lead to overfitting or insufficient generalization of the model. Through the topological mapping function, the high-dimensional feature vector can be reduced to a feature space that is more suitable for subsequent analysis.

[0084] Next, this method defines a topology mapping function

[0085]

[0086] Among them, Yi is the mapped feature vector, W is the weight matrix, b is the bias vector, and m is the mapped feature dimension. This mapping function allows the original features to be transformed into a new feature space, which may result in better expression.

[0087] In a chromosome classification task, researchers used a deep learning model to extract a 2048-dimensional feature vector for each ploidy image. However, these high-dimensional feature vectors are not only computationally complex, but also easily affected by noise. Through the topological mapping function, the system maps these feature vectors to a 128-dimensional feature space, which not only reduces the dimension but also enhances the distinguishability of the features.

[0088] Through linear mapping, the system can reduce the dimensionality of high-dimensional feature vectors to a lower-dimensional space while retaining key information and reducing computational complexity. For example, when processing large-scale ploid image data, the reduced-dimensional feature vector can significantly improve computational efficiency and shorten analysis time. The mapped feature vector has better distinguishability in the new space and can more accurately identify different types of ploid samples. For example, in chromosome classification tasks, the reduced-dimensional feature vector can help the system more accurately distinguish normal chromosomes from abnormal chromosomes and improve classification accuracy.

[0089] The weight matrix W and the bias vector b can be optimized through a deep learning model, so that the mapping function can be adaptively adjusted according to the specific needs of ploidy analysis. For example, when processing different types of ploidy images, the system can automatically adjust the mapping parameters according to the characteristics of the image to further improve the accuracy of the analysis.

[0090] In ploidy analysis, the similarity between samples is crucial for subsequent clustering, classification and other tasks. The population consistency matrix can effectively capture the local similarities between samples and help the system better understand the overall structure of the sample population.

[0091] Finally, this method calculates the population consistency matrix C:

[0092]

[0093] Where σ is the Gaussian kernel parameter. This matrix captures the similarity between samples and provides important information for subsequent analysis. In practical applications, the choice of σ has an important impact on the results. Generally speaking, σ can be set to the median of the distance between samples, which ensures that the values ​​in the matrix are distributed in a suitable range.

[0094] In a chromosome clustering task, researchers need to classify different types of chromosomes into several groups. By constructing a group consistency matrix, the system can calculate the similarities between each pair of samples and perform clustering based on these similarities. For example, for a set of multiple chromosome images from the same patient, the group consistency matrix can help the system identify which chromosomes are of the same type and which chromosomes are abnormal.

[0095] The group consistency matrix can effectively capture the similarities between samples, especially the local similarities, which is very useful for subsequent clustering and classification tasks. For example, when processing chromosome images, the group consistency matrix can help the system more accurately identify which chromosomes are of the same type and which chromosomes are abnormal. Through the group consistency matrix, the system can better reflect the overall structure of the sample population, helping the subsequent spectrum embedding and optimization process to better maintain group consistency. For example, when processing large-scale ploid image data, the group consistency matrix can help the system identify the global and local relationships between samples and ensure the accuracy of the analysis results. The group consistency matrix can also be used for anomaly detection to identify abnormal samples that have low similarity with other samples, thereby improving the robustness of the analysis. For example, when processing a batch of chromosome images, if the similarity of a sample is significantly lower than that of other samples, the system can mark it as an abnormal sample for further analysis.

[0096] In ploidy analysis, the original feature vector may not be able to fully capture the complex relationship between samples, especially when there are slight differences between samples. By introducing the feature transformation group G, the system can transform the features from multiple different perspectives, thereby enhancing the diversity and expression ability of the features.

[0097] The group-based feature enhancement of the present invention specifically includes defining a feature transformation group and applying a group operation to each sample. First, the method defines a feature transformation group G:

[0098] G = {g k |k∈{1,2,…,K}},

[0099] Among them, g k is the kth transformation operation, and K is the total number of transformation operations. These transformation operations may include rotation, scaling, translation, etc. The specific selection depends on the specific requirements of ploidy analysis.

