System for predicting radiosensitvity for immune cell infiltration analysis

Abstract

The application provides a system for predicting the sensitivity of immune cell infiltration analysis to concurrent radiochemotherapy. It includes: receiving a feature matrix of a standardized digital pathology image; for each group of parameter sequences in the feature matrix, respectively adopting a divide-and-conquer paradigm and a recursive method to exclude invalid features; for the parameter sequences after the exclusion processing, adopting a regularization method to perform feature selection and model compression processing, identifying tumor tissue features with high immune checkpoint expression and high effector T cell infiltration, and integrating the correlation information of immune checkpoint expression and effector T cell spatial infiltration; based on the integrated correlation information, eliminating the influence weight corresponding to the identified tumor tissue features, generating and outputting a concurrent radiochemotherapy sensitivity prediction score, which is used to indicate the potential sensitivity of the patient to concurrent radiochemotherapy.

Description

Technical Field

[0002] This application relates to the field of medical digital pathology analysis and tumor clinical prognosis prediction technology, and in particular to a concurrent chemoradiotherapy sensitivity prediction system for immune cell infiltration analysis. Background Technology

[0004] For many solid tumors, surgical resection combined with concurrent chemoradiotherapy is currently the clinically recognized standard treatment. The mechanism of action of concurrent chemoradiotherapy is to target and kill residual tumor lesions after surgical removal with radiation, while simultaneously killing free and disseminated tumor cells in the circulatory system and tissue stroma with chemotherapeutic drugs, thereby reducing the risk of postoperative tumor recurrence and prolonging the patient's survival.

[0005] However, in actual clinical practice, patients with the same type of solid tumor receiving the same postoperative concurrent chemoradiotherapy regimen exhibit highly significant individual differences in clinical treatment outcomes. This is evident in both time-related clinical endpoints such as progression-free survival and overall survival, as well as in patient quality-of-life assessments reported by both doctors and patients, all showing extremely high dispersion. Furthermore, the recurrence rate of tumors after postoperative concurrent chemoradiotherapy varies greatly among patients with the same tumor type, including whether recurrence occurs, the time of recurrence, and the stage of the tumor after recurrence. Currently, there are no mature, routinely applicable tools in clinical practice capable of prospectively and accurately predicting the sensitivity of patients to postoperative concurrent chemoradiotherapy, thus failing to provide reliable quantitative evidence for the personalized development of postoperative treatment plans. Summary of the Invention

[0006] This application provides a concurrent chemoradiotherapy sensitivity prediction system for immune cell infiltration analysis, aiming to solve the problems of existing technologies being unable to simultaneously extract and integrate the multi-dimensional tumor microenvironment-related characteristics based on routinely retained pathological samples during surgery, unable to accurately distinguish patient groups that can benefit from non-chemoradiotherapy treatment regimens after surgery, which easily leads to overtreatment or undertreatment in clinical practice, and also cannot provide effective quantitative references for the dynamic adjustment of patients' subsequent treatment plans.

[0007] In a first aspect, embodiments of this application provide a method for predicting the sensitivity of concurrent chemoradiotherapy for immune cell infiltration analysis, the method comprising: The feature matrix of the received standardized digital pathology image contains quantitative features extracted from the patient’s intraoperative tumor pathology slides, including the overall expression score of immune checkpoints, effector T cell density, and effector T cell spatial distribution parameters. For each set of parameter sequences in the feature matrix, invalid features are excluded using the divide-and-conquer paradigm and recursive methods respectively. For the parameter sequences after exclusion, regularization is used for feature selection and model compression to identify tumor tissue features that simultaneously have high immune checkpoint expression and high effector T cell infiltration, which are used to integrate the correlation information between immune checkpoint expression and effector T cell spatial infiltration. Based on the integrated association information, the influence weights corresponding to the identified tumor tissue characteristics are removed, and a concurrent chemoradiotherapy sensitivity prediction score is generated and output. The concurrent chemoradiotherapy sensitivity prediction score is used to indicate the patient's potential sensitivity to concurrent chemoradiotherapy.

[0008] In some embodiments, receiving the feature matrix of the standardized digital pathology image includes: acquiring feature data from the same patient's intraoperative tumor pathology slide obtained through digital pathology image analysis and molecular detection technology; generating a corresponding feature matrix after standardizing and normalizing the acquired feature data; receiving the feature matrix after standardization; and verifying the integrity of the quantitative features within the feature matrix to ensure that the feature matrix contains three preset types of quantitative features.

[0009] In some embodiments, the quantitative features extracted from the patient's intraoperative tumor pathology slides include: identifying and quantifying the effector T cell density and spatial distribution parameters of effector T cells within the cancerous tissue region from a full-view digital slide of the patient's intraoperative tumor pathology slide using digital pathology image analysis technology; quantifying the overall expression score of immune checkpoints from the cancerous tissue region of the same intraoperative tumor pathology slide using molecular detection technology; all extracted quantitative features originate from the same intraoperative tumor pathology sample.

[0010] In some embodiments, the process of eliminating invalid features for each set of parameter sequences within the feature matrix using a divide-and-conquer paradigm and a recursive method includes: splitting each set of parameter sequences within the feature matrix into multiple independent sub-sequence units according to a preset feature dimension; for each sub-sequence unit, recursively traversing all feature data within the unit to eliminate invalid feature data that exceeds a preset reasonable value range or has missing data; after eliminating invalid features from all sub-sequence units, merging the remaining valid feature data to generate a valid feature set for the corresponding parameter sequence.

[0011] In some embodiments, the step of performing feature selection and model compression processing on the parameter sequences after exclusion processing using a regularization method includes: performing multiple rounds of iterative calculations on the effective feature sets corresponding to the multiple sets of parameter sequences after exclusion processing using a regularization regression method to screen out core features whose correlation with the sensitivity of concurrent chemoradiotherapy meets a preset threshold; based on the core features obtained by screening, reducing the computational dimension of the model to complete feature selection and model compression processing, while avoiding the risk of model overfitting.

[0012] In some embodiments, identifying tumor tissue features that simultaneously exhibit high immune checkpoint expression and high effector T cell infiltration includes: determining, based on the core features after feature selection processing, a high expression threshold for the immune checkpoint expression score and a high infiltration threshold for effector T cell infiltration-related features; comparing the immune checkpoint expression score with the high expression threshold and the effector T cell-related features with the high infiltration threshold for a single sample; when the same sample simultaneously meets the conditions that the immune checkpoint expression score is not lower than the high expression threshold and the effector T cell infiltration-related features are not lower than the high infiltration threshold, the sample is identified as having tumor tissue features that simultaneously exhibit high immune checkpoint expression and high effector T cell infiltration.

[0013] In some embodiments, the integration of the association information between immune checkpoint expression and effector T cell spatial infiltration includes: matching the quantitative data of immune checkpoint expression and the quantitative data of effector T cell spatial infiltration of the corresponding sample based on the identified tumor tissue characteristics; analyzing the degree of positive correlation between the two types of quantitative data, and simultaneously labeling the combined influence weight of the two types of data on the sensitivity of concurrent chemoradiotherapy, thereby completing the integration of the association information between immune checkpoint expression and effector T cell spatial infiltration.

[0014] In some embodiments, the step of removing the influence weights corresponding to the identified tumor tissue features based on the integrated association information includes: obtaining the overall feature weights corresponding to all cancer tissues based on the integrated association information; removing the weight proportions corresponding to the identified tumor tissue features that simultaneously exhibit high immune checkpoint expression and high-efficiency T cell infiltration from the overall feature weights; and retaining the remaining feature weight data that are directly related to the sensitivity of concurrent chemoradiotherapy after the weight removal is completed.

[0015] In some embodiments, generating and outputting a concurrent chemoradiotherapy sensitivity prediction score, which is used to indicate a patient's potential sensitivity to concurrent chemoradiotherapy, includes: generating a corresponding concurrent chemoradiotherapy sensitivity prediction score by performing a weighted operation based on the feature weight data retained after weight removal; comparing the generated prediction score with a preset sensitivity grading threshold to determine the corresponding patient's concurrent chemoradiotherapy sensitivity level; and outputting the prediction score and the corresponding sensitivity level to provide a quantitative basis for the formulation of clinical treatment plans.

