A Dynamic Optimization and Identification Analysis System for Early Tumor Screening Targeted Panels

Abstract

This invention relates to the field of early tumor screening analysis technology, and more particularly to a dynamic optimization and identification analysis system for early tumor screening targeted panels. The system is used for: weighted fusion processing of tumor characterization signal mapping based on multimodal tumor-related data to generate weighted tumor characterization signal mapping data; analysis and optimization of early tumor screening targeted panel configuration data based on the weighted tumor characterization signal mapping data; panel constraint identification of tumor characterization signal feature mapping data based on the optimized early tumor screening targeted panel configuration data and the weighted tumor characterization signal mapping data; and establishment of dynamic early tumor screening targeted panel data based on panel constraint identification of tumor characterization signal feature mapping data and optimized early tumor screening targeted panel configuration data. This invention achieves dynamic construction of early tumor screening targeted panels and improves the accuracy and adaptability of targeted panels.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of tumor early screening analysis, and particularly relates to a tumor early screening targeted panel dynamic optimization and identification analysis system. BACKGROUND

[0002] Tumor early screening is a key means to improve tumor cure rate and reduce mortality. The precise detection technology based on targeted panel has become a research hotspot and application mainstream in the field of tumor early screening, with the advantages of high specificity, high detection efficiency and controllable cost. The targeted panel can realize early warning, type discrimination and risk assessment of tumors by screening tumor-specific targets and combining multi-modal tumor-related data, thereby providing an important reference for clinical diagnosis and treatment. However, the existing tumor early screening targeted panel related analysis technology does not fully consider the signal stability difference of different modal data and the contribution difference to tumor representation, which cannot provide reliable support for subsequent target analysis, ignores the precise analysis of target stable domain, does not effectively screen and optimize the tumor representation steady state interval samples, and does not mine the core features of candidate targets through clustering analysis, which easily leads to candidate target screening deviation and affects the rationality of subsequent panel configuration. Moreover, the configuration design of the existing tumor early screening targeted panel is mostly in a fixed mode, lacking a dynamic optimization mechanism, which can only detect specific tumor types or fixed target combinations, and cannot dynamically adjust according to the changes of multi-modal data, the differences of tumor types and the needs of detection scenarios, so the adaptability is poor. SUMMARY

[0003] Therefore, the present application provides a tumor early screening targeted panel dynamic optimization and identification analysis system to solve at least one of the above technical problems.

[0004] To achieve the above-mentioned purpose, a tumor early screening targeted panel dynamic optimization and identification analysis system comprises the following modules: A tumor representation signal analysis module is configured to acquire multi-modal tumor-related data pre-stored in a medical database, perform weighted fusion processing of tumor representation signal mapping based on the multi-modal tumor-related data, and generate weighted tumor representation signal mapping data. A tumor candidate target stable domain analysis module is configured to perform tumor candidate target stable domain analysis based on the weighted tumor representation signal mapping data, and generate tumor candidate target stable domain data. A tumor early screening targeted panel configuration analysis module is configured to perform tumor early screening targeted panel configuration analysis and optimization processing on the tumor candidate target stable domain data, and generate optimized tumor early screening targeted panel configuration data. Panel-constrained identification of tumor signature signal analysis module, configured to perform Panel-constrained identification of tumor signature signal feature analysis on the weighted tumor signature signal mapping data based on the optimized tumor early screening targeted Panel configuration data, to generate Panel-constrained identification of tumor signature signal feature mapping data; Dynamic tumor early screening targeted Panel establishment module, configured to perform dynamic iteration of tumor early screening targeted Panel establishment on the optimized tumor early screening targeted Panel configuration data based on the Panel-constrained identification of tumor signature signal feature mapping data, to generate dynamic tumor early screening targeted Panel data.

[0005] The application has the advantages that the tumor characterization signal analysis module can accurately obtain multi-dimensional, comprehensive and multi-modal tumor-related data such as tumor types, cfDNA mutation points, methylation region distribution and the like pre-stored in a medical database, provides a comprehensive and reliable data basis for subsequent tumor characterization signal analysis, and effectively avoids analysis deviation caused by one-sided single-modal data information. The modal heterogeneity relationship analysis and multi-modal signal analysis of the multi-modal tumor-related data are performed, reasonable modal heterogeneity mapping rules are designed, and logical space mapping processing is completed, which solves the technical pain points of heterogeneity of different modal data and difficulty in direct fusion analysis, realizes the transformation of multi-modal data into a unified logical space, and lays a foundation for subsequent signal fusion. Through signal stability analysis and contribution discriminant analysis of the multi-modal tumor-related signal mapping data, the stability difference of each modal signal and the contribution difference to tumor characterization can be accurately captured, and then the tumor-related signal attribute characteristics are analyzed through the steady-state and discriminant joint driving mode, and a scientific and reasonable weight parameter is designed, which solves the limitations of simple fusion of multi-modal data and lack of targeted weight design. The accuracy, reliability and pertinence of the generated weighted tumor characterization signal mapping data can be significantly improved through the weight parameter for weighted fusion of the characterization signal, effectively filtering invalid signal interference and highlighting the core tumor characterization information. The tumor candidate target stable domain analysis module can accurately decompose different scale fluctuation characteristics of the tumor characterization signal through multi-scale signal fluctuation decomposition processing, and clearly distinguish effective tumor characterization signals from noise interference, providing accurate data support for subsequent signal amplitude and noise ratio analysis. Through analysis of the amplitude-noise ratio relationship of the tumor characterization signal, tumor characterization steady-state interval samples can be scientifically selected, and samples with large signal fluctuations and low reliability can be effectively removed. At the same time, through analysis of the signal potential redundancy relationship of the steady-state interval samples, redundant samples are removed by using mutual information optimization processing, further improving the purity and effectiveness of the steady-state interval samples and avoiding interference caused by redundant samples and low-quality samples in subsequent analysis. Through specificity analysis and clustering processing of the optimized steady-state interval samples, the core features of tumor candidate targets in different samples can be accurately mined, the stable domain range of the tumor candidate target is determined, accurate tumor candidate target stable domain data is generated, and the problems of ignoring target stable domain analysis and large candidate target screening deviation in the prior art are effectively solved. The tumor early screening target panel configuration analysis module realizes accurate division of tumor candidate target stable domain combination units by performing feature analysis, adjacency relationship analysis and coexistence feature analysis on the target stable domain data, effectively mines the internal correlation between different target stable domains, and avoids the problem of unreasonable panel configuration caused by single target or disordered target combination.In combination with tumor type data, the tumor type characteristics are analyzed, and the multi-target early screening target panel efficiency decision is designed accordingly, which can fully adapt to the detection needs of different tumor types, breaking the limitations of the lack of tumor type targeting of existing panel configurations. Using this efficiency decision, the configuration of the target stable domain combination unit can be analyzed to generate initial panel configuration data that meets the clinical detection needs and has the optimal efficiency. At the same time, through risk scenario simulation, the initial panel configuration is optimized, which can predict the potential risks of the panel configuration in different detection scenarios in advance, further improving the stability and applicability of the panel configuration, and generating optimized tumor early screening target panel configuration data, which provides a scientific and reliable configuration basis for subsequent panel constraint identification and dynamic panel establishment. The beneficial effects of the tumor characterization signal analysis module in panel constraint identification are that by accurately analyzing the measurable feature boundary of the panel configuration, the effective range and judgment standard of the panel detection are determined, effectively avoiding the problem of ambiguous tumor characterization signal recognition range and large deviation under unconstrained conditions. By combining the measurable feature boundary data with the weighted tumor characterization signal mapping data, the tumor characterization signal feature is analyzed first, and then the signal response difference of the panel combination is analyzed in depth, which can accurately capture the response law of different panel combinations to tumor characterization signals, and distinguish between effective response signals and invalid interference signals. Through equivalent correction processing, the response deviation between different panel combinations is eliminated. The generated panel constraint identification tumor characterization signal feature mapping data can accurately match the detection needs of the optimized panel configuration, clearly present the correspondence between the tumor characterization signal and the panel configuration, and effectively solve the technical problems of inaccurate tumor characterization signal recognition and low adaptation degree to the panel configuration. The dynamic tumor early screening target panel establishment module integrates the tumor early screening target panel configuration data and the panel constraint identification tumor characterization signal feature mapping data, accurately analyzes the discrimination path of the tumor early screening target panel, and clearly determines the association logic between the panel configuration and the tumor characterization signal recognition, providing a clear direction for the dynamic optimization of the panel. By obtaining the tumor early screening target panel prior evaluation data and combining the multi-target panel parallel discrimination evaluation method, the detection efficiency of different panel configurations can be evaluated comprehensively and objectively, avoiding the evaluation deviation caused by a single evaluation method, and accurately identifying the deficiencies and optimization space in the panel configuration.Simultaneously, a reverse dynamic iteration mechanism continuously adjusts the optimized panel configuration, overcoming the limitation of fixed panel configurations in existing technologies. This allows for dynamic optimization of the panel configuration based on changes in multimodal data, differences in tumor types, the needs of different testing scenarios, and prior assessment results. This ensures that the generated dynamic tumor early screening targeted panel data continuously adapts to clinical testing needs, significantly improving the panel's adaptability, stability, and discrimination accuracy. This guarantees that the established dynamic tumor early screening targeted panel can accurately address the needs of different individuals, different tumor stages, and different testing scenarios.

[0006] Therefore, the tumor early screening targeted panel dynamic optimization and identification analysis system of this invention fully considers the differences in signal stability and contribution to tumor characterization of tumor-related data of different modalities. Through a combination of steady-state and discriminative approaches, it designs scientifically reasonable weight parameters to perform weighted fusion processing on multimodal data, effectively improving the accuracy and reliability of tumor characterization signal mapping data and providing solid data support for subsequent tumor candidate target analysis. By performing multi-scale fluctuation decomposition, steady-state interval sample screening optimization, and cluster analysis on the weighted tumor characterization signals, it achieves accurate analysis of the stable domain of tumor candidate targets, effectively avoiding candidate target screening bias and ensuring the rationality and scientific nature of subsequent panel configurations. Furthermore, it combines tumor type characteristics to design... This system employs multi-objective performance decision-making and optimizes panel configuration through risk scenario simulation, overcoming the limitations of fixed panel configurations. By analyzing measurable feature boundaries of panel configurations and correcting signal response differences, it significantly improves the accuracy of tumor characterization signal feature recognition, enabling precise matching of early tumor characteristics and fully leveraging the clinical value of multimodal tumor-related data. Utilizing prior assessment data for parallel discrimination evaluation of multiple target panels and continuously optimizing panel configurations through a reverse dynamic iteration mechanism, it achieves dynamic adjustment of target panels, significantly improving the system's adaptability to different tumor types, detection scenarios, and individuals. This effectively addresses the issues of insufficient stability and limited discrimination accuracy in existing panels. Furthermore, it effectively solves core problems in existing technologies such as unscientific data fusion, inaccurate target analysis, fixed configurations without dynamic optimization, and insufficient recognition accuracy, significantly improving the accuracy, stability, and applicability of early tumor screening and providing efficient and reliable technical support for precise early tumor screening. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the module flow of a tumor early screening targeted panel dynamic optimization and identification analysis system according to the present invention; The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0008] The technical method of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0009] Furthermore, the accompanying drawings are merely illustrative of the invention and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. Functional entities may be implemented in software, or in one or more hardware modules or integrated circuits, or in different network and / or processor methods and / or microcontroller methods. The term "and / or" as used herein includes any and all combinations of one or more of the associated items listed.