[0100] The method then applies the group operation to each sample:

[0101]

[0102] Among them, Z iis the enhanced feature vector. This step enriches the feature expression by applying a series of transformations, which helps to improve the accuracy of subsequent analysis. In practical applications, the selection of K needs to balance the computational complexity and feature enhancement effect. Generally speaking, K can be set between 10 and 20, which can achieve good enhancement effect at a reasonable computational cost.

[0103] The feature transformation group G can transform features from multiple angles, allowing the system to capture more detailed information, especially some minor structural differences. This helps to improve the expressiveness and distinguishability of features, thereby improving classification accuracy. By introducing multiple transformation operations, the system can better cope with different types of samples, especially when the sample distribution is more complex. This helps to improve the robustness of the system, making it more stable and reliable when processing new data. The feature transformation group G provides a variety of transformation operations. The system can dynamically select the most appropriate transformation operation based on the specific sample type and application scenario, and further optimize the feature extraction process. This helps to improve the flexibility and adaptability of the system. The feature transformation group G can transform features from multiple angles, so that the similarity between samples is enhanced, especially when dealing with samples with complex structures. This helps to improve the accuracy of clustering and classification tasks.

[0104] The synergy of these technologies makes the method of the present invention have significant advantages in ploidy analysis, and can effectively improve the accuracy and reliability of the analysis. By introducing the feature transformation group G, the system can not only better capture the complex relationship between samples, but also cope with changes in various experimental environments, ensuring the stability and reliability of the analysis results.

[0105] Through the above steps, the method of the present invention can effectively use population information to improve the quality of ploidy analysis. In particular, the combination of population feature topological mapping and group-based feature enhancement enables the method to better capture the relationship between samples and the essential attributes of ploidy characteristics. This method based on swarm intelligence has higher robustness and accuracy than traditional single sample analysis methods.

[0106] In practical applications, the method of the present invention can adjust parameters according to specific ploidy analysis requirements. For example, in the topological mapping function, the weight matrix W and the bias vector b can be optimized by deep learning methods. In group theory based feature enhancement, the choice of transformation operation can be determined according to the physical meaning of the ploidy feature. This flexibility enables the method to adapt to different ploidy analysis scenarios.

[0107] In a chromosome classification task, researchers need to classify different types of chromosomes. By constructing a dynamic weighted Laplacian matrix, the system can dynamically adjust the weights according to the distance between samples and group consistency to ensure the accuracy of the classification results. For example, for a set of multiple chromosome images from the same patient, the dynamic weighted Laplacian matrix can help the system identify which chromosomes are of the same type and which chromosomes are abnormal.

[0108] The dynamic weighted Laplacian matrix construction of the present invention specifically includes calculating the adjacency matrix, calculating the degree matrix and constructing the Laplacian matrix. First, the method calculates the adjacency matrix A:

[0109]

[0110] Among them, τ is the temperature parameter, C ij is the group consistency matrix element calculated above. This adjacency matrix not only considers the distance between samples, but also incorporates the group consistency information, thereby more comprehensively describing the relationship between samples. In practical applications, the choice of τ has an important impact on the results. Preferably, τ can be set to the standard deviation of the distance between samples, so that the values ​​of the adjacency matrix can be distributed in a suitable range.

[0111] Next, this method calculates the degree matrix D:

[0112]

[0113] The degree matrix D is a diagonal matrix whose diagonal elements represent the sum of the connection strengths of each sample with all other samples. This matrix provides information about the importance of the sample in the population.

[0114] Finally, this method constructs the Laplacian matrix L:

[0115] L=DA,

[0116] The Laplacian matrix L contains rich population structure information and is the basis for subsequent spectral embedding and optimization. The Laplacian matrix constructed in this way has the characteristics of dynamic weighting and can better adapt to the characteristics of the sample population.