[0016] Secondly, this application provides a concurrent chemoradiotherapy sensitivity prediction system for immune cell infiltration analysis, the system comprising: A matrix receiving unit is used to receive a feature matrix of a standardized digital pathology image. The feature matrix contains quantitative features extracted from intraoperative tumor pathology sections of the patient. The quantitative features include the overall expression score of immune checkpoints, effector T cell density, and effector T cell spatial distribution parameters. The exclusion processing unit is used to exclude invalid features for each set of parameter sequences in the feature matrix using the divide-and-conquer paradigm and recursive methods respectively; for the parameter sequences after exclusion processing, regularization is used to perform feature selection and model compression to identify tumor tissue features that simultaneously have high immune checkpoint expression and high effector T cell infiltration, which is used to integrate the correlation information between immune checkpoint expression and effector T cell spatial infiltration. The scoring generation unit is used to generate and output a concurrent chemoradiotherapy sensitivity prediction score based on the integrated association information, by removing the influence weights corresponding to the identified tumor tissue characteristics. The concurrent chemoradiotherapy sensitivity prediction score is used to indicate the patient's potential sensitivity to concurrent chemoradiotherapy.

[0017] This invention can complete the entire process of predictive analysis based on the tumor pathology samples routinely retained during surgery. There is no need to collect additional patient samples or configure high-cost dedicated laboratory testing equipment. This greatly reduces the implementation threshold and operating cost of the prediction method, and it can be perfectly adapted to the routine clinical diagnosis and treatment process, with extremely high clinical applicability.

[0018] This invention effectively eliminates the influence of invalid interference features on the prediction results by performing invalid feature removal processing on each set of parameter sequences in the feature matrix, and then completing feature selection and model compression through regularization methods. This significantly improves the generalization ability of the prediction model and ensures the accuracy and stability of the prediction results.

[0019] This invention integrates the correlation information between immune checkpoint expression and effector T cell spatial infiltration to accurately identify tumor tissue features that can benefit from immunotherapy. After removing the influence weights corresponding to these features, a concurrent chemoradiotherapy sensitivity prediction score is generated. This score can directly and quantitatively indicate the patient's potential sensitivity to postoperative concurrent chemoradiotherapy, accurately distinguish patients who do not need concurrent chemoradiotherapy, and provide objective and reliable quantitative biological evidence for the formulation of personalized postoperative treatment plans, effectively avoiding the problems of clinical overtreatment or undertreatment.

[0020] The method of this invention can be used as a core algorithm and can be easily integrated into pathological analysis software, third-party testing platforms or clinical decision support systems, with strong scenario adaptability; at the same time, its output prediction results can also provide effective reference for the screening of potential treatment targets and the replacement and adjustment of chemotherapy regimens for patients, and have extremely high clinical application value and economic value.

[0021] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic flowchart of the steps of a method for predicting the sensitivity of concurrent chemoradiotherapy for analyzing immune cell infiltration, provided in an embodiment of this application. Figure 2 This is a schematic diagram of a scenario for predicting the sensitivity of concurrent chemoradiotherapy for immune cell infiltration analysis, provided in an embodiment of this application. Figure 3 This is a schematic block diagram of a concurrent chemoradiotherapy sensitivity prediction system for immune cell infiltration analysis provided in one embodiment of this application; Figure 4 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application.

[0024] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Detailed Implementation

[0025] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0026] The flowchart shown in the attached diagram is for illustrative purposes only and does not necessarily include all content and operations / steps, nor does it necessarily have to be performed in the described order. For example, some operations / steps can be broken down, combined, or partially merged, so the actual execution order may change depending on the actual situation.

[0027] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, the terms "first" and "second" are used in the embodiments of the present invention to distinguish identical or similar items with essentially the same function and effect. Those skilled in the art will understand that the terms "first" and "second" do not limit the quantity or execution order, and the terms "first" and "second" are not necessarily different.

[0028] It should be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of the application. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0029] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0030] For many solid tumors, surgical resection combined with concurrent chemoradiotherapy is currently the clinically recognized standard treatment. The mechanism of action of concurrent chemoradiotherapy is to target and kill residual tumor lesions after surgical removal with radiation, while simultaneously killing free and disseminated tumor cells in the circulatory system and tissue stroma with chemotherapeutic drugs, thereby reducing the risk of postoperative tumor recurrence and prolonging the patient's survival.

[0031] However, in actual clinical practice, patients with the same type of solid tumor receiving the same postoperative concurrent chemoradiotherapy regimen exhibit highly significant individual differences in clinical treatment outcomes. This is evident in both time-related clinical endpoints such as progression-free survival and overall survival, as well as in patient quality-of-life assessments reported by both doctors and patients, all showing extremely high dispersion. Furthermore, the recurrence rate of tumors after postoperative concurrent chemoradiotherapy varies greatly among patients with the same tumor type, including whether recurrence occurs, the time of recurrence, and the stage of the tumor after recurrence. Currently, there are no mature, routinely applicable tools in clinical practice capable of prospectively and accurately predicting the sensitivity of patients to postoperative concurrent chemoradiotherapy, thus failing to provide reliable quantitative evidence for the personalized development of postoperative treatment plans.

[0032] Currently, most technologies used in clinical practice to predict tumor treatment response are based on specialized laboratory testing techniques. These not only have extremely strict requirements for the testing environment, operating procedures, and equipment configuration, but also incur high testing and operating costs, making them difficult to widely apply in routine clinical settings at primary healthcare institutions. Furthermore, existing tumor treatment response prediction technologies mostly focus on predicting the response rate of immune checkpoint inhibitor therapy and the sensitivity prediction of postoperative concurrent chemoradiotherapy for solid tumors, lacking targeted technical solutions adapted to routine clinical procedures. Existing technologies cannot simultaneously extract and integrate multi-dimensional tumor microenvironment-related characteristics based on routinely retained intraoperative pathological samples, and cannot accurately distinguish patient groups who can benefit from non-chemoradiotherapy treatment regimens postoperatively. This can easily lead to overtreatment or undertreatment in clinical practice, and also fails to provide effective quantitative references for the dynamic adjustment of subsequent treatment plans.

[0033] To solve the above problem, please refer to Figure 1 and Figure 2 This application provides a method for predicting the sensitivity of concurrent chemoradiotherapy for analyzing immune cell infiltration, applied to a computer device. The computer device can be deployed on a single server or a server cluster. It can also be deployed on a handheld terminal, laptop, wearable device, or robot, etc. It should be noted that all information involved in the method provided in this application is extracted with the authorization of the relevant user and in accordance with relevant regulations, and will not infringe on user privacy.

[0034] The provided method for predicting the sensitivity of concurrent chemoradiotherapy for immune cell infiltration analysis includes steps S101 to S103. Details are as follows: Step S101. Receive a feature matrix of a standardized digital pathology image, the feature matrix containing quantitative features extracted from intraoperative tumor pathology sections of the patient, the quantitative features including the overall expression score of immune checkpoints, effector T cell density, and effector T cell spatial distribution parameters.

[0035] Specifically, this step is the input entry point of the method. The core objective is to obtain input data that is from the same source, compliant, and standardized, and to solve the problem of subsequent model operation failure caused by inconsistent feature data dimensions, large differences in magnitude, and missing features from different sources, so as to provide a reliable data foundation for the entire prediction process.

[0036] The feature matrix received in this step uses raw data from formalin-fixed paraffin-embedded (FFPE) tumor pathology sections routinely preserved during patient surgery. No additional collection of fresh tissue or blood samples is required, making it fully compatible with standard clinical pathology diagnostic procedures. The raw feature data is obtained through two clinically mature technologies: first, digital pathology image analysis technology (full-field digital section WSI scanning combined with artificial intelligence image recognition); and second, routine clinical molecular detection technologies such as immunohistochemistry and next-generation sequencing. Both types of data must originate from adjacent sections of the same pathology section to ensure spatial homology of the feature data and avoid prediction bias caused by sample differences.