[0010] To achieve the above objectives, please refer to Figure 1 This invention provides a dynamic optimization and identification analysis system for tumor early screening targeting panels. In the embodiments of this invention, please refer to... Figure 1 The diagram shown is a flowchart of a dynamic optimization and identification analysis system for early tumor screening targeting panels according to the present invention. The system includes the following modules: S1: Tumor characterization signal analysis module, used to acquire multimodal tumor-related data pre-stored in medical databases; and to perform weighted fusion processing of tumor characterization signal mapping based on the multimodal tumor-related data to generate weighted tumor characterization signal mapping data; In this embodiment of the invention, multi-dimensional and multi-type tumor-related data are extracted from clinically validated medical databases, covering tumor pathology data, gene-level data, peripheral blood test data, and tumor structure data. All extracted data undergoes integrity verification, removing data lacking core information to ensure analyzability and reliability, providing a foundation for subsequent signal processing. Subsequently, heterogeneous relationship analysis is performed on the extracted multimodal tumor-related data to clarify the correlation strength and intrinsic correlation patterns between different modalities, distinguishing between core, secondary, and weakly correlated modalities. Based on the correlation analysis results, corresponding heterogeneous mapping rules are designed to clarify the mapping region, coordinate allocation principles, and parameter processing standards for different modalities in a unified logical space, achieving the transformation of multimodal heterogeneous data into a unified logical space. Next, signal analysis is performed on each modality, extracting the core signal feature parameters corresponding to each data type, eliminating data redundancy and irrelevant interference information, and normalizing the analyzed multimodal tumor-related signals to eliminate processing biases caused by differences in the magnitude of signal parameters across different modalities. Subsequently, considering the stability of each modality signal and its discriminative contribution to tumor characterization, a reasonable signal attribute weighting scheme was designed. Core correlated modal signals were assigned higher weights, while weakly correlated modal signals were assigned lower weights, ensuring that the weighting matched the tumor characterization value of the signals. Finally, a weighted fusion method was used to integrate the normalized multimodal tumor-related signals. Signals closely correlated within the logical space were fused and enhanced, invalid signals with excessively low intensity were removed, and effective signals with tumor characterization value were retained. The core features, logical space coordinates, and corresponding modal identifiers of all effective signals were integrated to generate weighted tumor characterization signal mapping data.

[0011] S2: Tumor candidate target stability domain analysis module, used to perform tumor candidate target stability domain analysis based on weighted tumor characterization signal mapping data, and generate tumor candidate target stability domain data; In this embodiment of the invention, weighted tumor characterization signal mapping data is used as the core input. First, multi-scale fluctuation decomposition is performed on the tumor characterization signals. This decomposition distinguishes between effective feature fluctuations and noise interference fluctuations, clarifying the fluctuation patterns of tumor characterization signals at different scales, removing noise interference to the effective signals, and improving signal purity and reliability. Subsequently, amplitude and noise ratio analysis is performed on the decomposed effective signals to clarify the amplitude differences between effective and noise signals. Reasonable signal screening criteria are established to select steady-state signal samples with high signal stability and low noise interference, while eliminating non-steady-state samples with large signal fluctuations and severe noise interference, forming a set of steady-state interval samples for tumor characterization. Then, redundancy optimization is performed on the steady-state interval samples, analyzing the signal feature redundancy relationships between samples, identifying and eliminating highly similar redundant samples, while retaining the core feature parameters and clinical correlation information of the samples, improving the purity and effectiveness of the sample set. Next, specificity analysis is performed on the optimized steady-state interval samples to clarify the specific signal features corresponding to different tumor types, distinguishing the signal differences between different tumor types, ensuring that each sample can be associated with its tumor type through its specific signal features, providing a basis for subsequent sample clustering. Subsequently, clustering was used to classify specific samples, grouping samples with similar signal features and corresponding tumor types into one category, forming sample cluster sets corresponding to different tumor types. The core signal features and logical space distribution range of each cluster set were then clarified. Finally, stability domain feature analysis, adjacency relationship analysis, and coexistence feature analysis were performed on each cluster set to clarify the boundary features, core target information, and correlation patterns between adjacent stability domains. Based on the analysis results, tumor candidate target stability domain combination units were divided, and the feature parameters, correlation relationships, and combination unit information of all stability domains were integrated to generate tumor candidate target stability domain data.

[0012] S3: Tumor early screening targeted panel configuration analysis module, used to perform tumor early screening targeted panel configuration analysis and optimization processing on tumor candidate target stability domain data, and generate optimized tumor early screening targeted panel configuration data; In this embodiment of the invention, by combining tumor type-related data, the clinical characteristics, detection challenges, and core target requirements of different tumor types are analyzed. The screening criteria for core targets and panel configuration design requirements corresponding to each tumor type are clarified, providing guidance for panel configuration analysis. Subsequently, based on the combined unit information and core target information in the stable domain data of tumor candidate targets, combined with the results of tumor type characteristic analysis, a multi-target early screening targeted panel efficacy decision system is designed. The core efficacy objectives and weight allocation of the panel configuration are clarified. The efficacy objectives cover detection sensitivity, detection specificity, and target adaptability, ensuring that the panel configuration can meet the core needs of clinical early tumor screening. Next, based on the efficacy decision system, panel configurations are designed for the combined units of the stable domains of tumor candidate targets. Core targets and auxiliary targets that meet the efficacy objectives are screened out. The target composition, signal response judgment criteria, and detection range of each panel configuration are determined, forming a preliminary early tumor screening targeted panel configuration scheme. Subsequently, the initial panel configuration was optimized through risk scenario simulation, simulating various interference scenarios that may occur in clinical testing, including interference from low-concentration samples, noise enhancement interference, and target mutation heterogeneity interference. The performance of the initial configuration under these risk scenarios was tested, and deficiencies were identified, such as insufficient target coverage and unreasonable signal response thresholds. Based on the issues identified in the simulation, the panel configuration was specifically optimized and adjusted, modifying the target composition, signal response threshold, and detection range to improve the stability and performance of the panel configuration under various risk scenarios. After optimization, the performance of the adjusted panel configuration was validated to ensure that its performance indicators met the preset performance targets. The target composition, performance parameters, risk scenario adaptability, and optimization records of the optimized panel configuration were integrated to generate optimized tumor early screening targeted panel configuration data.

[0013] S4: Panel constraint recognition tumor characterization signal analysis module, used to perform panel constraint recognition tumor characterization signal feature analysis on weighted tumor characterization signal mapping data based on optimized tumor early screening targeted panel configuration data, and generate panel constraint recognition tumor characterization signal feature mapping data. In this embodiment of the invention, measurable feature boundary analysis is performed on optimized tumor early screening targeted panel configuration data to clarify the core feature parameter range, logical space boundary, and signal response judgment criteria of the detectable tumor characterization signals for each panel configuration. This defines the measurable signal range for each panel configuration, providing a clear boundary basis for subsequent signal constraint identification. Subsequently, the defined measurable feature boundaries of the panel configurations are correlated and matched with weighted tumor characterization signal mapping data. Each signal in the weighted tumor characterization signals is verified one by one, and measurable constraint signals whose signal feature parameters are within the measurable boundary range of the corresponding panel configuration and meet the signal response judgment criteria are selected. Invalid signals that exceed the measurable boundary or do not meet the judgment criteria are eliminated, forming a preliminary constrained tumor characterization signal feature mapping set. Then, signal response difference analysis is performed on the pre-constrained signal set to compare the differences in core feature parameters and signal response patterns of the measurable constraint signals corresponding to different panel configurations, clarifying the sources of difference and influencing factors, providing a basis for subsequent signal correction. To address the signal response differences identified in the analysis, targeted correction methods were employed to eliminate response variations between signals corresponding to different panel configurations. This ensured the compatibility of signal feature parameters with the panel configuration and verified whether the corrected signals still met the measurable feature boundary requirements, guaranteeing the effectiveness of the corrected signals. After correction, all corrected signals were fused, integrating tumor characterization signal features corresponding to different panel configurations to form a unified and highly reliable set of tumor characterization signal features. This enhanced the tumor discriminative power and stability of the signals, eliminated invalid weak signals after fusion, and normalized the feature parameters of the fused signals to eliminate differences in parameter magnitude. Finally, based on the fused signal feature set and the measurable feature boundaries of the panel configurations, panel-combined tumor characterization signal feature constraint identification was performed. This clarified the matching relationship between each signal and the panel configuration, identified core tumor characterization signal features that met the constraint requirements, labeled the corresponding tumor types and panel configuration compatibility, and integrated all constraint identification results, signal feature parameters, and matching details to generate panel constraint identification tumor characterization signal feature mapping data.

[0014] S5: Dynamic Tumor Early Screening Targeted Panel Establishment Module, used to dynamically iterate the establishment of tumor early screening targeted panels based on panel constraint recognition tumor characterization signal feature mapping data to optimize tumor early screening targeted panel configuration data, and generate dynamic tumor early screening targeted panel data.

[0015] In this embodiment of the invention, by combining optimized tumor early screening targeted panel configuration data and panel constraint recognition tumor characterization signal feature mapping data, a tumor early screening targeted panel discrimination path is constructed. This clarifies the complete logical process from panel configuration adaptation to accurate tumor type discrimination, defines the discrimination nodes, signal parameter judgment criteria, and fault tolerance range for each discrimination path, and clarifies the discrimination logic corresponding to different panel configurations, the correspondence rules between signal parameters and tumor types, ensuring that the discrimination path can achieve accurate tumor type discrimination based on the signal features after constraint recognition. Subsequently, prior evaluation data of clinically validated tumor early screening targeted panels is obtained. Clinical test sample data consistent with the current panel configuration, panel efficacy evaluation results, and historical discrimination path evaluation data are collected. The collected prior evaluation data is screened and verified, eliminating abnormal and incomplete data to ensure the authenticity, completeness, and relevance of the data, serving as a benchmark reference for subsequent panel discrimination path evaluation and configuration optimization. Next, based on prior evaluation data, the constructed tumor early screening targeted panel discrimination path was subjected to multi-target panel parallel tumor early screening discrimination evaluation. The discrimination paths corresponding to the two panel configurations ran in parallel and were applied to tumor type discrimination of the prior evaluation samples. The discrimination accuracy, performance consistency, and operational stability of the discrimination paths were evaluated. Deficiencies in the discrimination paths and panel configurations were identified, and abnormal discrimination cases and their causes were recorded, generating a complete parallel discrimination evaluation report. Subsequently, based on the parallel discrimination evaluation report, the optimized tumor early screening targeted panel configuration data underwent reverse dynamic iterative processing. For the problems identified in the evaluation, the target composition, signal response threshold, measurable feature boundaries, and logical nodes of the discrimination paths in the panel configurations were adjusted. Signal fusion and constraint recognition rules were optimized to improve the discrimination efficiency and stability of the panel configurations. During the iteration process, after each round of adjustments, a parallel discrimination evaluation was performed again to verify the optimization effect. Clear iteration termination conditions were set until the discrimination efficiency, accuracy, and stability of the panel configurations reached the preset targets, at which point the iteration stopped. Finally, by integrating and iteratively optimizing the panel configuration parameters, discrimination path logic, performance evaluation results, and iteration records, the core configuration, applicable scope, and performance indicators of the dynamic early cancer screening targeted panel are clarified, and dynamic early cancer screening targeted panel data is generated. This data can adapt to the dynamic needs of clinical early cancer screening and achieve precise optimization of the early cancer screening targeted panel.

[0016] Furthermore, the tumor characterization signal analysis module includes the following functions: Retrieve pre-stored multimodal tumor-related data from medical databases; In this embodiment of the invention, multimodal tumor-related data pre-stored in a medical database is acquired. The medical database adopts a distributed storage architecture, and the stored data has been clinically validated and fully annotated. The tumor type data includes pathological classification data for three common tumors: lung cancer, gastric cancer, and colorectal cancer. Each pathological classification corresponds to 1000 clinical sample data, and each sample data is annotated with tumor differentiation degree, invasion depth, and clinical stage information. The cfDNA mutation site data consists of specific gene site mutation information in the peripheral blood of each sample, covering six core mutation sites: TP53 gene at positions 175 and 248, KRAS gene at positions 12 and 13, and EGFR gene exon 19 deletion and exon 21 L858R mutation. For each site, the base mutation type, mutation frequency, and mutation abundance are recorded. The mutation frequency detection accuracy is controlled at 0.01%, and the mutation abundance is calculated using the percentage of sequencing reads. The methylation region distribution data targeted the specific methylation gene promoter regions of the three tumor types mentioned above, including the CDKN2A gene promoter region and the MGMT gene promoter region, totaling eight core regions. For each region, the number of methylation sites, methylation coverage, and methylation intensity were recorded. Methylation coverage was calculated as the proportion of methylated sites to the total number of sites in that region, and methylation intensity was quantified using absorbance values ​​from quantitative fluorescence detection. DNA fragment length statistics recorded the length distribution of cfDNA fragments in each sample, ranging from 100-200 bp, divided into 20 length intervals at 10 bp intervals. The number and proportion of fragments of corresponding length were recorded for each interval. Tumor structural variation data recorded three types of variations: chromosomal translocation, deletion, and duplication, clearly identifying the start and end positions and length of the variant fragments. Peripheral blood background noise data recorded the intensity of non-tumor-related signals generated during the detection process, quantified as absorbance values, with a range controlled between 0.001 and 0.005.