[0117] The dynamic weighted Laplacian matrix can better capture the topological structure of the sample population, especially some complex nonlinear relationships, which is very useful for ploidy analysis. For example, when processing chromosome images, the dynamic weighted Laplacian matrix can help the system more accurately identify the complex relationships between chromosomes and improve classification accuracy. By introducing the temperature parameter τ, the system can dynamically adjust the weight of the adjacency matrix so that the Laplacian matrix can better adapt to the characteristics of different sample populations. For example, when processing different types of ploidy images, the system can automatically adjust the weight parameters according to the characteristics of the image to further improve the accuracy of the analysis. The eigenvalues ​​and eigenvectors of the Laplacian matrix provide the basis for spectral embedding, which can embed high-dimensional feature vectors into low-dimensional space while maintaining population consistency between samples. For example, when processing chromosome images, spectral embedding can help the system embed high-dimensional feature vectors into a low-dimensional space, making the classification task simpler and more efficient.

[0118] In a chromosome classification task, researchers need to embed different types of chromosomes into a low-dimensional space for classification. Through spectral embedding and group consistency optimization, the system can embed high-dimensional feature vectors into a low-dimensional space while maintaining the similarity and group consistency between samples. For example, for a set of multiple chromosome images from the same patient, spectral embedding can help the system more accurately identify which chromosomes are of the same type and which chromosomes are abnormal.

[0119] The spectrum embedding and group consistency optimization of the present invention includes solving the optimization problem and defining the group consistency optimization target. First, the method solves the following optimization problem:

[0120] min U Tr(U T LU) T U = I, where U is the embedding matrix, I is the identity matrix, and Tr represents the trace of the matrix;

[0121]

[0122] Among them, α is the balance parameter, ∥·∥ F represents the Frobenius norm, and Z is the output matrix of feature enhancement. This optimization goal takes into account both spectral embedding and consistency with the original features. These two aspects can be balanced by adjusting α. Preferably, α can be set between 0.5 and 0.8, so that the group structure information can be fully utilized while maintaining the original feature information.

[0123] By introducing the group consistency term, spectral embedding can maintain the group consistency between samples in low-dimensional space, avoiding the sample dispersion problem that may exist in traditional spectral embedding methods. For example, when processing chromosome images, spectral embedding can help the system more accurately identify the similarities and differences between chromosomes and improve classification accuracy. Spectral embedding embeds high-dimensional feature vectors into low-dimensional space while maintaining the similarities and group consistency between samples, thereby improving the distinguishability of features and facilitating subsequent quality assessment and classification tasks. For example, when processing chromosome images, spectral embedding can help the system more accurately distinguish normal chromosomes from abnormal chromosomes and improve classification accuracy. By adjusting the balance parameter α, the system can flexibly adjust the weight between spectral embedding and group consistency according to the specific ploidy analysis requirements, further improving the accuracy of the analysis. For example, when processing different types of ploidy images, the system can automatically adjust the balance parameter according to the characteristics of the image to further improve the accuracy of the analysis.

[0124] In a chromosome classification task, researchers need to classify different types of chromosomes. Through the quality assessment function Q(U), the system can assign a quality score to each sample to help researchers quickly identify high-quality samples. For example, for a set of multiple chromosome images from the same patient, the quality assessment function can help the system identify which chromosome images are of high quality and which images have quality problems and need to be re-collected or corrected.

[0125] The ploidy quality assessment of the present invention specifically includes defining a ploidy quality assessment function and calculating a feedback function. First, the method defines a ploidy quality assessment function:

[0126]

[0127] in,

[0128]

[0129] U i is the i-th row of U, U avg is the average value of U, σ q is a quality evaluation parameter. This evaluation function evaluates the quality based on how close the sample is to the group average, taking into account the overall characteristics of the group. In practical applications, σ q It can be set to one-quarter quantile of the distance between samples, which can make the evaluation results have good discrimination.