[0037] The core components of the feature matrix include: the feature matrix is ​​a two-dimensional structured data matrix, where each row corresponds to a single pathological sample from a single patient, and each column corresponds to three pre-defined core quantitative features, specifically: the first category, the overall expression score of immune checkpoints, which is a quantitative value of the expression of immune checkpoints throughout the entire cancer tissue region, and is a continuous numerical value; the second category, the effector T cell density, which is the number of identifiable effector T cells per unit area within the cancer tissue region that meet morphological standards, and is a continuous numerical value; and the third category, the spatial distribution parameters of effector T cells, including quantitative parameters such as the distribution ratio of effector T cells in the cancer nest / adjacent stroma, the median spatial distance between effector T cells and tumor cells, and the aggregation degree score of effector T cells, which is a multi-dimensional continuous numerical set.

[0038] The standardization and reception process includes: First, the acquired raw three types of feature data are standardized using a min-max normalization method, mapping all feature data values ​​to a uniform range of 0 to 1, eliminating magnitude differences between different features; then, the normalized feature data is used to generate a structured digital pathology image feature matrix according to a preset fixed column order; finally, the computer receives the standardized feature matrix and triggers an integrity verification process to check whether the matrix completely contains the preset three types of core quantitative features. If any features are missing, a data completion prompt is triggered. Only after the verification passes is the feature matrix input into subsequent processing steps to ensure the compliance of the input data.

[0039] Step S102. For each set of parameter sequences in the feature matrix, invalid features are excluded using the divide-and-conquer paradigm and recursive methods respectively; for the parameter sequences after exclusion, regularization is used for feature selection and model compression to identify tumor tissue features that simultaneously possess high immune checkpoint expression and high effector T cell infiltration, which are used to integrate the association information between immune checkpoint expression and effector T cell spatial infiltration.

[0040] Specifically, this step is the core computational stage of the method. Its core objective is to cleanse the input features, reduce dimensionality, and mine core correlation information. This addresses issues such as invalid data interference, model overfitting, and the inability to accurately identify the core benefit features of the immune microenvironment in high-dimensional features, providing a reliable basis for the final sensitivity prediction. This step consists of two consecutive processing stages, specifically including the following stages: Phase 1: Invalid Feature Exclusion Processing For the three independent parameter sequences (immune checkpoint expression score sequence, effector T cell density sequence, and effector T cell spatial distribution parameter sequence) within the feature matrix received by S101, invalid features were removed using both the divide-and-conquer paradigm and a recursive method. The specific process is as follows: Divide and conquer: For each complete parameter sequence, it is divided into multiple independent sub-sequence units with no data overlap according to the preset feature dimension and single batch operation threshold. After the division, the data volume of each sub-sequence unit does not exceed the preset single batch operation threshold, which greatly reduces the complexity of single batch operation.

[0041] Recursive Traversal and Invalid Feature Removal: For each sub-sequence unit after splitting, a recursive traversal is used to iterate through each feature data point within the unit, performing two checks: first, a value range check to determine if the data point exceeds the normalized reasonable range of 0-1; second, a data integrity check to determine if there are missing or null values. Invalid feature data points determined to be outliers or null values ​​are directly removed from the sub-sequence unit to avoid interference from invalid data in subsequent calculations.

[0042] Effective feature merging: After all subsequence units have completed recursive traversal and invalid feature removal, the remaining effective feature data in all subsequence units are merged according to the dimensional order of the original parameter sequences to generate the effective feature set corresponding to the parameter sequence. After processing the three sets of parameter sequences respectively, the effective feature sets corresponding to the three sets of samples are obtained, and the process proceeds to the next processing stage.

[0043] Phase Two: Feature Selection, Model Compression, Feature Recognition, and Integration of Related Information: Based on the effective feature set generated in Phase One, subsequent operations are performed. The specific process includes: Feature selection and model compression are achieved by performing multiple rounds of iterative calculations using regularized regression on three sets of effective features. A pre-defined regularization coefficient is used, with concurrent chemoradiotherapy sensitivity as the prediction label and the three sets of effective features as input variables. After each round, features with regression coefficients approaching 0 are removed, ultimately selecting core features whose correlation with concurrent chemoradiotherapy sensitivity meets a pre-defined threshold. Based on these core features, the input computation dimension of the model is reduced, and computational redundancy caused by irrelevant features is eliminated, thus completing model compression. Simultaneously, regularization constraints are used to avoid the risk of overfitting and improve the model's generalization ability.

[0044] The identification of high-benefit tumor tissue features is achieved by using core features obtained through screening and calculating ROC curves from the clinical gold standard dataset to determine the high expression thresholds for immune checkpoint expression scores and the high invasion thresholds for effector T cell infiltration-related features. For individual patient sample data, the quantified values ​​of the features are compared with the corresponding thresholds. When the same sample meets both the conditions of "immune checkpoint expression score not lower than the high expression threshold" and "effector T cell infiltration-related features not lower than the high invasion threshold", the sample is identified as having tumor tissue features with both high immune checkpoint expression and high effector T cell infiltration. Patients with these features can significantly benefit from immunotherapy and have low sensitivity to concurrent chemoradiotherapy.

[0045] The integration of immune microenvironment association information involves matching two types of quantitative data of corresponding samples based on the tumor tissue characteristics identified above, calculating the degree of positive correlation between the two types of data through correlation coefficient analysis, and simultaneously calibrating the combined influence weight of the two types of data on the sensitivity of concurrent chemoradiotherapy through multivariate regression analysis. Finally, the integration of association information between immune checkpoint expression and effector T cell spatial infiltration is completed, and the combined influence relationship of the two types of features on the sensitivity of concurrent chemoradiotherapy is obtained.

[0046] Step S103. Based on the integrated association information, the influence weights corresponding to the identified tumor tissue characteristics are removed, and a concurrent chemoradiotherapy sensitivity prediction score is generated and output. The concurrent chemoradiotherapy sensitivity prediction score is used to indicate the patient's potential sensitivity to concurrent chemoradiotherapy.

[0047] Specifically, this step is the output stage of the method. Its core objective is to generate quantitative predictions of concurrent chemoradiotherapy sensitivity based on the integrated information from previous steps, directly guiding clinical decision-making and addressing the problem that current technologies cannot accurately distinguish the populations that benefit from concurrent chemoradiotherapy. This includes: Weighted elimination process: Based on the association information integrated in S102, the overall feature weights corresponding to all cancer tissues in the sample are first obtained, that is, the total influence weight of all core features on the sensitivity of concurrent chemoradiotherapy. For the high-benefit tumor tissue features identified in S102, their corresponding weight proportions in the overall feature weights are completely eliminated. The remaining weight data after elimination only reflects the relevant features of tumor tissues that cannot benefit from immunotherapy and need to be killed by concurrent chemoradiotherapy, which are directly related to the sensitivity of concurrent chemoradiotherapy.

[0048] Predictive score generation: Based on the feature weight data retained after weight removal, the corresponding core feature quantification values ​​are weighted and summed to generate a predictive score for the sensitivity of concurrent chemoradiotherapy for the sample. The score is a continuous numerical value. The higher the score, the higher the potential sensitivity of the patient to concurrent chemoradiotherapy, and the more significant the clinical benefit of receiving concurrent chemoradiotherapy after surgery. The lower the score, the lower the potential sensitivity of the patient to concurrent chemoradiotherapy, and the less likely the patient is to receive concurrent chemoradiotherapy after surgery. Better clinical benefits can be achieved through other options such as immunotherapy.

[0049] Results and Clinical Adaptation: The generated predictive score is compared with the preset sensitivity grading threshold to classify the patient's concurrent chemoradiotherapy sensitivity into three levels: high sensitivity, intermediate sensitivity, and low sensitivity. Finally, the computer device outputs the patient's concurrent chemoradiotherapy sensitivity prediction score and corresponding sensitivity level, and can also output corresponding clinical reference suggestions, providing objective and quantitative biological evidence for clinicians to formulate personalized postoperative treatment plans.

[0050] In some embodiments, receiving the feature matrix of the standardized digital pathology image includes: acquiring feature data from the same patient's intraoperative tumor pathology slide obtained through digital pathology image analysis and molecular detection technology; generating a corresponding feature matrix after standardizing and normalizing the acquired feature data; receiving the feature matrix after standardization; and verifying the integrity of the quantitative features within the feature matrix to ensure that the feature matrix contains three preset types of quantitative features.