[0017] Based on multimodal tumor-related data, logical space mapping analysis of tumor-related signals is performed to generate multimodal tumor-related signal mapping data; In this embodiment of the invention, modal heterogeneity analysis is performed on multimodal tumor-related data. The linear correlation coefficient between each modality and tumor type data is calculated. Specifically, the correlation coefficient between cfDNA mutation site data and tumor type data is controlled at 0.75-0.85; the correlation coefficient between methylation region distribution data and tumor type data is controlled at 0.70-0.80; and the correlation coefficients between DNA fragment length statistics, tumor structural variation data, peripheral blood background noise data, and tumor type data are controlled at 0.60-0.70, 0.65-0.75, and 0.10-0.20, respectively. The heterogeneous association strength between each modality is determined based on the correlation coefficients. Subsequently, multimodal tumor-related signal analysis is performed on the multimodal tumor-related data to extract the core feature parameters of each modality. The cfDNA mutation site data is extracted... Two core parameters, mutation frequency and mutation abundance, are extracted from methylation region distribution data to extract methylation coverage and methylation intensity. Fragment length statistics are used to extract the proportion of fragments in each length interval. Tumor structural variation data is used to extract variation length parameters. Peripheral blood background noise data is used to extract noise intensity parameters. Based on modal heterogeneity correlation strength, tumor-related modal heterogeneity mapping rules are designed. cfDNA mutation site data and methylation region distribution data with correlation coefficients greater than 0.70 are mapped to the core region of the logical space; DNA fragment length statistics and tumor structural variation data with correlation coefficients between 0.60 and 0.70 are mapped to the middle region of the logical space; and peripheral blood background noise data with correlation coefficients less than 0.20 are mapped to the edge region of the logical space, generating multimodal tumor-related signal mapping data.

[0018] Signal stability analysis was performed on multimodal tumor-related signal mapping data under different measurement conditions for each modality to generate multimodal tumor-related signal stability data; In this embodiment of the invention, signal stability analysis is performed on multimodal tumor-related signal mapping data under various modal difference measurement conditions. Three fixed difference measurement conditions are set: the first measurement condition is to maintain an ambient temperature of 25°C and humidity of 50%, repeated 5 times; the second measurement condition is to maintain an ambient temperature of 23°C and humidity of 45%, repeated 5 times; and the third measurement condition is to maintain an ambient temperature of 27°C and humidity of 55%, repeated 5 times. For each modality of signal mapping data, the deviation between the characteristic parameter value measured each time under the three measurement conditions and the average value of the characteristic parameter of the signal mapping data for that modality is calculated. The deviation value is calculated using the absolute difference method. The average deviation value and the maximum deviation value of each modality of signal mapping data under the three measurement conditions are calculated. The average deviation is the arithmetic mean of all single-measurement deviations for that modality, and the maximum deviation is the maximum value among all single-measurement deviations for that modality. Signal stability criteria are defined: modal signals with an average deviation ≤ 0.02 and a maximum deviation ≤ 0.05 are considered stable; modal signals with an average deviation between 0.02 and 0.03 and a maximum deviation between 0.05 and 0.06 are considered basically stable; and modal signals with an average deviation > 0.03 or a maximum deviation > 0.06 are considered unstable. The above operations are performed on the signal mapping data for all modalities, recording the average deviation, maximum deviation, and stability determination results for each modality. These data are then integrated to generate multimodal tumor-related signal stability data.

[0019] We perform contribution discriminant analysis on the multimodal tumor-related signal mapping data for each modality to generate multimodal tumor-related signal contribution discriminant data. In this embodiment of the invention, pathological classification and clinical stage in tumor type data are used as the discrimination criteria. The correlation between the characteristic parameter values ​​of each modality signal mapping data and tumor data of different pathological classifications and clinical stages is calculated. The correlation is calculated using the mutual information calculation method. The larger the mutual information value, the stronger the discriminative contribution of the modality signal to tumor characterization. For cfDNA mutation site signal mapping data, the mutual information values ​​of its mutation frequency and mutation abundance parameters with three pathological classifications (lung cancer, gastric cancer, and colorectal cancer) and with clinical stages I, II, III, and IV are calculated to ensure that each parameter corresponds to the mutual information value of different pathological classifications and clinical stages. For methylation region distribution signal mapping data, the mutual information values ​​of its methylation coverage and methylation intensity parameters with three pathological classifications and four clinical stages are calculated. For D For NA fragment length statistical signal mapping data, the proportion of fragments in each length interval was calculated, along with the mutual information values ​​for three pathological subtypes and four clinical stages. For tumor structural variation signal mapping data, the variation length parameter was calculated, along with the mutual information values ​​for three pathological subtypes and four clinical stages. For peripheral blood background noise signal mapping data, the noise intensity parameter was calculated, along with the mutual information values ​​for three pathological subtypes and four clinical stages. Contribution discriminant criteria were established: modal signals with a mutual information value ≥ 0.8 were classified as high contribution discriminant signals, modal signals with a mutual information value between 0.6 and 0.8 were classified as medium contribution discriminant signals, and modal signals with a mutual information value < 0.6 were classified as low contribution discriminant signals. All mutual information values ​​and contribution discriminant judgment results for each modality were recorded and integrated to generate multimodal tumor-related signal contribution discriminant data.

[0020] Based on the stability data and discriminative contribution data of multimodal tumor-related signals, a joint steady-state and discriminative analysis of tumor-related signal attribute features is performed to generate tumor-related signal attribute feature data, and tumor-related signal attribute weight parameters are designed based on the tumor-related signal attribute feature data. In this embodiment of the invention, core parameters are extracted from two types of data. The average deviation and maximum deviation of each modal signal are extracted from the stability data, and the average mutual information value of each modal signal is extracted from the contribution discriminative data. The average deviation value is converted into a stability coefficient: stability coefficient = 1 - average deviation value / 0.03. The average mutual information value is converted into a contribution coefficient: contribution coefficient = average mutual information value / 0.8. When the average mutual information value ≥ 0.8, the attribute feature comprehensive value of each modal signal is calculated through a joint steady-state and discriminative driving model: comprehensive value = stability coefficient × 0.4 + contribution coefficient × 0.6. This model achieves joint consideration of stability and contribution through fixed weight allocation, with stability accounting for 40% and contribution accounting for 60%, ensuring that attribute feature analysis takes into account both signal reliability and discriminative value. Based on the comprehensive value, each modal signal is divided into core attribute signals. The system identifies three categories of signal attributes: core attributes (≥0.8), important attributes (0.6-0.8), and general attributes (<0.6). It generates tumor-related signal attribute feature data, including the attribute type, stability coefficient, contribution coefficient, and overall value for each modality. Based on this data, it designs weight parameters for tumor-related signal attributes: core attribute signals have a weight of 0.8-0.9, important attribute signals have a weight of 0.6-0.8, and general attribute signals have a weight of 0.3-0.6. The specific values ​​of the weight parameters are linearly distributed according to the overall value; the higher the overall value, the larger the weight parameter. Signals of the same attribute type are assigned corresponding weights according to their overall value, ensuring a perfect match between the weight parameters and the signal attribute features. The total weight parameter is controlled to be 1, generating a clear table of tumor-related signal attribute weight parameters.

[0021] The multimodal tumor-related signal mapping data is weighted and fused using tumor-related signal attribute weight parameters to generate weighted tumor characterization signal mapping data.

[0022] In this embodiment of the invention, the weight parameter values ​​corresponding to each modality signal mapping data are clearly defined. The specific weight values ​​of each modality signal are extracted from the weight parameter table. The weight of the cfDNA mutation site signal in the core attribute signal is set to 0.85, and the weight of the methylation region distribution signal is set to 0.82. The weight of the tumor structural variation signal in the important attribute signal is set to 0.72, and the weight of the DNA fragment length statistics signal is set to 0.68. The weight of the peripheral blood background noise signal in the general attribute signal is set to 0.35. For each modality signal feature parameter value corresponding to each logical space coordinate position, the corresponding weight parameter value is multiplied by the weight parameter value to calculate the weighted sum of each coordinate position. The above weighted sum calculation operation is performed on all coordinate positions in the logical space. During the fusion process, the signal data corresponding to the coordinate positions with a weighted sum value < 0.1 are removed, and such data is judged as invalid signals. The signal data corresponding to the coordinate positions with a weighted sum value ≥ 0.1 are retained. Finally, weighted tumor characterization signal mapping data is generated. This data includes the coordinate positions, weighted sum values, and corresponding original modality signal identifiers of the effective signals in the logical space, which can accurately reflect the core signal characteristics of tumor characterization.

[0023] Furthermore, the multimodal tumor-related data includes tumor type data, cfDNA mutation site data, methylation region distribution data, DNA fragment length statistics, tumor structural variation data, and peripheral blood background noise data.

[0024] Furthermore, the logical space mapping analysis of tumor-related signals based on multimodal tumor-related data includes: Modal heterogeneity analysis is performed based on multimodal tumor-related data to generate tumor-related modal heterogeneity data; In this embodiment of the invention, the multimodal tumor-related data includes tumor type data, cfDNA mutation site data, methylation region distribution data, DNA fragment length statistics, tumor structural variation data, and peripheral blood background noise data. Using tumor type data as the baseline reference data, the Pearson linear correlation coefficient calculation method is employed to calculate the heterogeneous association strength between the other five modalities and the tumor type data. During the calculation process, outliers in each modality are removed. Outliers are defined as data deviating from the modality's average by more than three standard deviations to ensure the accuracy of the calculation results. Specifically, the correlation coefficient calculation between cfDNA mutation site data and tumor type data focuses on the mutation frequency and abundance parameters of six core mutation sites; the methylation region distribution data focuses on the methylation coverage and methylation intensity parameters of eight core regions; the DNA fragment length statistics focus on the fragment proportion parameters of 20 length intervals; the tumor structural variation data focuses on variation length parameters; and the peripheral blood background noise data focuses on noise intensity parameters. The correlation coefficients between cfDNA mutation site data and tumor type data were calculated to be consistently between 0.78 and 0.82; between methylation region distribution data and tumor type data, between 0.73 and 0.77; between DNA fragment length statistics and tumor type data, between 0.63 and 0.67; between tumor structural variation data and tumor type data, between 0.68 and 0.72; and between peripheral blood background noise data and tumor type data, between 0.13 and 0.17. Pairwise correlation coefficients among the five modalities were also calculated: the correlation coefficient between cfDNA mutation site data and methylation region distribution data was 0.65-0.69; the correlation coefficient between cfDNA mutation site data and tumor structural variation data was 0.58-0.62; and the pairwise correlation coefficients among the remaining modalities were all below 0.55. All calculated correlation coefficients, core parameters of each modality, and outlier removal records were integrated to generate tumor-related modal heterogeneous relationship data.

[0025] Multimodal tumor-related signal analysis is performed based on multimodal tumor-related data to generate multimodal tumor-related signal data; In this embodiment of the invention, core signal parameters are extracted according to fixed analytical standards for tumor-related data of each modality to ensure the uniformity and accuracy of the analytical results. For cfDNA mutation site data, three core signal parameters are analyzed for each sample: mutation type, mutation frequency, and mutation abundance. The mutation frequency is accurate to 0.001%, and the mutation abundance is accurate to 0.01% as a percentage of sequencing reads. Missense mutations and nonsense mutations are clearly distinguished and quantified using 01 and 02, respectively. For methylation region distribution data, three core signal parameters are analyzed for the number of methylation sites, methylation coverage, and methylation intensity in the promoter regions of eight core genes. The methylation coverage is calculated as the proportion of methylated sites to the total number of sites in the region, accurate to 0.01%. The methylation intensity is quantified using absorbance values ​​from quantitative fluorescence detection, controlled within the range of 0.1-0.9, accurate to 0.001. For DNA fragment length statistics, the system analyzes two core signal parameters: the number of fragments and the fragment percentage for each of 20 length intervals (100-200 bp, 10 bp intervals). The fragment percentage is accurate to 0.01%, ensuring that the sum of the fragment percentages for all intervals is 100%. For tumor structural variation data, the system analyzes three core signal parameters: the start position, termination position, and variation length for three types of variations: chromosomal translocation, deletion, and duplication. The position is accurate to the base pair, and the variation length is accurate to 10 bp. For peripheral blood background noise data, the system analyzes the intensity of non-tumor-related signals generated during the detection process, quantifying it as absorbance values ​​within the range of 0.002-0.004, accurate to 0.0001. For tumor type data, four core signal parameters were analyzed for each sample: pathological type, differentiation degree, invasion depth, and clinical stage. In pathological type analysis, lung cancer, gastric cancer, and colorectal cancer were quantified using 001, 002, and 003, respectively. In differentiation degree analysis, high, medium, and low were quantified using 1, 2, and 3, respectively. Invasion depth was measured precisely in millimeters. Clinical stages I, II, III, and IV were quantified using 10, 20, 30, and 40, respectively. All analyzed core signal parameters, quantification rules, and analysis standards were integrated to generate multimodal tumor-related signal data.