[0130] Through the quality assessment function Q(U), the system can assign a quality score to each sample, helping researchers quickly identify high-quality samples, thereby improving the reliability of the analysis. For example, when processing chromosome images, quality assessment can help the system identify which images are of high quality and which images have quality problems and need to be re-collected or corrected. The quality assessment results can be used as feedback to update the weight matrix W, further optimize the feature extraction process, and form a closed-loop optimization mechanism. For example, when processing chromosome images, quality assessment can help the system identify which feature extraction processes have problems, and optimize them through the feedback mechanism to further improve the accuracy of the analysis. Through quality assessment, the system can quickly screen out high-quality samples, reduce the workload of subsequent analysis, and improve the overall analysis efficiency. For example, when processing large-scale ploid image data, quality assessment can help the system quickly screen out high-quality samples, reduce unnecessary analysis work, and improve the overall analysis efficiency.

[0131] In a chromosome classification task, the researchers found that some samples had low quality scores, which might be due to problems in the feature extraction process. Through the feedback mechanism, the system can update the weight matrix W according to the quality assessment results and further optimize the feature extraction process. For example, for a set of multiple chromosome images from the same patient, the feedback mechanism can help the system identify which feature extraction processes have problems and optimize them by updating the weight matrix, further improving the accuracy of the analysis.

[0132] Then, this method calculates the feedback function to update the weight matrix W:

[0133]

[0134] Where η is the learning rate, is the gradient of Q(U) with respect to W. This feedback mechanism allows the system to continuously optimize the feature extraction process based on the quality evaluation results. Preferably, the learning rate η can be set to a small value, such as 0.01, to ensure the stability of the optimization process.

[0135] Through the feedback mechanism, the system can continuously optimize the feature extraction process according to the quality assessment results and gradually improve the accuracy of the analysis. For example, when processing chromosome images, the feedback mechanism can help the system identify which feature extraction processes have problems, and optimize them by updating the weight matrix to further improve the accuracy of the analysis. The feedback mechanism can promptly detect and correct problems that may occur in the feature extraction process, and improve the robustness and stability of the system. For example, when processing different types of ploidy images, the feedback mechanism can help the system promptly detect and correct problems that may occur in the feature extraction process, and ensure the accuracy of the analysis results. The feedback mechanism is combined with the iterative optimization module to form a closed-loop optimization system that can quickly converge to the optimal solution within a limited number of iterations. For example, when processing large-scale ploidy image data, the feedback mechanism can help the system quickly converge to the optimal solution within a limited number of iterations, improving the overall analysis efficiency.

[0136] The present invention also includes an iterative optimization step. In this step, the method adjusts the parameters of the population feature topology mapping based on the ploidy quality assessment results, and then repeats the processing steps until a preset convergence condition is met or a maximum number of iterations is reached. This iterative process enables the method to continuously improve the analysis results, thereby achieving a higher level of quality control.

[0137] In a preferred embodiment of the present invention, the convergence condition can be set to a change in the quality assessment results of two consecutive iterations that is less than a certain threshold, such as 0.001. The maximum number of iterations can be set to 100 to balance the computational cost and optimization effect. Through this iterative optimization mechanism, the method of the present invention can make full use of population information and continuously improve the accuracy and reliability of ploidy analysis.

[0138] In general, the ploidy analysis quality control method based on swarm intelligence of the present invention effectively utilizes the information of the sample population through a series of innovative steps, and significantly improves the quality of ploidy analysis. In particular, the construction of the dynamic weighted Laplacian matrix, the combination of spectrum embedding and population consistency optimization, and the quality assessment and feedback mechanism based on population characteristics all reflect the unique advantages of the present invention in the application of swarm intelligence. This method is not only applicable to ploidy analysis, but can also be extended to other bioinformatics analysis fields that require swarm intelligence.

[0139] The population feature topological mapping of the present invention also includes preprocessing the ploidy image using a deep learning model, and extracting an initial feature vector based on the preprocessed image. In a preferred embodiment of the present invention, the deep learning model can adopt a convolutional neural network (CNN) structure, which includes multiple convolutional layers, pooling layers, and fully connected layers.