[0051] This embodiment is a specific implementation of the feature matrix of the standardized digital pathology image described in step S101. The core is to achieve homology verification, standardization processing, and compliance verification of the feature data, ensuring the consistency and integrity of the input data and avoiding prediction deviations caused by differences in data sources, dimensional differences, or missing features. The specific implementation is as follows: Homologous feature data acquisition is achieved by first connecting computer equipment to the clinical pathology system and molecular detection system to obtain two sets of feature data from the same patient's intraoperative FFPE tumor pathology slide. The first set is effector T cell-related feature data obtained by processing WSI digital slides with artificial intelligence image recognition algorithms, and the second set is immune checkpoint expression quantification data obtained from the same slide through immunohistochemical detection. When acquiring data, the computer equipment simultaneously verifies the consistency of the pathology slide number and patient number of the two sets of data. Only when the two sets of data come from the same slide and the same patient are they determined to be homologous and compliant data and enter subsequent processing to ensure the spatial and sample homology of the feature data.

[0052] Standardization and normalization processing is performed on the acquired compliant feature data using a min-max normalization method. Specifically, the normalized value is calculated as follows: Normalized value = (Original value - Physiological minimum value of the feature) / (Physiological maximum value of the feature - Physiological minimum value of the feature). The physiological maximum and minimum values ​​of the feature are derived from the preset physiological value limits of this type of feature, which are calibrated using a large-sample clinical dataset. Through this processing, all feature data of different dimensions and magnitudes are uniformly mapped to a value range of 0 to 1, eliminating the magnitude differences between different features and ensuring the fairness of subsequent model calculations.

[0053] The feature matrix generation and reception process involves generating a two-dimensional structured feature matrix from standardized and normalized feature data, following a fixed column order of "overall expression score of immune checkpoints, effector T cell density, and effector T cell spatial distribution parameters." Each row of the matrix corresponds to a single sample of data from a patient, and each column corresponds to a core feature. The computer receives this standardized feature matrix and simultaneously triggers an integrity verification process, checking each row of data in the matrix to ensure it fully contains the three core features. If any feature is missing, a data anomaly alert is immediately triggered and feedback is sent to the corresponding operator. Only after all sample data has passed the integrity verification is the feature matrix input into subsequent processing steps, completing the feature matrix reception process.

[0054] In some embodiments, the quantitative features extracted from the patient's intraoperative tumor pathology slides include: identifying and quantifying the effector T cell density and spatial distribution parameters of effector T cells within the cancerous tissue region from a full-view digital slide of the patient's intraoperative tumor pathology slide using digital pathology image analysis technology; quantifying the overall expression score of immune checkpoints from the cancerous tissue region of the same intraoperative tumor pathology slide using molecular detection technology; all extracted quantitative features originate from the same intraoperative tumor pathology sample.

[0055] This embodiment is a specific implementation of the extraction of quantitative features from intraoperative tumor pathology slides described in step S101. The core principle is to simultaneously extract three core quantitative features from the same intraoperative pathology slide using routine clinical detection techniques, without requiring additional patient samples. This aligns with standard clinical diagnostic and treatment procedures, ensuring the accuracy and accessibility of the feature data. The specific implementation method is as follows: Pathological section preprocessing involves selecting FFPE tumor pathological sections routinely preserved after tumor tissue is removed during surgical procedures. The sections are 3-5 μm thick, meeting the standards for routine clinical pathological testing. After routine preprocessing such as dewaxing and antigen retrieval, parallel sections with two adjacent sections are prepared for digital pathological image analysis and molecular detection, respectively, to ensure the homology of the samples.

[0056] Effector T cell-related feature extraction was performed by staining the first parallel slide with CD8 immunohistochemistry and then scanning the stained slide with a standard clinical digital pathology scanner at 20x or 40x magnification to generate a WSI digital pathology image. A trained artificial intelligence image recognition model was used to analyze the WSI image. First, semantic segmentation identified the cancerous tissue region and adjacent stroma region in the image. Then, target detection identified CD8+ effector T cells within the cancerous tissue region. The following parameters were quantified and statistically analyzed: ① the number of effector T cells per unit area within the cancerous tissue region, i.e., effector T cell density; ② the distribution ratio of effector T cells within the cancer nest and adjacent stroma, the median spatial distance between effector T cells and tumor cells, and the effector T cell aggregation score, i.e., effector T cell spatial distribution parameters. This completed the extraction of effector T cell-related features.

[0057] The extraction of immune checkpoint expression scores was achieved by performing PD-L1 immunohistochemical staining on a second adjacent parallel slice. The stained slices were then scanned and quantitatively analyzed using a routine clinical pathological image analysis system. The proportion of PD-L1 positive cells in the entire cancerous tissue area was calculated, i.e., the tumor proportion score (TPS), which served as the overall expression score of immune checkpoints, thus completing the extraction of immune checkpoint-related features.

[0058] Feature data integration combines three core quantitative features extracted from adjacent slices of the same paraffin block into a feature set corresponding to the patient. All features come from the same intraoperative tumor pathology sample, eliminating the need to collect additional patient samples. It is fully compatible with routine clinical pathology testing procedures and can directly interface with the output data of existing clinical pathology systems.

[0059] In some embodiments, the process of eliminating invalid features for each set of parameter sequences within the feature matrix using a divide-and-conquer paradigm and a recursive method includes: splitting each set of parameter sequences within the feature matrix into multiple independent sub-sequence units according to a preset feature dimension; for each sub-sequence unit, recursively traversing all feature data within the unit to eliminate invalid feature data that exceeds a preset reasonable value range or has missing data; after eliminating invalid features from all sub-sequence units, merging the remaining valid feature data to generate a valid feature set for the corresponding parameter sequence.

[0060] This embodiment is a specific implementation of step S102, which describes the process of eliminating invalid features for each set of parameter sequences within the feature matrix using a divide-and-conquer paradigm and a recursive method. The core principle is to reduce computational complexity through divide-and-conquer decomposition and to achieve accurate removal of all invalid features through recursive traversal, thus solving the problem of outliers and missing values ​​interfering with subsequent model calculations in high-dimensional feature data. The specific implementation method is as follows: The parameter sequence splitting and divide-and-conquer process splits the feature matrix received in step S101 into three independent parameter sequences: immune checkpoint expression score sequence, effector T cell density sequence, and effector T cell spatial distribution parameter sequence. For each parameter sequence, a divide-and-conquer split is performed according to a preset feature dimension and a single-batch computation threshold. Specifically, for the single-dimensional immune checkpoint expression score sequence and effector T cell density sequence, it is split into multiple equal-length sub-sequence units in units of 1000 samples. For the multi-dimensional effector T cell spatial distribution parameter sequence, it is split into distribution proportion sub-units, spatial distance sub-units, and clustering degree sub-units according to the feature sub-dimension. Each sub-unit contains only feature data of a single sub-dimension. After splitting, the data volume of each sub-sequence unit does not exceed the preset single-batch computation threshold, thus achieving divide-and-conquer processing and significantly reducing the complexity of single-batch computation.

[0061] Recursive traversal and invalid feature removal: For each sub-sequence unit after splitting, a recursive traversal program is initiated. The specific execution flow is as follows: ① The program first locates the first feature data point within the sub-sequence unit and performs data verification; ② The verification rule is: determine whether the value of the data point is within the normalized reasonable range of 0 to 1, and at the same time determine whether the data point is null or missing; ③ If the data point meets the verification rule, it is determined to be valid data and is retained. The program recursively locates the next data point and repeats the verification process; ④ If the data point does not meet the verification rule, it is determined to be invalid data and is directly removed from the sub-sequence unit. The program recursively locates the next data point and repeats the verification process; ⑤ When the program has traversed all data points within the sub-sequence unit, the recursion terminates, completing the invalid feature removal for that sub-sequence unit.

[0062] The effective feature set is generated by merging the remaining effective feature data in all subsequence units according to the dimensional order and sample order of the original parameter sequence after all subsequence units have completed recursive traversal and invalid feature removal. After the three sets of parameter sequences have completed the above processing, three sets of effective feature sets corresponding to the sample order are obtained, completing the full process of invalid feature removal and providing clean, reliable, and effective data for subsequent feature selection.

[0063] In some embodiments, the step of performing feature selection and model compression processing on the parameter sequences after exclusion processing using a regularization method includes: performing multiple rounds of iterative calculations on the effective feature sets corresponding to the multiple sets of parameter sequences after exclusion processing using a regularization regression method to screen out core features whose correlation with the sensitivity of concurrent chemoradiotherapy meets a preset threshold; based on the core features obtained by screening, reducing the computational dimension of the model to complete feature selection and model compression processing, while avoiding the risk of model overfitting.