[0026] Tumor-related modal heterogeneity mapping rules are designed based on tumor-related modal heterogeneity relation data, and then used to perform logical space mapping processing on multimodal tumor-related signal data to generate multimodal tumor-related signal mapping data.

[0027] In this embodiment of the invention, based on the correlation coefficient in tumor-related modal heterogeneous relationship data, a three-level heterogeneous mapping rule is designed to clarify the mapping region, coordinate allocation standard, and parameter normalization requirements of each modality signal in the logical space. The first-level mapping rule targets cfDNA mutation site signals and methylation region distribution signals with a correlation coefficient ≥ 0.73, mapping them to the core region of the three-dimensional logical space (X-axis 0.7-1.0, Y-axis 0.7-1.0, Z-axis 0.7-1.0). Specifically, cfDNA mutation site signals are assigned to the X-axis 0.85-1.0, Y-axis 0.85-1.0, and Z-axis 0.85-1.0 intervals, while methylation region distribution signals are assigned to the X-axis 0.7-0.85, Y-axis 0.7-0.85, and Z-axis 0.7-0.85 intervals. The second-level mapping rule targets DNA fragment length statistical signals and tumor structural variation signals with correlation coefficients between 0.63 and 0.72, mapping them to the middle region of the three-dimensional logical space (X-axis 0.4-0.7, Y-axis 0.4-0.7, Z-axis 0.4-0.7). Specifically, tumor structural variation signals are assigned to the intervals of X-axis 0.6-0.7, Y-axis 0.6-0.7, and Z-axis 0.6-0.7, while DNA fragment length statistical signals are assigned to the intervals of X-axis 0.4-0.6, Y-axis 0.4-0.6, and Z-axis 0.4-0.6. The third-level mapping rule targets peripheral blood background noise signals with correlation coefficients between 0.13 and 0.17, mapping them to the edge region of the three-dimensional logical space (X-axis 0.0-0.4, Y-axis 0.0-0.4, Z-axis 0.0-0.4). During the mapping process, the core parameters of each modality signal correspond to a coordinate point in the logical space. The above mapping operation is performed on signal data of all modalities, and the coordinate position, original parameter value and modality identifier of each signal parameter in the logical space are recorded. After integration, multimodal tumor-related signal mapping data is generated.

[0028] Furthermore, the tumor candidate target stability domain analysis module includes the following functions: Multi-scale signal fluctuation decomposition processing is performed on weighted tumor characterization signal mapping data to generate multi-scale signal fluctuation data of tumor characterization. In this embodiment of the invention, the weighted tumor characterization signal mapping data includes the coordinate positions of effective signals in a three-dimensional logical space, the weighted sum values, and the corresponding original modal signal identifiers. The weighted sum values ​​are controlled within the range of 0.1-1.0, and the coordinate positions are distributed in the core, middle, and edge regions of the logical space. Multi-scale signal fluctuation decomposition employs a wavelet decomposition algorithm, setting four fixed decomposition scales: Scale 1 (low-frequency scale, corresponding to a signal fluctuation period of 1-5 units), Scale 2 (mid-low-frequency scale, corresponding to a signal fluctuation period of 6-10 units), Scale 3 (mid-high-frequency scale, corresponding to a signal fluctuation period of 11-15 units), and Scale 4 (high-frequency scale, corresponding to a signal fluctuation period of 16-20 units). Each decomposition scale corresponds to a fixed wavelet basis function: Scale 1 uses the db1 wavelet basis function, Scale 2 uses the db2 wavelet basis function, Scale 3 uses the db3 wavelet basis function, and Scale 4 uses the db4 wavelet basis function, ensuring accurate decomposition of signal fluctuations at different scales. During the decomposition process, all weighted sums in the weighted tumor characterization signal mapping data undergo detrending to remove fluctuation interference caused by linear trends. The detrending process employs a multinomial fitting method with a fitting order of 2. After detrending, the weighted sum corresponding to each logical space coordinate position is sequentially decomposed into wavelet components at four scales. After each scale decomposition, the corresponding fluctuation component and residual component are extracted. The fluctuation component represents the signal fluctuation characteristics at that scale, and the residual component represents the signal components not decomposed at that scale, which are then further decomposed at the next scale. After scale 4 decomposition, the fluctuation components and final residual components corresponding to each of the four scales are obtained. Each fluctuation component records four core parameters: the corresponding scale, logical space coordinate position, fluctuation amplitude, and fluctuation period. The fluctuation amplitude is the difference between the maximum and minimum values ​​of the fluctuation component, accurate to 0.001. The fluctuation period is recorded within a fixed range according to the corresponding scale. The final residual component records the logical space coordinate position and residual value. By integrating all the four scale fluctuation components obtained from the decomposition, the final residual components, and each parameter, multi-scale signal fluctuation data of tumor characterization is generated. This data clearly presents the fluctuation characteristics of tumor characterization signals at different scales and clearly distinguishes between low-frequency stable fluctuations and high-frequency noise fluctuations.

[0029] The amplitude-noise ratio relationship of tumor characterization signals is analyzed using multi-scale signal fluctuation data of tumor characterization, and tumor characterization signal amplitude-noise ratio data is generated. In this embodiment of the invention, the multi-scale signal fluctuation data of tumor characterization includes fluctuation components at four scales, a final residual component, and corresponding core parameters. Scales 1 and 2 are low-frequency and mid-low-frequency fluctuation components, corresponding to effective tumor characterization signal fluctuations; scales 3 and 4 are mid-high-frequency and high-frequency fluctuation components, corresponding to noise signal fluctuations; and the final residual component is an invalid signal component. The fluctuation amplitude parameters of each scale fluctuation component are extracted: 1000 effective fluctuation amplitude data points are extracted for scale 1, 1000 for scale 2, 800 for scale 3, and 800 for scale 4. The residual value of the final residual component is extracted and determined to be an invalid signal, not participating in amplitude and noise ratio calculations. The effective signal amplitude of tumor characterization is calculated; the effective signal amplitude is the arithmetic mean of the fluctuation amplitudes of scale 1 and scale 2 fluctuation components. During the calculation, abnormal data with fluctuation amplitudes <0.01 in scale 1 and scale 2 fluctuation components are removed. The noise signal amplitude is calculated as the arithmetic mean of the fluctuation amplitudes of the scale 3 and scale 4 fluctuation components, with outliers of fluctuation amplitude <0.005 being removed. The tumor characterization signal amplitude-to-noise ratio (SNR) is calculated using the formula SNR = 20 × lg(effective signal amplitude / noise signal amplitude), accurate to 0.1 dB, where lg is a common logarithmic operation. For the signal corresponding to each logical space coordinate position, the above calculation operations for effective signal amplitude, noise signal amplitude, and amplitude-to-noise ratio are performed, while recording the corresponding coordinate position, original weighted sum value, and fluctuation amplitude at each scale, clarifying the magnitude of the amplitude-to-noise ratio of the signal at each coordinate position and distinguishing the amplitude difference between the effective signal and the noise signal. The amplitude-to-noise ratio data, effective signal amplitude data, noise signal amplitude data, and corresponding parameters for all coordinate positions are integrated to generate tumor characterization signal amplitude-to-noise ratio data. This data clearly presents the amplitude relationship between the tumor characterization signal and the noise signal, providing a clear basis for subsequent steady-state interval sample selection.

[0030] Based on the amplitude-noise ratio data of tumor characterization signals, sample screening and optimization of the steady-state interval of tumor characterization are performed to obtain optimized steady-state interval sample data of tumor characterization. In this embodiment of the invention, the tumor characterization signal amplitude-noise ratio data includes the amplitude-noise ratio, effective signal amplitude, noise signal amplitude, and corresponding parameters of the signal at each logical space coordinate position. Each sample corresponds to a signal coordinate point in three-dimensional logical space, and each sample is associated with a corresponding clinical sample identifier and tumor-related signal parameters. First, a sample screening process is performed within the steady-state interval of tumor characterization. Fixed sample screening criteria are set: samples with an amplitude-noise ratio ≥ 20.0 dB, effective signal amplitude ≥ 0.3, and noise signal amplitude ≤ 0.05 are considered steady-state samples and included in the steady-state interval of tumor characterization; samples with an amplitude-noise ratio between 15.0 and 20.0 dB, effective signal amplitude between 0.2 and 0.3, and noise signal amplitude between 0.05 and 0.08 are considered borderline steady-state samples and are not included in the steady-state interval; samples with an amplitude-noise ratio < 15.0 dB, effective signal amplitude < 0.2, or noise signal amplitude > 0.08 are considered non-steady-state samples and are directly removed from subsequent analysis. During the screening process, samples at all logical space coordinate positions were evaluated one by one, and the screening results for each sample were recorded. The total number of steady-state samples was counted to ensure that the number of steady-state samples was not less than 60% of the total number of samples. A total of 1200 steady-state samples were obtained in this screening, all of which were associated with corresponding clinical sample identifiers, logical space coordinate positions, amplitude-to-noise ratios, and signal amplitude parameters. These steady-state samples were integrated to generate steady-state interval sample data for tumor characterization. Subsequently, the steady-state interval samples were optimized, with a focus on analyzing the potential redundancy relationship data of tumor sample signals in the steady-state interval sample data for tumor characterization. The criteria for redundancy relationship determination were that the logical space coordinate distance between two samples was ≤0.02, the effective signal amplitude difference was ≤0.01, and the amplitude-to-noise ratio difference was ≤0.5dB. Two samples that met these criteria were determined to have a potential redundancy relationship, with one sample being a redundant sample. Redundancy was eliminated using a mutual information optimization method. The mutual information value between samples with potential redundancy was calculated, focusing on two core parameters: effective signal amplitude and amplitude-to-noise ratio. Sample pairs with a mutual information value ≥ 0.9 were considered highly redundant, and samples with higher mutual information values ​​and more stable effective signal amplitudes were retained, while the other redundant sample was removed. Sample pairs with mutual information values ​​between 0.8 and 0.9 were considered moderately redundant, and samples with more complete clinical data were retained. During the optimization process, all steady-state samples were examined one by one, and 180 redundant samples were removed, ultimately retaining 1020 non-redundant, highly stable steady-state samples. All relevant parameters of these samples were integrated to generate optimized steady-state interval sample data for tumor characterization.

[0031] Sample specificity analysis was performed on the optimized tumor characterization steady-state interval sample data to generate optimized tumor characterization steady-state interval sample specificity data; In this embodiment of the invention, the optimized tumor characterization steady-state interval sample data includes 1020 non-redundant, highly stable steady-state samples. Each sample is associated with a clinical sample identifier, tumor type (lung cancer, gastric cancer, colorectal cancer), logical spatial coordinates, and various signal parameters. Sample specificity analysis focuses on the specific association characteristics between the sample and its corresponding tumor type, clarifying the differences in specific signal parameters among samples of different tumor types, ensuring that each sample can accurately correspond to its tumor type through specific parameters. First, the core signal parameters of each sample are extracted, including the mutation frequency of cfDNA mutation site signals, the methylation coverage of methylation region distribution signals, the variation length of tumor structural variation signals, the proportion of core interval fragments and effective signal amplitude of DNA fragment length statistics signals, and the amplitude-noise ratio parameter. All parameters retain their original precise values. For samples of each tumor type, the mean, standard deviation, and confidence interval of each core signal parameter are calculated. The confidence interval is set at 95% and determined by t-test to ensure the reliability of the parameter statistical results. A specificity criterion was established: a sample was considered specific for its tumor type if all core signal parameters fell within the 95% confidence interval of its tumor type parameters and did not overlap with the 95% confidence intervals of two other tumor types. If one or two core signal parameters fell within the 95% confidence interval of its tumor type and partially overlapped with the 95% confidence intervals of two other tumor types, it was considered a suspected specific sample. If most core signal parameters exceeded the 95% confidence interval of its tumor type or highly overlapped with the 95% confidence intervals of two other tumor types, it was considered a non-specific sample and was discarded. For suspected specific samples, clinicopathological details (differentiation degree, invasion depth) of the tumor type were supplemented to further confirm their specificity, ultimately identifying 995 specific samples. The core signal parameters, tumor type, parameter confidence interval matching, and specificity determination results for each specific sample were recorded and integrated to generate optimized tumor characterization steady-state interval sample specificity data.