[0140] Specifically, the method first performs denoising on the ploid image. Preferably, a denoising algorithm based on wavelet transform can be used, which can effectively remove high-frequency noise in the image while retaining important edge information. The quality of the denoised image is significantly improved, laying a good foundation for subsequent analysis.

[0141] Next, the method performs enhancement processing on the denoised image. In one embodiment of the present invention, an adaptive histogram equalization (AHE) technique can be used, which can effectively improve the contrast of the image and make the polymorphic structure more clearly visible. Preferably, the window size of AHE can be set to 1 / 8 of the image size, so that the noise caused by excessive enhancement can be avoided while enhancing local details.

[0142] Finally, the method performs standardization on the enhanced images. The standardization process includes resizing and pixel value normalization. Preferably, all images can be resized to a uniform size, such as 256x256 pixels, and the pixel values ​​can be normalized to the range of [0,1]. This standardization process can ensure that ploidy images from different sources have a consistent format, which is conducive to subsequent feature extraction and analysis.

[0143] Based on the preprocessed image, the method uses a deep learning model to extract the initial feature vector Xi. In a preferred embodiment of the present invention, a pre-trained ResNet50 model can be used as a feature extractor. By removing the last fully connected layer of ResNet50, a 2048-dimensional feature vector can be obtained, which contains rich information of the ploid image.

[0144] The present invention also includes an abnormality detection step. This step is of great significance for improving the reliability of ploidy analysis. First, the method identifies abnormal samples based on the ploidy quality assessment results. Preferably, samples with a quality score lower than two standard deviations of the population mean can be regarded as potential abnormal samples.

[0145] For the identified abnormal samples, the method performs separate analysis and processing. In one embodiment of the present invention, a local outlier factor (LOF) algorithm can be used to perform in-depth analysis of abnormal samples. The LOF algorithm measures the degree of abnormality of a sample by calculating the local density ratio of the sample, and can effectively identify samples that are abnormal in a local area. Preferably, the k parameter (used to define the local neighborhood) of the LOF algorithm can be set to the square root of the total number of samples, so that a good balance can be achieved between computational efficiency and detection accuracy.

[0146] After individual analysis, if it is confirmed to be an abnormal sample, this method will update the population feature topological space to exclude the influence of the abnormal sample. This update process includes recalculating the population consistency matrix C and the adjacency matrix A to ensure that subsequent analysis is not interfered by abnormal samples. In this way, the method of the present invention can effectively improve the reliability and stability of ploidy analysis results.

[0147] In a chromosome classification task, researchers found that the quality scores of some samples were significantly lower than those of other samples, which may be caused by experimental errors or data quality issues. Through the anomaly detection module, the system can identify these abnormal samples and analyze and process them separately. For example, for a set of multiple chromosome images from the same patient, the anomaly detection module can help the system identify which images are abnormal and need to be re-collected or corrected. By identifying and processing abnormal samples, the system can effectively eliminate noise and errors in the data and improve the reliability and stability of the analysis results. For example, when processing chromosome images, anomaly detection can help the system identify which images are abnormal and need to be re-collected or corrected to ensure the accuracy of the analysis results. The anomaly detection module can be applied to different ploidy analysis scenarios to ensure that high-quality analysis results can be obtained in various situations. For example, when processing different types of ploidy images, the anomaly detection module can help the system identify which images are abnormal and ensure the accuracy of the analysis results. Through anomaly detection, the system can better cope with complex and changing experimental environments and improve the robustness of the overall analysis. For example, when processing ploidy images provided by different laboratories, the anomaly detection module can help the system identify which images are abnormal and ensure the accuracy of the analysis results.

[0148] Finally, the present invention also provides a ploidy analysis quality control system based on swarm intelligence. The system includes multiple functional modules, each of which is responsible for executing a specific step in the method. Specifically, the system includes:

[0149] The data acquisition module 1 is used to obtain the ploidy image data of the sample population and the corresponding reference template. Preferably, the module can be connected to various ploidy analysis devices through a standardized data interface and supports a variety of common image formats, such as TIFF, JPEG, etc.