[0064] This embodiment is a specific implementation of the feature selection and model compression processing using regularization on the parameter sequence after exclusion processing described in step S102. The core is to use regularized regression to screen out key features strongly correlated with the sensitivity of concurrent chemoradiotherapy, while simultaneously reducing the model's computational dimensionality, avoiding the risk of overfitting, and improving the model's generalization ability and prediction accuracy. The specific implementation is as follows: Dataset preprocessing involves merging the three sets of effective feature sets into the model input dataset and matching them with corresponding clinical label data. The label data represents the clinical outcomes of patients who received concurrent chemoradiotherapy after surgery, including whether there was a recurrence and progression-free survival. The clinical outcomes are binarized and used as the model's prediction labels. The dataset is then randomly divided into a training set and a validation set in a 7:3 ratio. The training set is used for iterative training of the model, and the validation set is used for verifying the model's generalization ability.

[0065] Regularized Regression Model Training and Feature Selection: Lasso regression was used as the regularized regression model. The regularization penalty coefficient of the model was preset, and the optimal penalty coefficient value was determined by ten-fold cross-validation. The effective feature data of the training set was used as the input independent variable, and the clinical outcome label of concurrent chemoradiotherapy was used as the dependent variable. Multiple rounds of iterative regression calculations were performed. After each round of calculation, the model automatically compressed the regression coefficients of features with extremely low correlation with the predicted label to 0. After the iterative calculations converged, all features with regression coefficients of 0 were removed, and features with non-zero regression coefficients were retained as core features whose correlation with the sensitivity of concurrent chemoradiotherapy met the preset threshold, thus completing the feature selection.

[0066] Model compression and generalization validation: Based on the core features obtained through screening, the regression model is reconstructed. The input dimension of the model is reduced from the original high-dimensional features to low-dimensional features containing only the core features, thus completing model compression and significantly reducing computational redundancy. The feature data of the validation set is input into the compressed model to verify the model's predictive accuracy and generalization ability. The predictive performance of the model is evaluated by the area under the ROC curve and the consistency index, ensuring that the area under the ROC curve of the compressed model in the validation set is not less than 0.75, which indicates good predictive accuracy and generalization ability. At the same time, regularization constraints effectively avoid the risk of model overfitting, ensuring that the model maintains stable predictive performance in clinical data from different batches and sources.

[0067] In some embodiments, identifying tumor tissue features that simultaneously exhibit high immune checkpoint expression and high effector T cell infiltration includes: determining, based on the core features after feature selection processing, a high expression threshold for the immune checkpoint expression score and a high infiltration threshold for effector T cell infiltration-related features; comparing the immune checkpoint expression score with the high expression threshold and the effector T cell-related features with the high infiltration threshold for a single sample; when the same sample simultaneously meets the conditions that the immune checkpoint expression score is not lower than the high expression threshold and the effector T cell infiltration-related features are not lower than the high infiltration threshold, the sample is identified as having tumor tissue features that simultaneously exhibit high immune checkpoint expression and high effector T cell infiltration.

[0068] This embodiment is a specific implementation of the method described in step S102 for identifying tumor tissue characteristics that simultaneously exhibit high immune checkpoint expression and high effector T cell infiltration. The core is to accurately identify patient groups that can benefit from immunotherapy using objective thresholds calibrated from clinical data, providing a basis for subsequently removing the influence weight of this characteristic. Specifically, it includes: Feature threshold calibration: Based on the core features obtained through screening, a large-scale clinical dataset of solid tumors was selected. This dataset includes patients' immune checkpoint expression scores, effector T cell infiltration-related feature data, and corresponding immunotherapy clinical outcome data. For the immune checkpoint expression score, the optimal critical value for distinguishing between patients who benefited from and did not benefit from immunotherapy was calculated by plotting ROC curves, and this critical value was set as the threshold for high immune checkpoint expression. For effector T cell density and effector T cell spatial distribution parameters, the same ROC curve analysis method was used to calculate the corresponding optimal critical values, which were set as the thresholds for high effector T cell infiltration. All thresholds were calibrated based on clinical gold standard data and have clear clinical significance.

[0069] Single-sample feature comparison: For sample data of a single patient, the quantitative values ​​of the patient's immune checkpoint expression score, effector T cell density, and effector T cell spatial distribution parameters are extracted from the core features after feature selection processing. First, the immune checkpoint expression score is compared with a preset high expression threshold to determine whether the patient's immune checkpoint expression meets the high expression standard. Then, the effector T cell density and effector T cell spatial distribution parameters are compared with the corresponding high infiltration thresholds to determine whether the patient's effector T cell infiltration meets the high infiltration standard.

[0070] Target Feature Identification: For the same patient sample, if both of the following conditions are met simultaneously: ① Immune checkpoint expression score is greater than or equal to the preset high expression threshold; ② Effector T cell density and effector T cell spatial distribution parameters are both greater than or equal to the corresponding high infiltration threshold, the sample is identified as having tumor tissue characteristics of simultaneous high immune checkpoint expression and high effector T cell infiltration, and the patient is determined to be a potential beneficiary of immunotherapy with low potential sensitivity to concurrent chemoradiotherapy; if neither of the above two conditions is met simultaneously, the sample is determined to not possess the target feature and is a potential suitable candidate for concurrent chemoradiotherapy.

[0071] In some embodiments, the integration of the association information between immune checkpoint expression and effector T cell spatial infiltration includes: matching the quantitative data of immune checkpoint expression and the quantitative data of effector T cell spatial infiltration of the corresponding sample based on the identified tumor tissue characteristics; analyzing the degree of positive correlation between the two types of quantitative data, and simultaneously labeling the combined influence weight of the two types of data on the sensitivity of concurrent chemoradiotherapy, thereby completing the integration of the association information between immune checkpoint expression and effector T cell spatial infiltration.

[0072] This embodiment is a specific implementation of the integration of the association information between immune checkpoint expression and effector T cell spatial infiltration described in step S102. The core is to quantitatively analyze the association between the two types of features, assign weights to their combined impact on the sensitivity of concurrent chemoradiotherapy, and provide accurate weighting basis for subsequent predictive scoring. The specific implementation is as follows: Target sample data matching: Based on the identified samples that simultaneously exhibit high immune checkpoint expression and high effector T cell infiltration, quantitative data on immune checkpoint expression and quantitative data on spatial infiltration of effector T cells are extracted for these samples. The two types of data are matched one-to-one according to the sample number to form an association analysis dataset, ensuring the sample homology of the data.

[0073] Feature correlation analysis: For the correlation analysis dataset, Pearson correlation coefficient analysis was used to calculate the correlation coefficient between the quantitative data of immune checkpoint expression and the quantitative data of effector T cell spatial infiltration. This clarified the degree of positive correlation between the two types of data. The closer the correlation coefficient is to 1, the stronger the positive correlation between the two types of features, the clearer the immunosuppressive state of the tumor microenvironment, and the higher the probability that patients will benefit from immunotherapy. At the same time, scatter plot fitting was used to clarify the linear correlation between the two types of features, providing a basis for subsequent joint weighting.

[0074] Joint Influence Weighting: Using concurrent chemoradiotherapy sensitivity as the dependent variable and quantitative data on immune checkpoint expression and effector T cell spatial infiltration as independent variables, along with clinical covariates such as patient age, gender, tumor stage, and pathological type, a multivariate Cox regression model was constructed. Through iterative calculations of the regression model, the hazard ratio and regression coefficient of the combined effect of immune checkpoint expression and effector T cell spatial infiltration on concurrent chemoradiotherapy sensitivity were calculated. This regression coefficient was then assigned as the joint influence weight of the two data types on concurrent chemoradiotherapy sensitivity. Finally, the association information between immune checkpoint expression and effector T cell spatial infiltration was integrated to obtain the degree of association and joint influence weight of the two features, providing a precise quantitative basis for subsequent weight elimination.

[0075] In some embodiments, the step of removing the influence weights corresponding to the identified tumor tissue features based on the integrated association information includes: obtaining the overall feature weights corresponding to all cancer tissues based on the integrated association information; removing the weight proportions corresponding to the identified tumor tissue features that simultaneously exhibit high immune checkpoint expression and high-efficiency T cell infiltration from the overall feature weights; and retaining the remaining feature weight data that are directly related to the sensitivity of concurrent chemoradiotherapy after the weight removal is completed.