[0032] Based on the specific data of the optimized steady-state interval of tumor characterization samples, the candidate tumor characterization samples are clustered to generate clustered tumor characterization sample data. In this embodiment of the invention, the optimized tumor characterization steady-state interval sample-specific data includes 995 specific samples. Each sample is associated with its tumor type, core signal parameters, logical space coordinates, and clinical sample identifier. The purpose of clustering is to group samples with similar signal characteristics into one category, clarify the differences in signal characteristics between different categories, and provide a categorical basis for subsequent target stability domain analysis. The clustering process uses the K-means clustering algorithm, setting the number of clusters to 3, corresponding to lung cancer, gastric cancer, and colorectal cancer, respectively, to ensure accurate matching between clustering results and tumor types. During the clustering process, the core signal parameters of the samples are selected as clustering feature indicators, including cfDNA mutation frequency, methylation coverage, effective signal amplitude, amplitude-noise ratio, logical space X-axis coordinate value, Y-axis coordinate value, and Z-axis coordinate value, totaling 7 feature indicators. A clustering termination condition was set: when the change in cluster centers is ≤0.001 and the average distance error of samples in each category is ≤0.01, the clustering iteration was stopped to ensure the stability of the clustering results. The upper limit of the number of clustering iterations was set to 50. In this case, the termination condition was reached after 32 iterations. After clustering, three clusters were obtained: Category 1 corresponds to 338 lung cancer samples, Category 2 corresponds to 327 gastric cancer samples, and Category 3 corresponds to 330 colorectal cancer samples. The cluster center of each cluster was calculated. The cluster center is the arithmetic mean of seven feature indicators of all samples in the category, accurate to 0.001. At the same time, the distance between each sample and its cluster center was calculated using the Euclidean distance formula. Samples with an Euclidean distance ≤0.05 were considered qualified samples for clustering, and samples with an Euclidean distance >0.05 were considered abnormal samples. In this case, two abnormal samples were obtained. After supplementing their core signal parameters, they were re-clustered, and finally, three clusters without abnormal samples were obtained. By integrating all clustering-related data, clustered tumor characterization sample data is generated. This data enables the classification of samples with similar signal characteristics and clarifies the differences in clustering characteristics among samples of different tumor types.

[0033] By clustering tumor characterization sample data, tumor candidate target stability domain analysis is performed to generate tumor candidate target stability domain data.

[0034] In this embodiment of the invention, the clustered tumor characterization sample data includes 3 cluster categories (corresponding to three tumor types) and 995 samples. Each category has a clear cluster center and characteristic index statistical values. The stable region of tumor candidate targets is defined as the logical space region where the core signal parameters of samples of each tumor type are clustered. The sample signal characteristics within this region are stable and highly correlated with the corresponding tumor type, and can be used as the core region of tumor candidate targets. First, the core data of each cluster category is extracted, including the coordinates of the cluster center (X-axis, Y-axis, Z-axis), the average value and standard deviation of 7 characteristic indices, where the coordinates of the cluster center directly correspond to the core position in the logical space, and the average value of the characteristic indices reflects the signal characteristic benchmark of the samples in this category. For each cluster category, the boundary range of the stable region of tumor candidate targets is determined. The boundary range is calculated based on the coordinates of the cluster center and the standard deviation of the characteristic indices. The boundary range of the logical space X-axis, Y-axis, and Z-axis is the coordinates of the cluster center ± 2 × the standard deviation of the characteristic indices (corresponding to the standard deviation of the logical space coordinates), ensuring that the stable region contains more than 95% of the samples of this category. The stable domain boundaries for Category 1 (lung cancer) are 0.83-0.91 on the X-axis, 0.82-0.90 on the Y-axis, and 0.84-0.92 on the Z-axis; for Category 2 (gastric cancer), the boundaries are 0.78-0.86 on the X-axis, 0.77-0.85 on the Y-axis, and 0.79-0.87 on the Z-axis; and for Category 3 (colorectal cancer), the boundaries are 0.80-0.88 on the X-axis, 0.81-0.89 on the Y-axis, and 0.82-0.90 on the Z-axis. Subsequently, the signal characteristics of samples within each stable domain were analyzed, and parameters related to core targets, such as cfDNA mutation sites and methylation regions, were extracted for all samples within the stable domain. This determined the core candidate targets for each stable domain: for Category 1, targets related to the TP53 and EGFR genes; for Category 2, targets related to the KRAS and CDKN2A genes; and for Category 3, targets related to the MGMT and TP53 genes. Stability metrics were calculated for each stable region, including the mean amplitude-to-noise ratio and the standard deviation of the effective signal amplitude within the stable region. Stable regions with a mean amplitude-to-noise ratio ≥ 22.0 dB and an effective signal amplitude standard deviation ≤ 0.02 were classified as high-stability regions. All three stable regions in this study met this criterion and were therefore classified as high-stability regions. All relevant data were integrated to generate stable region data for tumor candidate targets. This data clearly presents the range and characteristics of stable regions for candidate targets in different tumor types.

[0035] Furthermore, the sample screening and optimization process for the steady-state interval of tumor characterization based on tumor characterization signal amplitude-noise ratio data includes: Based on the amplitude-noise ratio data of tumor characterization signals, sample screening processing of the steady-state interval of tumor characterization is performed to obtain sample data of the steady-state interval of tumor characterization. In this embodiment of the invention, the tumor characterization signal amplitude-noise ratio data includes the amplitude-noise ratio, effective signal amplitude, noise signal amplitude, and corresponding original modal signal identifier for each logical space coordinate position signal. Each sample corresponds to a signal coordinate point in three-dimensional logical space, and each sample is associated with a clinical sample identifier, tumor type, and core signal parameters, aligning with the high-quality sample requirements of the tumor early screening targeted panel dynamic optimization and identification analysis system. A fixed judgment standard is set for the screening process: samples with an amplitude-noise ratio ≥ 20.0 dB, effective signal amplitude ≥ 0.3, and noise signal amplitude ≤ 0.05 are judged as steady-state samples and included in the steady-state range of tumor characterization; samples with an amplitude-noise ratio between 15.0 and 20.0 dB, effective signal amplitude between 0.2 and 0.3, and noise signal amplitude between 0.05 and 0.08 are judged as borderline steady-state samples and are temporarily excluded; samples with an amplitude-noise ratio < 15.0 dB, effective signal amplitude < 0.2, or noise signal amplitude > 0.08 are judged as non-steady-state samples and are directly removed. During the screening process, samples at all logical space coordinate positions are evaluated one by one, and the screening results of each sample are recorded to ensure that the number of steady-state samples is not less than 60% of the total number of samples. All relevant data of these steady-state samples are integrated to generate steady-state interval sample data for tumor characterization.

[0036] The potential redundancy relationship data of tumor sample signals in the steady-state interval of tumor characterization is analyzed, and the mutual information optimization processing of the steady-state interval samples of tumor characterization is performed through the potential redundancy relationship data of tumor sample signals to generate optimized steady-state interval sample data of tumor characterization.

[0037] In this embodiment of the invention, firstly, for steady-state samples in the steady-state interval of tumor characterization sample data, the potential redundancy relationship between samples is analyzed, and a fixed redundancy judgment standard is set: the logical spatial coordinate distance between two samples is ≤0.02, the effective signal amplitude difference is ≤0.01, the amplitude-noise ratio difference is ≤0.5dB, and the cfDNA mutation frequency difference is ≤0.005% and the methylation coverage difference is ≤0.01%. Two samples that meet all the above conditions are judged to have a potential redundancy relationship, and one of the samples is a redundant sample. Subsequently, the mutual information optimization processing method is used to eliminate redundancy. The mutual information value of sample pairs with potential redundancy relationship is calculated. The mutual information value calculation focuses on four core parameters: effective signal amplitude, amplitude-noise ratio, cfDNA mutation frequency, and methylation coverage. Sample pairs with mutual information values ​​≥0.9 are judged to be highly redundant, and the sample with higher mutual information value and smaller fluctuation in effective signal amplitude is retained, while the other redundant sample is removed; sample pairs with mutual information values ​​between 0.8 and 0.9 are judged to be moderately redundant, and the sample with more complete clinical pathological information and clearer tumor staging is retained. During the optimization process, all steady-state samples are examined one by one, redundant samples are eliminated, and finally, non-redundant and highly stable steady-state samples are retained. The clinical identifiers, logical spatial coordinates, various signal parameters and screening and optimization records of these samples are integrated to generate optimized tumor characterization steady-state interval sample data.

[0038] Furthermore, the tumor early screening targeted panel configuration analysis module includes the following functions: Based on the stable domain data of tumor candidate targets, the combined unit analysis of stable domains of tumor candidate targets is performed to generate combined unit data of stable domains of tumor candidate targets. In this embodiment of the invention, the tumor candidate target stability domain data includes the stability domain range, core candidate targets, stability indices, and sample signal characteristic statistics for three cluster categories (corresponding to lung cancer, gastric cancer, and colorectal cancer). The core candidate targets cover five gene-related targets: TP53, EGFR, KRAS, CDKN2A, and MGMT, which aligns with the requirements of the tumor early screening targeted panel dynamic optimization and identification analysis system for precise target combinations. First, the tumor candidate target stability domain data undergoes candidate target stability domain feature analysis, extracting core feature parameters for each stability domain, including the logical space boundary range (X-axis, Y-axis, Z-axis coordinates), cluster center coordinates, average amplitude-noise ratio of samples within the stability domain, standard deviation of effective signal amplitude, and signal response threshold of the core candidate targets. The signal response threshold for the core targets in the lung cancer stability domain is set to ≥0.8, ≥0.78 for gastric cancer, and ≥0.79 for colorectal cancer. All feature parameters are accurate to 0.001. Subsequently, adjacency analysis of candidate target stable domains was performed based on the candidate target stable domain feature data. The Euclidean distance method was used to calculate the distance between cluster centers of different stable domains. The distance calculation focused on the X, Y, and Z axis coordinates of the logical space. Adjacency judgment criteria were set: stable domains with a cluster center distance ≤ 0.08 and a core target signal response threshold difference ≤ 0.02 were determined to be adjacent stable domains. It was found that the stable domains of lung cancer and colorectal cancer were adjacent stable domains, while both, along with the stable domain of gastric cancer, were non-adjacent stable domains. Next, coexistence feature analysis of the tumor candidate target stable domain adjacency data was performed, statistically analyzing the common core candidate targets and signal characteristics of adjacent stable domains. It was found that the common core candidate target of the lung cancer and colorectal cancer stable domains was the TP53 gene-related target, with common signal characteristics including an average amplitude-to-noise ratio ≥ 22.3 dB and an effective signal amplitude standard deviation ≤ 0.018. No common core candidate targets were found in non-adjacent stable domains. Finally, based on the stable domain feature data and coexistence feature data of candidate targets, combinatorial unit analysis was performed. According to the principle of "merging adjacent stable domains and keeping non-adjacent stable domains independent", combinatorial units were divided. The stable domains of lung cancer and colorectal cancer were merged into one combinatorial unit, while the stable domain of gastric cancer was an independent combinatorial unit. At the same time, the core candidate targets, stable domain boundaries, adjacency relationships, coexistence features and sample signal statistics within each combinatorial unit were integrated. Each combinatorial unit was clearly labeled with the corresponding tumor type, the number of core targets and the stability level, generating tumor candidate target stable domain combinatorial unit data, which provides an accurate target combination basis for subsequent panel configuration analysis.