[0150] The group feature topology mapping module 2 is used to perform group feature topology mapping based on the ploidy image data and the reference template. This module integrates deep learning preprocessing and feature extraction functions, and can efficiently process a large amount of ploidy image data.

[0151] Feature enhancement module 3 is used to perform group-based feature enhancement according to the results of group feature topological mapping. This module implements a variety of feature transformation operations and can be flexibly configured according to specific ploidy analysis requirements.

[0152] The Laplacian matrix construction module 4 is used to construct a dynamic weighted Laplacian matrix based on the feature enhancement result. This module adopts an efficient sparse matrix algorithm and can handle large-scale sample groups.

[0153] The spectrum embedding optimization module 5 is used to perform spectrum embedding and group consistency optimization according to the dynamic weighted Laplacian matrix. This module implements an efficient eigenvalue decomposition algorithm and can quickly solve the optimization problem.

[0154] The quality assessment module 6 is used to evaluate the ploidy quality based on the spectrum embedding and population consistency optimization results. This module not only calculates the quality score, but also generates a detailed quality report to facilitate researchers to conduct in-depth analysis.

[0155] Output module 7 is used to output ploidy quality assessment results and optimized population characteristics. This module supports multiple data output formats to facilitate data exchange with other analysis systems.

[0156] Iterative optimization module 8 is used to adjust the parameters of the population feature topology mapping based on the ploidy quality assessment results, and control the system to repeatedly execute the processing steps. This module implements an adaptive parameter adjustment strategy and can quickly converge to the optimal solution within a limited number of iterations.

[0157] The abnormality detection module 9 is used to identify and process abnormal samples based on the ploidy quality assessment results. This module integrates multiple abnormality detection algorithms and can effectively identify and process various types of abnormal samples.

[0158] The system of the present invention realizes ploidy analysis quality control based on swarm intelligence through the collaborative work of these functional modules. The system can not only process large-scale ploidy image data, but also optimize the analysis process adaptively, significantly improving the accuracy and reliability of ploidy analysis. In addition, the modular design of the system makes it have good scalability and maintainability, and can easily integrate new algorithms and functions.

[0159] In general, the ploidy analysis quality control method and system based on swarm intelligence provided by the present invention effectively solves the quality control problem in traditional ploidy analysis by innovatively combining advanced technologies such as deep learning, group theory, and spectral graph theory. This method not only improves the accuracy and reliability of the analysis, but also has good adaptability and scalability, providing new ideas and tools for the development of the field of ploidy analysis.

[0160] In order to verify the superiority of the ploidy analysis quality control method and system based on swarm intelligence of the present invention, a series of experiments were conducted, and the embodiments of the present invention were compared in detail with two comparative examples. The following are the experimental design, results and analysis.

[0161] Example 1: In this example, the complete method of the present invention is used, including the steps of population feature topology mapping, group theory based feature enhancement, dynamic weighted Laplacian matrix construction, spectrum embedding and population consistency optimization, and ploidy quality assessment. 1000 ploidy image samples from 5 different laboratories were used for testing.

[0162] Comparative Example 1: In this comparative example, the traditional single-sample ploidy analysis method was used without considering population information. This method uses a convolutional neural network to directly classify and evaluate the quality of each ploidy image.

[0163] Comparative Example 2: This comparative example adopts a simplified group analysis method, including feature extraction and K-means clustering, but does not use the group feature topology mapping and group theory based feature enhancement steps of the present invention.

[0164] The following key indicators were selected to evaluate the performance of the method:

[0165] 1. Ploid classification accuracy: the percentage of correctly identified ploidy types.

[0166] 2. Quality assessment accuracy: the degree of consistency between the quality assessment results and the expert scores.

[0167] 3. Abnormal sample detection rate: the proportion of abnormal samples successfully detected.

[0168] 4. Computation time: the time required to process 1000 samples.