[0076] This embodiment is a specific implementation method for removing the influence weights corresponding to the identified tumor tissue features based on the integrated association information described in step S103. The core is to accurately remove the influence weights of features related to immunotherapy benefit, retaining only the weight data directly related to the sensitivity of concurrent chemoradiotherapy, ensuring the specificity and accuracy of the final prediction score. The specific implementation method is as follows: Overall feature weight acquisition: Based on the completed prediction model and the integrated correlation information, for the sample data of a single patient, the overall feature weight corresponding to the full set of core features of the sample is obtained, that is, the total regression coefficient of all core features on the sensitivity of concurrent chemoradiotherapy. This overall feature weight reflects the comprehensive influence of all cancer tissue features on the sensitivity of concurrent chemoradiotherapy.

[0077] Target feature weight localization and elimination: Based on the identified tumor tissue features of high immune checkpoint expression and high-effect T cell infiltration in the sample, the weight ratio of this feature in the overall feature weight is located according to the labeled joint influence weight; this joint influence weight is completely eliminated from the overall feature weight. After elimination, the overall feature weight no longer includes the influence of features related to immunotherapy benefit, and only the feature weight directly related to the sensitivity of concurrent chemoradiotherapy is retained.

[0078] Valid weight data retention: After weight removal, the remaining feature weight data is verified to ensure that the features corresponding to the remaining weight data are all core features directly related to the clinical benefits of concurrent chemoradiotherapy, eliminating the influence of all tumor tissue features that can be effectively killed by immunotherapy; the remaining feature weight data that has passed verification is used as the core basis for generating the subsequent concurrent chemoradiotherapy sensitivity prediction score, ensuring that the final prediction score only reflects the patient's true potential sensitivity to concurrent chemoradiotherapy, avoiding interference from features related to immunotherapy benefits on the prediction results.

[0079] In some embodiments, generating and outputting a concurrent chemoradiotherapy sensitivity prediction score, which is used to indicate a patient's potential sensitivity to concurrent chemoradiotherapy, includes: generating a corresponding concurrent chemoradiotherapy sensitivity prediction score by performing a weighted operation based on the feature weight data retained after weight removal; comparing the generated prediction score with a preset sensitivity grading threshold to determine the corresponding patient's concurrent chemoradiotherapy sensitivity level; and outputting the prediction score and the corresponding sensitivity level to provide a quantitative basis for the formulation of clinical treatment plans.

[0080] This embodiment is a specific implementation of the generation and output of the concurrent chemoradiotherapy sensitivity prediction score described in step S103. The core is to generate quantitative prediction results that can directly guide clinical decision-making, while simultaneously completing sensitivity grading, providing clear quantitative evidence for the formulation of clinical treatment plans. The specific implementation method is as follows: The predictive score is generated by extracting the core feature quantification value corresponding to the patient based on the retained feature weight data directly related to the sensitivity of concurrent chemoradiotherapy using sample data from a single patient. A weighted summation operation is then performed, specifically: Concurrent chemoradiotherapy sensitivity prediction score = Σ (core feature quantification value × corresponding feature weight). This operation generates a continuous concurrent chemoradiotherapy sensitivity prediction score for the patient. The score ranges from 0 to 100. The higher the score, the higher the patient's potential sensitivity to concurrent chemoradiotherapy, and the more significant the clinical benefit of receiving concurrent chemoradiotherapy after surgery.

[0081] Sensitivity grading determination: Based on survival analysis results from a large-sample clinical dataset, three sensitivity grading thresholds were preset: a low sensitivity threshold of 30 points and a high sensitivity threshold of 70 points. The generated predicted score was compared with the grading thresholds to complete the sensitivity level classification: ① Predicted score < 30 points: classified as low sensitivity level, indicating that the patient has extremely low potential sensitivity to concurrent chemoradiotherapy, and does not need to receive concurrent chemoradiotherapy after surgery. Immunotherapy or other treatment options are recommended as the first choice; ② 30 points ≤ predicted score < 70 points: classified as moderate sensitivity level, indicating that the patient has a moderate degree of sensitivity to concurrent chemoradiotherapy. Whether to use concurrent chemoradiotherapy can be comprehensively judged based on the patient's physical condition and clinical characteristics; ③ Predicted score ≥ 70 points: classified as high sensitivity level, indicating that the patient has extremely high potential sensitivity to concurrent chemoradiotherapy, and the clinical benefit of receiving concurrent chemoradiotherapy after surgery is significant. Postoperative concurrent chemoradiotherapy is recommended.

[0082] The results output and clinical adaptation are achieved through computer equipment, which ultimately outputs the patient's concurrent chemoradiotherapy sensitivity prediction score and corresponding sensitivity level, along with corresponding clinical reference suggestions. The output results can be directly integrated into clinical pathology analysis software, third-party testing platforms, and hospital clinical decision support systems via API interfaces. This provides objective and quantitative biological evidence for clinicians to develop personalized postoperative treatment plans for patients, effectively avoiding the problems of overtreatment or undertreatment.

[0083] Please see Figure 3 As shown, Figure 3 This is a schematic diagram of the structure of a concurrent chemoradiotherapy sensitivity prediction system 200 for immune cell infiltration analysis provided in this application embodiment. The concurrent chemoradiotherapy sensitivity prediction system 200 for immune cell infiltration analysis is used to perform the steps of the concurrent chemoradiotherapy sensitivity prediction methods for immune cell infiltration analysis shown in the above embodiments. The concurrent chemoradiotherapy sensitivity prediction system 200 for immune cell infiltration analysis can be a single server or a server cluster, or it can be a terminal, such as a handheld terminal, laptop computer, wearable device, or robot.

[0084] like Figure 3 As shown, the concurrent chemoradiotherapy sensitivity prediction system 200 for immune cell infiltration analysis includes: The matrix receiving unit 201 is used to receive the feature matrix of a standardized digital pathology image. The feature matrix contains quantitative features extracted from the patient's intraoperative tumor pathology section. The quantitative features include the overall expression score of immune checkpoints, effector T cell density, and effector T cell spatial distribution parameters. The exclusion processing unit 202 is used to exclude invalid features for each set of parameter sequences in the feature matrix using the divide-and-conquer paradigm and recursive methods respectively; for the parameter sequences after exclusion processing, the regularization method is used for feature selection and model compression to identify tumor tissue features that simultaneously have high immune checkpoint expression and high effector T cell infiltration, which is used to integrate the correlation information between immune checkpoint expression and effector T cell spatial infiltration. The scoring generation unit 203 is used to generate and output a concurrent chemoradiotherapy sensitivity prediction score based on the integrated association information, by removing the influence weights corresponding to the identified tumor tissue characteristics. The concurrent chemoradiotherapy sensitivity prediction score is used to indicate the patient's potential sensitivity to concurrent chemoradiotherapy.

[0085] In some embodiments, receiving the feature matrix of the standardized digital pathology image includes: acquiring feature data from the same patient's intraoperative tumor pathology slide obtained through digital pathology image analysis and molecular detection technology; generating a corresponding feature matrix after standardizing and normalizing the acquired feature data; receiving the feature matrix after standardization; and verifying the integrity of the quantitative features within the feature matrix to ensure that the feature matrix contains three preset types of quantitative features.

[0086] In some embodiments, the quantitative features extracted from the patient's intraoperative tumor pathology slides include: identifying and quantifying the effector T cell density and spatial distribution parameters of effector T cells within the cancerous tissue region from a full-view digital slide of the patient's intraoperative tumor pathology slide using digital pathology image analysis technology; quantifying the overall expression score of immune checkpoints from the cancerous tissue region of the same intraoperative tumor pathology slide using molecular detection technology; all extracted quantitative features originate from the same intraoperative tumor pathology sample.

[0087] In some embodiments, the process of eliminating invalid features for each set of parameter sequences within the feature matrix using a divide-and-conquer paradigm and a recursive method includes: splitting each set of parameter sequences within the feature matrix into multiple independent sub-sequence units according to a preset feature dimension; for each sub-sequence unit, recursively traversing all feature data within the unit to eliminate invalid feature data that exceeds a preset reasonable value range or has missing data; after eliminating invalid features from all sub-sequence units, merging the remaining valid feature data to generate a valid feature set for the corresponding parameter sequence.