[0039] Tumor type characteristics are analyzed by tumor type data to obtain tumor type characteristic data, and multi-target early screening targeted panel efficacy decision is designed based on tumor type characteristic data; In this embodiment of the invention, the tumor type data includes pathological classification, differentiation degree, invasion depth, clinical stage, and clinical sample data of 1000 cases / type of tumors for lung cancer, gastric cancer, and colorectal cancer, consistent with the multimodal tumor-related data mentioned above. First, tumor type characteristic analysis is performed. For each tumor type, the proportion of pathological classifications in clinical samples, the distribution of sample numbers at different clinical stages, the mutation frequency of core candidate targets, and the methylation coverage are statistically analyzed. Simultaneously, the detection challenges for each tumor type are analyzed: for lung cancer, the challenge is weak signal in early-stage samples; for gastric cancer, the challenge is strong heterogeneity of core target mutations; and for colorectal cancer, the challenge is large differences in target signals at different stages. These statistical data, detection challenges, and tumor type-specific characteristics are integrated to generate tumor type characteristic data. Subsequently, based on this data, a multi-target early screening targeted panel efficacy decision was designed, clarifying three core efficacy targets and their weight allocations: detection sensitivity (40%), detection specificity (35%), and target suitability (25%). The detection sensitivity targets were set at ≥92% for lung cancer, ≥90% for gastric cancer, and ≥91% for colorectal cancer; the detection specificity targets were all set at ≥93%; and the target suitability target was a ≥95% match between the core target and the tumor type. The efficacy decision clearly defined the criteria and calculation methods for each efficacy target: detection sensitivity = number of true positive samples / (number of true positive samples + number of false negative samples); detection specificity = number of true negative samples / (number of true negative samples + number of false positive samples); and target suitability = number of matching core targets / total number of core targets. This ensured the operability and accuracy of the efficacy decision, providing a clear efficacy guide for subsequent panel configuration analysis.

[0040] Using the efficacy decision of multi-target early screening targeted panel, tumor candidate target stable domain combination unit data were used to perform tumor early screening targeted panel configuration analysis to generate tumor early screening targeted panel configuration data. In this embodiment of the invention, the tumor candidate target stability domain combination unit data includes two combination units (lung cancer + colorectal cancer combination unit and gastric cancer independent combination unit). The efficacy decision of the multi-target early screening targeted panel clarifies the efficacy target, weight, and judgment criteria. First, for each combination unit, core candidate targets that meet the efficacy target are screened. For the lung cancer + colorectal cancer combination unit, three gene-related targets, TP53, EGFR, and MGMT, are prioritized. Specifically, mutation sites at positions 175 and 248 of the TP53 gene, deletion sites at exon 19 and L858R mutation sites at exon 21 of the EGFR gene, and methylation sites in the promoter region of the MGMT gene are selected to ensure that the detection sensitivity and specificity of these targets meet the efficacy target. For the gastric cancer combination unit, two gene-related targets, KRAS and CDKN2A, are screened. Mutations at positions 12 and 13 of the KRAS gene and methylation sites in the promoter region of the CDKN2A gene are selected, which also meet the efficacy target requirements. Subsequently, the panel configuration for each combination unit was determined, adopting a "core target + auxiliary target" structure. The core targets were the five gene-related targets selected above, and the auxiliary targets were secondary targets with stable signal responses within each stable domain and adapted to the performance target. The lung cancer + colorectal cancer combination unit had two auxiliary targets, and the gastric cancer combination unit had one auxiliary target. The signal response threshold for the auxiliary targets was set to ≥0.75. Next, the overall performance value of each panel configuration scheme was calculated: Overall value = detection sensitivity × 0.4 + detection specificity × 0.35 + target adaptability × 0.25. Configuration schemes with an overall performance value ≥0.91 were selected, with the lung cancer + colorectal cancer combination unit having an overall performance value of 0.923 and the gastric cancer combination unit having 0.915, both meeting the requirements. Finally, the target composition, target mutation sites / methylation regions, signal response thresholds, overall performance values, and corresponding combination units of each panel configuration are recorded to clarify the criteria for determining the detection range of the panel configuration. After integration, tumor early screening targeted panel configuration data are generated.

[0041] The panel configuration of the tumor early screening targeted panel configuration data is optimized by performing risk scenario simulation, thereby generating optimized tumor early screening targeted panel configuration data.

[0042] In this embodiment of the invention, the tumor early screening targeted panel configuration data includes a panel configuration scheme, target composition, comprehensive efficacy value, and detection judgment criteria for two combined units. Risk scenario simulation focuses on interference scenarios that may occur in clinical testing, aligning with the clinical application needs of the system. Three fixed risk scenarios are set: the first scenario is a low-concentration cfDNA sample interference scenario, with the cfDNA concentration controlled at 5-10 ng / mL, simulating the weak signal of early tumor samples; the second scenario is a peripheral blood background noise enhancement scenario, with the noise signal amplitude controlled at 0.06-0.09, simulating environmental interference during detection; the third scenario is a target mutation heterogeneity interference scenario, with the core target mutation frequency fluctuation range set at ±8%, simulating individual differences in tumors. For each risk scenario, the existing panel configuration data is simulated and tested, and the detection sensitivity, specificity, target adaptability, and number of false positive and false negative samples for each panel configuration are recorded. Subsequently, panel configuration optimization was performed based on simulation results. For low-concentration scenarios, the core target signal response threshold was lowered by 0.02 to enhance the recognition of weak signals. For scenarios with enhanced background noise, a signal filtering function for auxiliary targets was added, and a signal response difference of ≤0.01 between auxiliary targets was defined as a valid signal to eliminate noise interference. For scenarios with target mutation heterogeneity, one auxiliary target was added to the gastric cancer panel to improve target coverage. After optimization, simulation tests were performed again in three risk scenarios to ensure that the overall efficiency value of the panel configuration was ≥0.92 in all scenarios, and that the detection sensitivity and specificity met the preset targets, with the number of false positives and false negatives ≤3. The target composition, signal response threshold, efficiency parameters, and simulation results of the risk scenarios for the optimized panel configuration were recorded. All optimization-related data were integrated to generate optimized tumor early screening targeted panel configuration data.

[0043] Furthermore, the analysis of tumor candidate target stability domain combination units based on tumor candidate target stability domain data includes: Perform stability domain feature analysis on tumor candidate target stability domain data to generate tumor candidate target stability domain feature data; In this embodiment of the invention, the tumor candidate target stability domain data includes the candidate target stability domain range, core candidate targets, stability indicators, and sample signal feature statistics for three cluster categories (corresponding to lung cancer, gastric cancer, and colorectal cancer, respectively). The core candidate targets cover five gene-related targets: TP53, EGFR, KRAS, CDKN2A, and MGMT. This aligns with the core requirement of the tumor early screening targeted panel dynamic optimization and identification analysis system for precise target combination, providing basic feature support for subsequent combination unit division. The feature analysis process strictly follows a fixed procedure. First, the logical space boundary features of each stable domain are extracted, and the coordinate ranges of the X-axis, Y-axis, and Z-axis in the three-dimensional logical space of each stable domain are determined. Specifically, the boundary of the lung cancer stable domain is X-axis 0.83-0.91, Y-axis 0.82-0.90, and Z-axis 0.84-0.92; the boundary of the gastric cancer stable domain is X-axis 0.78-0.86, Y-axis 0.77-0.85, and Z-axis 0.79-0.87; and the boundary of the colorectal cancer stable domain is X-axis 0.80-0.88, Y-axis 0.81-0.89, and Z-axis 0.82-0.90. All coordinate values ​​are accurate to 0.001. Subsequently, cluster center features were extracted for each stability region. The cluster center was the arithmetic mean of the logical space coordinates of all samples within the corresponding stability region. The cluster center coordinates for lung cancer were (0.87, 0.86, 0.88), for gastric cancer (0.82, 0.81, 0.83), and for colorectal cancer (0.84, 0.85, 0.86), all accurate to 0.001. Next, core stability indicators were extracted, including the average amplitude-to-noise ratio and the standard deviation of the effective signal amplitude within the stability region. The average amplitude-to-noise ratio for lung cancer was 22.5 dB, and the standard deviation of the effective signal amplitude was 0.017; for gastric cancer, it was 22.2 dB, and the standard deviation of the effective signal amplitude was 0.019; and for colorectal cancer, it was 22.4 dB, and the standard deviation of the effective signal amplitude was 0.018. The amplitude-to-noise ratio was accurate to 0.1 dB, and the standard deviation was accurate to 0.001. Finally, the signal features of core candidate targets were extracted, and the signal response thresholds of core targets in each stable domain were set: ≥0.8 for lung cancer, ≥0.78 for gastric cancer, and ≥0.79 for colorectal cancer. The signal response patterns of each core target in the corresponding stable domain were clarified, and tumor candidate target stable domain feature data were generated to ensure that the data accurately correspond to each stable domain, providing clear feature basis for subsequent adjacency relationship analysis.

[0044] Based on the characteristic data of the stable domain of tumor candidate targets, the adjacency relationship analysis of the stable domain of candidate targets is performed to generate the adjacency relationship data of the stable domain of tumor candidate targets. In this embodiment of the invention, the core features of the stable domains of tumor candidate targets, such as the cluster center coordinates and signal response thresholds of each stable domain, have been clearly defined. The core of the adjacency analysis is to determine the degree of adjacency between different stable domains in the logical space, providing an adjacency basis for subsequent coexistence feature analysis and unit division, thus aligning with the system's target panel configuration's requirements for analyzing the correlation of target regions. The analysis process uses the Euclidean distance calculation method, focusing on the cluster center coordinates (X-axis, Y-axis, Z-axis) of each stable domain, and calculating the straight-line distance between the cluster centers of any two stable domains. The cluster center distances for the stable domains of lung cancer and gastric cancer, lung cancer and colorectal cancer, and gastric cancer and colorectal cancer are calculated respectively, yielding a distance of 0.086 between the stable domains of lung cancer and gastric cancer, 0.072 between the stable domains of lung cancer and colorectal cancer, and 0.083 between the stable domains of gastric cancer and colorectal cancer. Simultaneously, fixed adjacency determination criteria were established, requiring the simultaneous fulfillment of two conditions: firstly, the distance between the cluster centers of the two stable domains must be ≤0.08; secondly, the difference in the signal response threshold of the core target points of the two stable domains must be ≤0.02. Failure to meet either condition precludes the domain from being classified as adjacent stable domains. If only one condition is met, or neither condition is met, the domain is classified as non-adjacent stable domains. The criteria were verified one by one. The distance between the stable domains of lung cancer and colorectal cancer was 0.072 ≤0.08, and the difference in the signal response threshold of their core target points was 0.01 ≤0.02, thus classifying them as adjacent stable domains. The distance between the stable domains of lung cancer and gastric cancer was 0.086 >0.08, and the distance between the stable domains of gastric cancer and colorectal cancer was 0.083 >0.08, both classifying them as non-adjacent stable domains. The cluster center distance, signal response threshold difference, and adjacency determination results for all stable domains were recorded to clarify the relationships between adjacent and non-adjacent stable domains. This data was then integrated to generate adjacency relationship data for tumor candidate target stable domains.