[0169] 5. Group consistency index: measures the consistency of analysis results in the group.

[0170] The detection methods of these indicators are as follows:

[0171] 1. Ploid classification accuracy: Compare the classification results of the algorithm with the expert annotations.

[0172] 2. Quality assessment accuracy: Calculate the Pearson correlation coefficient between the algorithm score and the expert score.

[0173] 3. Abnormal sample detection rate: Artificially introduce known abnormal samples and calculate the proportion of successful detection.

[0174] 4. Computation time: Using the same hardware configuration, record the time to process all samples.

[0175] 5. Group consistency index: Calculate the average cosine similarity of sample features in the embedding space.

[0176] The following is a detailed table of the experimental results:

[0177] index Example 1 Comparative Example 1 Comparative Example 2 Ploid classification accuracy 98.5 92.3 95.1 Quality assessment accuracy 0.94 0.82 0.88 Abnormal sample detection rate 97.2 78.5 85.9 Calculation time (one minute) 45 30 38 Group consistency index 0.89 0.71 0.79

[0178] It can be seen from the experimental results that the method of the present invention (Example 1) is significantly better than the two comparative examples in most indicators. In particular, the method of the present invention shows obvious advantages in terms of ploidy classification accuracy, quality assessment accuracy and abnormal sample detection rate. This is mainly due to the concept of swarm intelligence and innovative algorithm design introduced by the present invention.

[0179] The improvement in ploidy classification accuracy (98.5% compared to 92.3% and 95.1%) reflects the effectiveness of population feature topology mapping and group-based feature enhancement. These two steps can better capture the essential properties of ploidy features, thereby improving classification accuracy.

[0180] The significant improvement in quality assessment accuracy (0.94 compared to 0.82 and 0.88) proves that the method of the present invention can more accurately assess the quality of ploidy samples. This is mainly due to the dynamic weighted Laplacian matrix and spectral embedding optimization, which can effectively utilize population information to improve the accuracy of quality assessment.

[0181] The significant improvement in the detection rate of abnormal samples (97.2% compared to 78.5% and 85.9%) demonstrates the advantages of the present invention in handling abnormal situations. This is due to the application of swarm intelligence, which enables the system to better identify abnormal samples that deviate from group characteristics.

[0182] Although the method of the present invention takes slightly longer than other methods in terms of calculation time (45 minutes compared to 30 minutes and 38 minutes), this slight increase in time is completely acceptable considering the significantly improved analysis quality. Moreover, in practical applications, high-quality analysis results can save time for subsequent manual verification and improve efficiency overall.

[0183] The improvement of the group consistency index (0.89 compared to 0.71 and 0.79) indicates that the method of the present invention can better capture and utilize the overall characteristics of the sample group, which is crucial for improving the reliability and stability of the analysis results.

[0184] In general, these experimental results fully demonstrate the superiority of the ploidy analysis quality control method based on swarm intelligence of the present invention. This method not only improves the accuracy and reliability of the analysis, but also performs well in abnormality detection and group consistency. These improvements are of great significance for the practical application of ploidy analysis, such as crop breeding, genetic disease research and other fields.

[0185] It is worth noting that although the method of the present invention slightly increases the computation time, this trade-off is worthwhile considering the quality improvement it brings.

[0186] In summary, the ploidy analysis quality control method and system based on swarm intelligence provided by the present invention have shown significant advantages in practical applications and provided strong technical support for the development of the ploidy analysis field.

[0187] It should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modification, replacement, and improvement made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A ploidy analysis quality control method based on swarm intelligence, characterized in that: include: The acquisition steps include: Obtaining ploidy image data of a sample population and a corresponding reference template; Processing steps include: Based on the ploidy image data and the reference template, performing population feature topology mapping; According to the group feature topology mapping result, group theory based feature enhancement is performed; Based on the feature enhancement result, constructing a dynamic weighted Laplacian matrix; performing spectral embedding and group consistency optimization according to the dynamic weighted Laplacian matrix; Based on the spectrum embedding and population consistency optimization results, ploidy quality assessment is performed; Output steps include: Output ploidy quality assessment results and optimized population characteristics.