[0088] In some embodiments, the step of performing feature selection and model compression processing on the parameter sequences after exclusion processing using a regularization method includes: performing multiple rounds of iterative calculations on the effective feature sets corresponding to the multiple sets of parameter sequences after exclusion processing using a regularization regression method to screen out core features whose correlation with the sensitivity of concurrent chemoradiotherapy meets a preset threshold; based on the core features obtained by screening, reducing the computational dimension of the model to complete feature selection and model compression processing, while avoiding the risk of model overfitting.

[0089] In some embodiments, identifying tumor tissue features that simultaneously exhibit high immune checkpoint expression and high effector T cell infiltration includes: determining, based on the core features after feature selection processing, a high expression threshold for the immune checkpoint expression score and a high infiltration threshold for effector T cell infiltration-related features; comparing the immune checkpoint expression score with the high expression threshold and the effector T cell-related features with the high infiltration threshold for a single sample; when the same sample simultaneously meets the conditions that the immune checkpoint expression score is not lower than the high expression threshold and the effector T cell infiltration-related features are not lower than the high infiltration threshold, the sample is identified as having tumor tissue features that simultaneously exhibit high immune checkpoint expression and high effector T cell infiltration.

[0090] In some embodiments, the integration of the association information between immune checkpoint expression and effector T cell spatial infiltration includes: matching the quantitative data of immune checkpoint expression and the quantitative data of effector T cell spatial infiltration of the corresponding sample based on the identified tumor tissue characteristics; analyzing the degree of positive correlation between the two types of quantitative data, and simultaneously labeling the combined influence weight of the two types of data on the sensitivity of concurrent chemoradiotherapy, thereby completing the integration of the association information between immune checkpoint expression and effector T cell spatial infiltration.

[0091] In some embodiments, the step of removing the influence weights corresponding to the identified tumor tissue features based on the integrated association information includes: obtaining the overall feature weights corresponding to all cancer tissues based on the integrated association information; removing the weight proportions corresponding to the identified tumor tissue features that simultaneously exhibit high immune checkpoint expression and high-efficiency T cell infiltration from the overall feature weights; and retaining the remaining feature weight data that are directly related to the sensitivity of concurrent chemoradiotherapy after the weight removal is completed.

[0092] In some embodiments, generating and outputting a concurrent chemoradiotherapy sensitivity prediction score, which is used to indicate a patient's potential sensitivity to concurrent chemoradiotherapy, includes: generating a corresponding concurrent chemoradiotherapy sensitivity prediction score by performing a weighted operation based on the feature weight data retained after weight removal; comparing the generated prediction score with a preset sensitivity grading threshold to determine the corresponding patient's concurrent chemoradiotherapy sensitivity level; and outputting the prediction score and the corresponding sensitivity level to provide a quantitative basis for the formulation of clinical treatment plans.

[0093] It should be noted that those skilled in the art will understand that, for the sake of convenience and brevity, the specific working process of the concurrent chemoradiotherapy sensitivity prediction system and its modules for immune cell infiltration analysis described above can be found in the corresponding contents of the various embodiments of the concurrent chemoradiotherapy sensitivity prediction method for immune cell infiltration analysis, and will not be repeated here.

[0094] The aforementioned method for predicting the sensitivity of concurrent chemoradiotherapy for immune cell infiltration analysis can be implemented as a computer program, which can be used in, for example... Figure 3 It runs on the device shown.

[0095] Please see Figure 4 , Figure 4 This is a schematic block diagram of the structure of a computer device provided in an embodiment of this application. The computer device includes a processor, a memory, and a network interface connected via a device bus, wherein the memory may include a storage medium and internal memory.

[0096] The storage medium may store operating devices and computer programs. The computer program includes program instructions that, when executed, cause the processor to perform any method for predicting the sensitivity of concurrent chemoradiotherapy for analysis of immune cell infiltration.

[0097] The processor provides computing and control capabilities, supporting the operation of the entire computer device.

[0098] Internal memory provides an environment for the execution of computer programs in non-volatile storage media. When executed by a processor, the computer program enables the processor to perform any method for predicting the sensitivity of concurrent chemoradiotherapy for the analysis of immune cell infiltration.

[0099] This network interface is used for network communication, such as sending assigned tasks. Those skilled in the art will understand that... Figure 4 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the terminal to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0100] It should be understood that the processor can be a Central Processing Unit (CPU), but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among these, a general-purpose processor can be a microprocessor or any conventional processor.

[0101] In one embodiment, the processor is configured to run a computer program stored in memory to perform the following steps: The feature matrix of the received standardized digital pathology image contains quantitative features extracted from the patient’s intraoperative tumor pathology slides, including the overall expression score of immune checkpoints, effector T cell density, and effector T cell spatial distribution parameters. For each set of parameter sequences in the feature matrix, invalid features are excluded using the divide-and-conquer paradigm and recursive methods respectively. For the parameter sequences after exclusion, regularization is used for feature selection and model compression to identify tumor tissue features that simultaneously have high immune checkpoint expression and high effector T cell infiltration, which are used to integrate the correlation information between immune checkpoint expression and effector T cell spatial infiltration. Based on the integrated association information, the influence weights corresponding to the identified tumor tissue characteristics are removed, and a concurrent chemoradiotherapy sensitivity prediction score is generated and output. The concurrent chemoradiotherapy sensitivity prediction score is used to indicate the patient's potential sensitivity to concurrent chemoradiotherapy.

[0102] In some embodiments, receiving the feature matrix of the standardized digital pathology image includes: acquiring feature data from the same patient's intraoperative tumor pathology slide obtained through digital pathology image analysis and molecular detection technology; generating a corresponding feature matrix after standardizing and normalizing the acquired feature data; receiving the feature matrix after standardization; and verifying the integrity of the quantitative features within the feature matrix to ensure that the feature matrix contains three preset types of quantitative features.

[0103] In some embodiments, the quantitative features extracted from the patient's intraoperative tumor pathology slides include: identifying and quantifying the effector T cell density and spatial distribution parameters of effector T cells within the cancerous tissue region from a full-view digital slide of the patient's intraoperative tumor pathology slide using digital pathology image analysis technology; quantifying the overall expression score of immune checkpoints from the cancerous tissue region of the same intraoperative tumor pathology slide using molecular detection technology; all extracted quantitative features originate from the same intraoperative tumor pathology sample.

[0104] In some embodiments, the process of eliminating invalid features for each set of parameter sequences within the feature matrix using a divide-and-conquer paradigm and a recursive method includes: splitting each set of parameter sequences within the feature matrix into multiple independent sub-sequence units according to a preset feature dimension; for each sub-sequence unit, recursively traversing all feature data within the unit to eliminate invalid feature data that exceeds a preset reasonable value range or has missing data; after eliminating invalid features from all sub-sequence units, merging the remaining valid feature data to generate a valid feature set for the corresponding parameter sequence.

[0105] In some embodiments, the step of performing feature selection and model compression processing on the parameter sequences after exclusion processing using a regularization method includes: performing multiple rounds of iterative calculations on the effective feature sets corresponding to the multiple sets of parameter sequences after exclusion processing using a regularization regression method to screen out core features whose correlation with the sensitivity of concurrent chemoradiotherapy meets a preset threshold; based on the core features obtained by screening, reducing the computational dimension of the model to complete feature selection and model compression processing, while avoiding the risk of model overfitting.

[0106] In some embodiments, identifying tumor tissue features that simultaneously exhibit high immune checkpoint expression and high effector T cell infiltration includes: determining, based on the core features after feature selection processing, a high expression threshold for the immune checkpoint expression score and a high infiltration threshold for effector T cell infiltration-related features; comparing the immune checkpoint expression score with the high expression threshold and the effector T cell-related features with the high infiltration threshold for a single sample; when the same sample simultaneously meets the conditions that the immune checkpoint expression score is not lower than the high expression threshold and the effector T cell infiltration-related features are not lower than the high infiltration threshold, the sample is identified as having tumor tissue features that simultaneously exhibit high immune checkpoint expression and high effector T cell infiltration.