[0045] Perform coexistence feature analysis on the adjacency relationship data of stable domains of tumor candidate targets to generate coexistence feature data of stable domains of tumor candidate targets; In this embodiment of the invention, the adjacency relationship data of tumor candidate target stable domains has clearly identified lung cancer and colorectal cancer stable domains as adjacent stable domains, while both and the gastric cancer stable domain are non-adjacent stable domains. Coexistence feature analysis focuses on the shared core targets and shared signal features of adjacent stable domains, exploring the correlation between adjacent stable domains to provide a basis for subsequent combination unit division, thus assisting the system in accurately designing target combination schemes for targeting panels. The analysis process prioritizes adjacent stable domains (lung cancer and colorectal cancer stable domains). First, the core candidate target information for both is extracted. The core candidate targets for the lung cancer stable domain are TP53 and EGFR gene-related targets, while the core candidate targets for the colorectal cancer stable domain are TP53 and MGMT gene-related targets. The core target types of both are compared one by one, and the shared core candidate target is identified as the TP53 gene-related target. The specific parameters of this shared target are determined, including the mutation sites at positions 175 and 248 of the TP53 gene, and the corresponding mutation frequency threshold ≥0.68%. Subsequently, signal characteristic parameters of adjacent stable domains were extracted, including average amplitude-to-noise ratio, effective signal amplitude standard deviation, and overlapping region of logical space coordinates. The average amplitude-to-noise ratio of the lung cancer stable domain was 22.5 dB, and the effective signal amplitude standard deviation was 0.017. The average amplitude-to-noise ratio of the colorectal cancer stable domain was 22.4 dB, and the effective signal amplitude standard deviation was 0.018. The average value of the signal characteristic parameters of the two domains was calculated, and the common signal characteristic standard was determined to be an average amplitude-to-noise ratio ≥ 22.3 dB and an effective signal amplitude standard deviation ≤ 0.018. At the same time, the logical space coordinates of the two domains were analyzed, and the coordinate overlap region was found to be 0.83-0.88 on the X-axis, 0.82-0.89 on the Y-axis, and 0.84-0.90 on the Z-axis. This region is the core region where the signals of the two domains coexist. For non-adjacent stable domains, core target points and signal features were extracted from the stable domains of lung cancer and gastric cancer, and gastric cancer and colorectal cancer, respectively. After comparison, it was found that there were no common core candidate targets among the non-adjacent stable domains, the average values ​​of signal feature parameters differed significantly, and there were no obvious signal coexistence regions, so there was no need to record coexistence features. All analysis results were integrated to generate coexistence feature data for tumor candidate target stable domains.

[0046] Based on the characteristic data of the stable domain of tumor candidate targets and the coexistence characteristic data of the stable domain of tumor candidate targets, the combined unit analysis of the stable domain of tumor candidate targets is performed to generate the combined unit data of the stable domain of tumor candidate targets.

[0047] In this embodiment of the invention, the characteristic parameters, adjacency relationships, and coexistence characteristics of stable domains of tumor candidate targets have been clearly defined for each stable domain. The combined unit analysis follows the fixed principle of "merging adjacent stable domains and keeping non-adjacent stable domains independent." The combined unit division is optimized by combining coexistence characteristics, providing a precise target combined unit basis for subsequent tumor early screening targeted panel configuration analysis, which meets the core requirement of system dynamic optimization. The combined unit division process first clarifies the specific execution criteria of the division principle. Adjacent stable domains that share a core target and signal features are merged into one combined unit. After merging, the shared core target and signal features of both are retained, while their respective exclusive core targets and signal features are integrated to ensure that the combined unit covers the core information of both. Non-adjacent stable domains have no coexistence characteristics and are treated as separate combined units, fully retaining all information such as their own core target, signal features, and logical space boundaries. Following the principle of processing each stable domain individually, the lung cancer and colorectal cancer stable domains were considered adjacent stable domains and shared TP53 gene-related targets and signal characteristics. Therefore, they were merged into a single unit, named the Lung Cancer-Colorectal Cancer Combined Unit. The core candidate targets of this combined unit were TP53, EGFR, and MGMT gene-related targets. The shared signal characteristics were a mean amplitude-to-noise ratio ≥22.3 dB and an effective signal amplitude standard deviation ≤0.018. The logical space boundary was the merged range of the original boundaries of the two units (X-axis 0.80-0.91, Y-axis 0.81-0.90, Z-axis 0.82-0.92). The gastric cancer stable domain was a non-adjacent stable domain and was treated as a separate combined unit, named the Gastric Cancer Combined Unit. Its core candidate targets (KRAS and CDKN2A gene-related targets), signal characteristics (mean amplitude-to-noise ratio 22.2 dB, effective signal amplitude standard deviation 0.019), and original logical space boundary were fully preserved. Each combined unit is labeled with the corresponding tumor type, number of core targets, stability level and coexistence characteristics. The lung cancer-colorectal cancer combined unit has a high stability level, and the gastric cancer combined unit has a high stability level. All core parameters and division criteria of each combined unit are recorded, and the data of tumor candidate target stability domain combined units are generated after integration.

[0048] Furthermore, the Panel constraint recognition tumor characterization signal analysis module includes the following functions: Perform panel configuration measurable feature boundary analysis on optimized tumor early screening targeted panel configuration data to generate panel configuration measurable feature boundary data. In this embodiment of the invention, the optimized tumor early screening targeted panel configuration data includes the optimized target composition, signal response threshold, efficacy parameters, and risk scenario simulation results of two combined units (lung cancer-colorectal cancer combined unit and gastric cancer combined unit). The core targets cover five gene-related targets: TP53, EGFR, MGMT, KRAS, and CDKN2A. This aligns with the core requirement of the tumor early screening targeted panel dynamic optimization and identification analysis system for panel constraint identification. The core of the measurable feature boundary analysis is to clarify the boundary range of detectable tumor characterization signals for each panel configuration, providing a clear constraint basis for subsequent signal feature analysis. The analysis process strictly followed a fixed procedure. First, the core measurable characteristic parameters of each panel configuration were extracted, including the core target signal response threshold, auxiliary target signal response threshold, effective signal amplitude range, amplitude-noise ratio range, and logical space coordinate correlation parameters. Specifically, for the lung cancer-colorectal cancer panel, the core target signal response threshold was ≥0.78, the auxiliary target threshold was ≥0.73, the effective signal amplitude range was 0.28-1.0, and the amplitude-noise ratio range was ≥19.8dB. For the gastric cancer panel, the core target signal response threshold was ≥0.76, the auxiliary target threshold was ≥0.73, the effective signal amplitude range was 0.27-1.0, and the amplitude-noise ratio range was ≥19.7dB. All parameters were accurate to 0.001 or 0.1dB. Subsequently, for the core measurable characteristic parameters of each panel configuration, the criteria for determining the measurable characteristic boundaries were established. Based on the core target signal response threshold, a fluctuation of 0.01 was used as the signal response boundary. The core target signal response boundary for the lung cancer-colorectal cancer panel was 0.77-0.79, and for the gastric cancer panel it was 0.75-0.77. Based on the effective signal amplitude range, the measurable effective signal amplitude boundary was defined. For the lung cancer-colorectal cancer panel, it was 0.27-1.01, and for the gastric cancer panel it was 0.26-1. 01; Based on the amplitude-to-noise ratio range, a downward adjustment of 0.2dB was used as the amplitude-to-noise ratio boundary, with both being ≥19.5dB. Simultaneously, combined with the logical space coordinate range, the logical space boundary of the measurable signal for each panel configuration was defined. The logical space boundary for the lung cancer-colorectal cancer panel was X-axis 0.79-0.92, Y-axis 0.80-0.91, Z-axis 0.81-0.93, and for the gastric cancer panel, it was X-axis 0.77-0.87, Y-axis 0.76-0.86, Z-axis 0.78-0.88. Next, the effectiveness of the measurable feature boundaries was verified. Signals from 50 clinical samples corresponding to each panel configuration and tumor type were selected to check whether their signal parameters were all within the set boundary range. The verification results showed that all sample signal parameters were within the boundary range, and the boundary effectiveness met the standard.The measurable feature parameters, measurable feature boundary range, boundary judgment criteria and validity verification results of each panel configuration are integrated, and the target type and signal parameters corresponding to each boundary are clearly marked to generate panel configuration measurable feature boundary data.

[0049] The panel configuration measurable feature boundary data is transmitted to the weighted tumor characterization signal mapping data to perform preliminary tumor characterization signal feature analysis of the panel configuration measurable boundary, and generate preliminary constrained tumor characterization signal feature mapping data. In this embodiment of the invention, the measurable feature boundary data of the panel configuration has clearly defined the measurable feature boundary ranges of two panel configurations. The weighted tumor characterization signal mapping data includes the coordinate positions, weighted sum values, and corresponding original modal signal identifiers of the effective signals in the three-dimensional logical space. Matching rules are defined, and each signal coordinate point, weighted sum value, and original modal signal parameter in the weighted tumor characterization signal mapping data is matched one by one with the corresponding boundary range in the panel configuration measurable feature boundary data. During the matching process, the core parameters of each signal are verified one by one, including whether the logical space coordinates are within the corresponding panel's measurable logical space boundary, whether the weighted sum value corresponds to the effective signal amplitude boundary, whether the amplitude-to-noise ratio reaches the panel's measurable amplitude-to-noise ratio boundary, and whether the core target signal response value is within the signal response boundary. Signals whose parameters all meet the requirements of the corresponding panel's measurable boundary are determined to be measurable constraint signals and included in the preliminary constraint analysis scope; signals whose parameters do not meet any boundary requirements are determined to be unmeasurable constraint signals and are directly discarded, not participating in subsequent analysis. Subsequently, the coordinates, weighted sum, original modal signal identifier, corresponding panel configuration, and boundary matching results of each measurable constraint signal were recorded. Simultaneously, core parameters such as cfDNA mutation frequency and methylation coverage for each signal were retained. These data were then correlated and integrated with the measurable boundary parameters to generate preliminary constraint tumor characterization signal feature mapping data. This data enables preliminary screening and constraint of weighted tumor characterization signals, eliminating invalid signals that do not meet the measurable boundaries, retaining valid signals that meet the requirements, and clarifying the matching relationship between signals and the panel's measurable boundaries.

[0050] Based on the measurable feature boundary data of the panel configuration, the feature mapping data of the preliminary constrained tumor characterization signal are analyzed to generate the feature response difference data of the tumor characterization signal of the panel combination. In this embodiment of the invention, the signals in the preliminary constraint tumor characterization signal feature mapping data are divided into two analysis units according to the corresponding panel configuration. Analysis unit one is the measurable constraint signal (lung cancer and colorectal cancer related signal) corresponding to the lung cancer-colorectal cancer panel, and analysis unit two is the measurable constraint signal (gastric cancer related signal) corresponding to the gastric cancer panel. Each analysis unit is analyzed for response differences separately, and the response differences between the two analysis units are compared. For each analysis unit, the mean and standard deviation of the core response parameters for all signals within the unit were calculated. The core response parameters included the weighted average, the average amplitude-to-noise ratio, the average core target signal response, and the average methylation coverage. For analysis unit one, the weighted average was 0.68, the average amplitude-to-noise ratio was 22.6 dB, the average core target signal response was 0.83, and the average methylation coverage was 62.3%. For analysis unit two, the weighted average was 0.65, the average amplitude-to-noise ratio was 22.1 dB, the average core target signal response was 0.80, and the average methylation coverage was 60.8%. All statistical data were accurate to 0.001, 0.1 dB, or 0.1%. Subsequently, the response differences of signals from different tumor types within the same analysis unit were calculated. In analysis unit one, the weighted sum difference between lung cancer and colorectal cancer signals was 0.04, the amplitude-to-noise ratio difference was 0.5 dB, and the core target signal response difference was 0.02, all of which were small and within a reasonable range. Analysis unit two contained no signals from different tumor types, and therefore no internal response differences. Next, the signal response differences between the two analysis units were calculated, clarifying that the core sources of the signal response differences between the two analysis units were the different core target types and the differences in measurable boundary parameters. Subsequently, a detailed analysis was conducted on the parameters with significant differences. The difference in core target signal response was mainly due to the fact that the core target signal response threshold (≥0.78) of the lung cancer-colorectal cancer panel was higher than that of the gastric cancer panel (≥0.76), resulting in a higher average value of the core target signal response in analysis unit one. The difference in methylation coverage was mainly due to the fact that the lung cancer-colorectal cancer panel covered the methylation target of the MGMT gene, while the gastric cancer panel focused on the methylation target of the CDKN2A gene. The methylation levels of the two genes have inherent differences in different tumor types. After integration, panel combination tumor characterization signal response difference data were generated.

[0051] Based on the difference data of the characteristic response of the panel combined tumor characterization signal, equivalent correction processing of the characteristic response of the panel combined tumor characterization signal is performed to generate corrected characteristic response data of the panel combined tumor characterization signal. In this embodiment of the invention, a targeted correction scheme is designed to address the signal response differences and parameter deviations between the two analysis units of the panel-combined tumor characterization signal feature response difference data. For the core target signal response differences, a linear correction algorithm is used to up-correct the core target signal response value of analysis unit two (the signal corresponding to the gastric cancer panel), with a correction coefficient set to 1.038, ensuring that the difference between the average core target signal response of analysis unit two and analysis unit one after correction is ≤0.01. For the methylation coverage differences, an offset correction method is used to add a fixed offset to the methylation coverage parameter of analysis unit two. For minor differences in weighted sum values ​​and amplitude-noise ratios, no separate correction is required; the deviation is further weakened through subsequent signal fusion. During the correction process, the parameters of each signal after correction are checked in real time to ensure they meet the measurable feature boundary requirements of the corresponding panel configuration. If the corrected parameters exceed the boundary range, the correction coefficient or offset is readjusted to ensure that all corrected signals meet the measurable constraints. After correction, corrected tumor characterization signal feature mapping data is generated.