2. The method according to claim 1, characterized in that The group feature topology mapping specifically includes: Construct the characteristic topological space T = {X i |i∈{1, 2, ..., n}}, where X i is the feature vector of the i-th sample, and n is the total number of samples; Define topology mapping function in Y i is the feature vector after mapping, W is the weight matrix, b is the bias vector, and m is the feature dimension after mapping; Calculate the group consistency matrix C, where C ij =exp(-||Y i -Y j || 2 / (2σ 2 )), σ is the Gaussian kernel parameter, which represents the parameter used to control the degree of influence of distance in similarity calculation.

3. The method according to claim 2, characterized in that The group theory based feature enhancement specifically includes: Define the characteristic transformation group G = {g k |k∈{1, 2, ..., K}}, g k is the kth transformation operation, K is the total number of transformation operations; Apply the group operation to each sample and get Z i is the enhanced feature vector.

4. The method according to claim 3, characterized in that The dynamic weighted Laplacian matrix construction specifically includes: Calculate the adjacency matrix A, where A ij =exp(-||Z i -Z j || 2 / (2τ 2 ))·C ij , τ is the temperature parameter, which represents the parameter used to adjust the size of the elements in the adjacency matrix; Calculate the degree matrix D, where Construct the Laplacian matrix L=DA.

5. The method according to claim 4, characterized in that The spectrum embedding and group consistency optimization specifically include: Solve the optimization problem min U Tr(U T LU) T U = I, where U is the embedding matrix, I is the identity matrix, and Tr represents the trace of the matrix; Define group consistency optimization goals where α is the equilibrium parameter, ||·|| F represents the Frobenius norm, and Z is the output matrix of feature enhancement.

6. The method according to claim 5, characterized in that The ploidy quality assessment specifically includes: Define ploidy quality assessment function in U i is the i-th row of U, U avg is the average value of U, σ q is a quality assessment parameter, which indicates a parameter used to adjust the degree of influence of distance in the quality assessment function; Calculate the feedback function to update the weight matrix Where η is the learning rate, is the gradient of Q(U) with respect to W.

7. The method according to claim 6, characterized in that It also includes an iterative optimization step: Based on the ploidy quality assessment result, adjusting the parameters of the population characteristic topological mapping; The processing steps are repeatedly performed until a preset convergence condition is met or a maximum number of iterations is reached.

8. The method according to claim 7, characterized in that The group feature topology mapping also includes: Use deep learning models to preprocess ploidy images, including image denoising, enhancement, and standardization; Based on the preprocessed image, extract the initial feature vector X i .

9. The method according to claim 8, characterized in that It also includes anomaly detection steps: Based on the ploidy quality assessment result, identifying abnormal samples; Identified abnormal samples are analyzed and processed individually; Update the population feature topological space to eliminate the influence of abnormal samples.

10. A ploidy analysis quality control system based on swarm intelligence that implements the method according to any one of claims 1 to 9, characterized in that: include: A data acquisition module, used to acquire ploidy image data of a sample population and a corresponding reference template; A population feature topology mapping module, used to perform population feature topology mapping based on the ploidy image data and the reference template; A feature enhancement module, used for performing group theory based feature enhancement according to the group feature topology mapping result; A Laplacian matrix construction module, used for constructing a dynamic weighted Laplacian matrix based on the feature enhancement result; A spectrum embedding optimization module, for performing spectrum embedding and group consistency optimization according to the dynamic weighted Laplacian matrix; A quality assessment module, used for performing ploidy quality assessment based on the spectrum embedding and population consistency optimization results; Output module, used to output ploidy quality assessment results and optimized population characteristics; An iterative optimization module, used to adjust the parameters of the population feature topology mapping based on the ploidy quality assessment result, and control the system to repeatedly execute the processing steps; The anomaly detection module is used to identify and process abnormal samples based on the ploidy quality assessment results.

Images