[0107] In some embodiments, the integration of the association information between immune checkpoint expression and effector T cell spatial infiltration includes: matching the quantitative data of immune checkpoint expression and the quantitative data of effector T cell spatial infiltration of the corresponding sample based on the identified tumor tissue characteristics; analyzing the degree of positive correlation between the two types of quantitative data, and simultaneously labeling the combined influence weight of the two types of data on the sensitivity of concurrent chemoradiotherapy, thereby completing the integration of the association information between immune checkpoint expression and effector T cell spatial infiltration.

[0108] In some embodiments, the step of removing the influence weights corresponding to the identified tumor tissue features based on the integrated association information includes: obtaining the overall feature weights corresponding to all cancer tissues based on the integrated association information; removing the weight proportions corresponding to the identified tumor tissue features that simultaneously exhibit high immune checkpoint expression and high-efficiency T cell infiltration from the overall feature weights; and retaining the remaining feature weight data that are directly related to the sensitivity of concurrent chemoradiotherapy after the weight removal is completed.

[0109] In some embodiments, generating and outputting a concurrent chemoradiotherapy sensitivity prediction score, which is used to indicate a patient's potential sensitivity to concurrent chemoradiotherapy, includes: generating a corresponding concurrent chemoradiotherapy sensitivity prediction score by performing a weighted operation based on the feature weight data retained after weight removal; comparing the generated prediction score with a preset sensitivity grading threshold to determine the corresponding patient's concurrent chemoradiotherapy sensitivity level; and outputting the prediction score and the corresponding sensitivity level to provide a quantitative basis for the formulation of clinical treatment plans.

[0110] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the steps of the concurrent chemoradiotherapy sensitivity prediction method for immune cell infiltration analysis as provided in any embodiment of this application.

[0111] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiments, such as the hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, SmartMedia Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the computer device.

[0112] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for predicting the sensitivity of concurrent chemoradiotherapy for immune cell infiltration analysis, characterized in that, include: The feature matrix of the received standardized digital pathology image contains quantitative features extracted from the patient’s intraoperative tumor pathology slides, including the overall expression score of immune checkpoints, effector T cell density, and effector T cell spatial distribution parameters. For each set of parameter sequences in the feature matrix, invalid features are excluded using the divide-and-conquer paradigm and recursive methods respectively. For the parameter sequences after exclusion, regularization is used for feature selection and model compression to identify tumor tissue features that simultaneously have high immune checkpoint expression and high effector T cell infiltration, which are used to integrate the correlation information between immune checkpoint expression and effector T cell spatial infiltration. Based on the integrated association information, the influence weights corresponding to the identified tumor tissue characteristics are removed, and a concurrent chemoradiotherapy sensitivity prediction score is generated and output. The concurrent chemoradiotherapy sensitivity prediction score is used to indicate the patient's potential sensitivity to concurrent chemoradiotherapy.

2. The method according to claim 1, characterized in that, The feature matrix of the received standardized digital pathology image includes: Acquire characteristic data from the same patient's intraoperative tumor pathology slides, obtained through digital pathology image analysis and molecular detection technology. After standardizing and normalizing the acquired feature data, a corresponding feature matrix is ​​generated, and the standardized feature matrix is ​​received. Verify the integrity of the quantitative features within the feature matrix to ensure that the feature matrix contains the three preset types of quantitative features.

3. The method according to claim 1, characterized in that, The quantitative features extracted from the patient's intraoperative tumor pathology sections include: Using digital pathological image analysis technology, the density of effector T cells in the cancerous tissue area and the spatial distribution parameters of effector T cells in the cancerous tissue area are identified and quantified from the full-view digital slices of the patient's intraoperative tumor pathology section. Using molecular detection technology, the overall expression score of immune checkpoints was quantitatively extracted from the cancerous tissue region of the same intraoperative tumor pathology section; all extracted quantitative features were derived from the same intraoperative tumor pathology sample.

4. The method according to claim 1, characterized in that, For each set of parameter sequences within the feature matrix, invalid features are eliminated using a divide-and-conquer paradigm and a recursive method, including: Each parameter sequence within the feature matrix is ​​split into multiple independent sub-sequence units according to a preset feature dimension; For each subsequence unit, all feature data within the unit are traversed recursively, and invalid feature data that exceeds the preset reasonable value range or has missing data is removed; After removing invalid features from all subsequence units, the remaining valid feature data are merged to generate a valid feature set for the corresponding parameter sequence.

5. The method according to claim 1, characterized in that, The process of regularizing the parameter sequence after exclusion for feature selection and model compression includes: For the effective feature sets corresponding to multiple parameter sequences after exclusion processing, a regularized regression method is used to perform multiple rounds of iterative calculations to screen out the core features whose correlation with the sensitivity of concurrent chemoradiotherapy meets the preset threshold. Based on the core features obtained through screening, the computational dimension of the model is reduced, feature selection and model compression are completed, and the risk of model overfitting is avoided.

6. The method according to claim 1, characterized in that, The identification of tumor tissue features exhibiting both high immune checkpoint expression and high effector T cell infiltration includes: Based on the core features after feature selection processing, the high expression thresholds for immune checkpoint expression scores and the high infiltration thresholds for effector T cell infiltration-related features were determined respectively. The immune checkpoint expression scores and high expression thresholds, as well as effector T cell-related features and high infiltration thresholds, were compared separately for individual samples. When the same sample simultaneously meets the criteria of immune checkpoint expression score not lower than the high expression threshold and effector T cell infiltration-related characteristics not lower than the high infiltration threshold, the sample is identified as having tumor tissue characteristics of both high immune checkpoint expression and high effector T cell infiltration.

7. The method according to claim 6, characterized in that, The integrated association information between immune checkpoint expression and effector T cell spatial infiltration includes: Based on the tumor tissue characteristics identified, the quantitative data of immune checkpoint expression and the quantitative data of spatial infiltration of effector T cells are matched with the corresponding samples. The degree of positive correlation between the two types of quantitative data was analyzed, and the combined influence weight of the two types of data on the sensitivity of concurrent chemoradiotherapy was determined. The association information between immune checkpoint expression and effector T cell spatial infiltration was integrated.

8. The method according to claim 1, characterized in that, The step of removing the influence weights corresponding to the identified tumor tissue features based on the integrated association information includes: Based on the integrated association information, the overall feature weights corresponding to all cancer tissues are obtained; For tumor tissue features that simultaneously exhibit high immune checkpoint expression and high-efficiency T cell infiltration, their corresponding weight percentage in the overall feature weight is removed. After weight removal, the remaining feature weight data that are directly related to the sensitivity of concurrent chemoradiotherapy are retained.

9. The method according to claim 1, characterized in that, The generation and output of a concurrent chemoradiotherapy sensitivity prediction score, which is used to indicate a patient's potential sensitivity to concurrent chemoradiotherapy, includes: Based on the feature weight data retained after weight removal, a weighted calculation is performed to generate the corresponding concurrent chemoradiotherapy sensitivity prediction score. The generated predicted score is compared with the preset sensitivity grading threshold to determine the sensitivity level of the corresponding patient's concurrent chemoradiotherapy. The system outputs predictive scores and corresponding sensitivity levels, providing a quantitative basis for the development of clinical treatment plans.

10. A system for predicting the sensitivity of concurrent chemoradiotherapy for analyzing immune cell infiltration, characterized in that, The method applied to any one of claims 1-9 includes: A matrix receiving unit is used to receive a feature matrix of a standardized digital pathology image. The feature matrix contains quantitative features extracted from intraoperative tumor pathology sections of the patient. The quantitative features include the overall expression score of immune checkpoints, effector T cell density, and effector T cell spatial distribution parameters. The exclusion processing unit is used to exclude invalid features from each set of parameter sequences in the feature matrix using both the divide-and-conquer paradigm and a recursive method. For the excluded parameter sequences, a regularization method is used for feature selection and model compression to identify tumor tissue features exhibiting both high immune checkpoint expression and high effector T cell infiltration, which is then used to integrate the correlation information between immune checkpoint expression and effector T cell spatial infiltration. The scoring generation unit, based on the integrated correlation information, removes the influence weights corresponding to the identified tumor tissue features, generates and outputs a concurrent chemoradiotherapy sensitivity prediction score, which indicates the patient's potential sensitivity to concurrent chemoradiotherapy.

Images