[0052] Based on the corrected panel combined tumor characterization signal feature response data, panel constraint recognition tumor characterization signal feature analysis is performed to generate panel constraint recognition tumor characterization signal feature mapping data.

[0053] In this embodiment of the invention, panel-based tumor characterization signal feature fusion processing is performed on the corrected tumor characterization signal feature mapping data. During the fusion process, for signals with similar coordinate positions in the three-dimensional logical space, their weighted fusion feature parameters are calculated, covering core parameters such as weighted sum values, core target signal response values, and methylation coverage. For independent signals with no similar coordinate positions, their corrected parameters are directly retained as fused parameters and labeled as independent feature signals. During the fusion process, signals with a weighted sum value < 0.2 after fusion are removed, as these signals are considered weak signals, and their removal improves the overall quality of the fused signal. At the same time, the parameter comparison data before and after fusion are retained to facilitate subsequent verification of the fusion effect. Subsequently, the stability index and discriminant index of the fused signal are calculated to verify the fusion effect, ensuring that the signal quality requirements of the system constraint recognition are met, and generating the fused characterization signal feature data corresponding to the corrected tumor characterization signal feature mapping data. Based on the fused tumor characterization signal feature data and the measurable feature boundary data of the panel configuration, constraint identification of panel-combined tumor characterization signal features is performed. The constraint identification process strictly follows a fixed identification procedure, extracting core feature parameters from the fused tumor characterization signal feature data and boundary parameters from the measurable feature boundary data of the panel configuration, establishing a constraint identification matching model, and clarifying the matching rules: if all fused signal feature parameters meet the measurable boundary requirements of a certain panel configuration and match the specific signal features of the tumor type corresponding to that panel configuration, it is determined to be an exclusive constraint signal of that panel configuration; if the fused signal feature parameters meet the measurable boundary requirements of two panel configurations simultaneously, it is determined to be a cross-constraint signal, and its main compatible panel configuration is further identified by combining its logical space coordinates and tumor type identifier; if the fused signal feature parameters do not meet the measurable boundary requirements of any panel configuration, it is determined to be an invalid constraint signal and is directly eliminated. During the identification process, the matching of the feature parameters, logical space coordinates, tumor type identifiers, and measurable boundaries of each fused signal with the panel configuration is verified one by one. The identification results are recorded, and the number of specific constraint signals, cross-constraint signals, invalid signals, and constraint identification accuracy for each panel configuration are calculated. After the identification is completed, the constraint identification results, statistical data of constraint signals for each panel configuration, signal matching details, identification accuracy, and invalid signal removal records are integrated. The tumor type, panel configuration, core feature parameters, and constraint matching basis corresponding to each specific constraint signal are clearly marked, generating panel constraint identification tumor characterization signal feature mapping data.

[0054] Furthermore, the dynamic tumor early screening targeting panel establishment module includes the following functions: Analyze the discrimination path data of the targeted panel for early tumor screening based on the optimized configuration data of the panel and the feature mapping data of the panel constraint recognition of tumor characterization signals. In this embodiment of the invention, the optimized tumor early screening targeted panel configuration data includes optimized configuration schemes for two combined units: a lung cancer-colorectal cancer panel (core targets TP53, EGFR, MGMT, core target signal response threshold ≥0.78, amplitude-to-noise ratio ≥19.8dB) and a gastric cancer panel (core targets KRAS, CDKN2A, core target signal response threshold ≥0.76, amplitude-to-noise ratio ≥19.7dB). It also includes the measurable feature boundaries, performance parameters, and risk scenario optimization records for each panel. The panel constraint recognition tumor characterization signal feature mapping data covers the signal data of the entire process of preliminary constraint, correction, fusion, and constraint recognition, with the core being the fused tumor characterization signal feature data and the panel combined tumor characterization signal feature constraint recognition result data. The analysis process employs a path parsing algorithm, using the panel constraint recognition result data as the core, and associating the target composition, measurable boundaries, and other parameters of the optimized panel configuration data to construct a three-level discrimination path. The first-level discrimination path is panel configuration adaptation discrimination, which clarifies the panel configuration corresponding to each fused tumor characterization signal. By matching the signal's logical space coordinates with the boundary of the panel's measurable logical space, it determines the panel unit to which the signal belongs. For example, lung cancer and colorectal cancer-related signals are assigned to the lung cancer-colorectal cancer panel. The second-level discrimination path is preliminary tumor type discrimination. Based on the core target signal response value of the assigned panel configuration, combined with parameters such as the signal's cfDNA mutation frequency and methylation coverage, it preliminarily determines the tumor type corresponding to the signal. The third-level discrimination path is precise tumor type discrimination. It verifies the preliminary discrimination results by combining the specific constraint signal identifier in the constraint recognition results, eliminates misclassification cases in cross-constraint signals, clarifies the specific tumor type and discrimination criteria corresponding to each signal, and sets an error tolerance standard for the discrimination path, controlling the error tolerance rate within 2% to ensure the reliability of the discrimination path. Record the discrimination nodes, judgment criteria, core parameter thresholds, fault tolerance range and path correlation of each discrimination path, clarify the discrimination logic, signal parameters and tumor type correspondence rules corresponding to different panel configurations, and generate tumor early screening targeted panel discrimination path data after integration.

[0055] Obtain prior assessment data from the targeted panel for early cancer screening; In this embodiment of the invention, the core of obtaining prior assessment data for targeted panels in early tumor screening is to collect clinically validated panel assessment data and historical testing data to ensure the authenticity, completeness, and relevance of the data. This provides a reliable reference benchmark for the subsequent parallel evaluation of panel discrimination pathways, aligning with the system's dynamic optimization requirements for clinical data support. The prior assessment data originates from the clinical early tumor screening database of tertiary hospitals. Clinical assessment data consistent with the panel configuration of this system (lung cancer-colorectal cancer panel, gastric cancer panel combination) are selected, covering clinical testing sample data and panel efficacy evaluation results from the past three years. A total of 1500 clinical samples were screened, including 500 lung cancer samples, 500 gastric cancer samples, and 500 colorectal cancer samples. Each sample underwent pathological biopsy confirmation to ensure the accuracy of the diagnostic results. The prior assessment data contains three main categories of core information: the first category is panel efficacy benchmark data, covering core efficacy parameters such as detection sensitivity, detection specificity, false positive rate, and false negative rate when the two panel configurations are used alone and in combination. The second category is historical discrimination pathway evaluation data. The third category is sample matching benchmark data, which includes core signal parameters, corresponding panel configuration identifiers, tumor type diagnosis results, and discrimination path matching of 1,500 clinical samples. This clarifies the distribution range of signal parameters, panel adaptation rules, and the adaptability of discrimination paths for different tumor types. During the acquisition process, the collected prior assessment data were screened and verified, removing abnormal data (data deviating from the mean by more than three standard deviations) and incomplete data (data missing core efficacy parameters or sample diagnosis results). This generated prior assessment data for tumor early screening targeted panels, ensuring that the data can provide a reliable benchmark reference for subsequent parallel discrimination assessment.

[0056] The tumor early screening targeted panel discrimination path data is processed by prior evaluation data of tumor early screening targeted panel to perform multi-target panel parallel tumor early screening discrimination evaluation, generating multi-target panel parallel tumor early screening discrimination evaluation data. In this embodiment of the invention, the evaluation process employs a parallel comparative evaluation algorithm. Using sample matching benchmark data and panel performance benchmark data from the prior evaluation data as references, the tumor early screening targeted panel discrimination path data is applied to the discrimination of valid prior samples. The discrimination paths corresponding to two panel configurations run in parallel, completing the tumor type discrimination for each sample respectively. Simultaneously, the discrimination results, discrimination time, core parameter matching status, and discrimination error for each path are recorded. The evaluation process sets three core evaluation indicators: discrimination accuracy, performance consistency, and path stability, with each indicator having clearly defined evaluation criteria. The accuracy assessment evaluates the degree of match between the tumor type results output by the discrimination pathway and the confirmed results of the prior samples. It is calculated as the number of accurately identified samples / the total number of evaluated samples. The accuracy of the discrimination pathways corresponding to the two panels must be ≥98%, and the overall accuracy of the combined pathway must be ≥98.5%. The efficacy consistency assessment evaluates the deviation between the detection sensitivity and specificity of the discrimination pathway after application and the prior efficacy benchmark data. The deviation must be ≤0.5% to ensure that the efficacy of the discrimination pathway is consistent with the clinical validation benchmark. The pathway stability assessment evaluates the fluctuations in discrimination time and parameter thresholds during parallel operation. The time fluctuation must be ≤0.1 seconds / sample, and the parameter threshold fluctuation must be ≤0.001 to ensure stable pathway operation. During the evaluation process, the evaluation indicators are verified one by one for each discrimination pathway, the evaluation results are recorded, and abnormal cases are identified. After the evaluation is completed, the specific values ​​of the three major evaluation indicators for each discrimination path, the deviation from the prior benchmark, the number of abnormal cases and the reasons for the abnormality are calculated to clarify the advantages and disadvantages of the discrimination path. At the same time, the synergistic efficacy of the two panels running in parallel is calculated to evaluate the complementarity of the combined discrimination paths. All evaluation data, abnormal case analysis results and efficacy comparison data are integrated to generate multi-target panel parallel tumor early screening discrimination evaluation data.

[0057] By using multi-target panel parallel tumor early screening discrimination evaluation data, the tumor early screening target panel configuration data is subjected to reverse dynamic iterative processing to generate dynamic tumor early screening target panel data.

[0058] In this embodiment of the invention, based on the deficiencies of the discrimination path and panel configuration identified by parallel evaluation, the parameters, target composition, and measurable boundaries of the panel configuration are adjusted and optimized in reverse. Through multiple iterations, dynamic optimization of the panel configuration is achieved, improving the discrimination efficacy, accuracy, and stability of the panel, thus meeting the core objective of dynamic optimization of targeted panels for early tumor screening in the system. The iterative processing employs a gradient descent iterative algorithm, with iteration termination conditions set as follows: panel combination discrimination accuracy ≥ 98.5%, efficacy deviation from prior benchmark ≤ 0.3%, and number of abnormal cases ≤ 3. The upper limit of the number of iterations is set to 5 to ensure iterative efficiency and optimization effect. The iterative process revolved around the problem of parallel evaluation data identification. The first iteration addressed the unreasonable parameter thresholds in the lung cancer-colorectal cancer panel discrimination path by adjusting the MGMT gene methylation coverage discrimination threshold from 61.0% to 60.8%, while simultaneously adjusting the discrimination parameter weights for colorectal cancer signals, increasing the methylation coverage weight from 0.45 to 0.48 to improve the discrimination accuracy of colorectal cancer signals. To address the insufficient fault tolerance in the gastric cancer panel discrimination path, the core target signal response threshold was lowered by 0.01, from ≥0.76 to ≥0.75, while the low-concentration signal identification algorithm was optimized to enhance the capture of weak signals. After the iteration, the two panel discrimination paths were reapplied to evaluate the prior samples in parallel to verify the optimization effect. After the first iteration, the combined discrimination accuracy improved to 98.2%, and the number of abnormal cases decreased to 4, failing to meet the termination criteria, thus entering the second iteration. The second iteration addressed the remaining four abnormal cases by adding an auxiliary target (a methylation site in the CDH1 gene promoter region) to the gastric cancer panel, improving target coverage. Simultaneously, the collaborative logic of the two parallel panels was adjusted to reduce misinterpretations of cross-constraint signals. The measurable logic space boundary of the lung cancer-colorectal cancer panel was optimized, and the iteration terminated when the conditions were met. After each iteration, the adjustments, their rationale, optimization effects, and evaluation data were recorded to generate dynamic early tumor screening targeted panel data.

[0059] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0060] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A dynamic optimization and identification analysis system for tumor early screening targeting panels, characterized in that, Includes the following modules: The tumor characterization signal analysis module is used to acquire multimodal tumor-related data pre-stored in the medical database; and to perform weighted fusion processing of tumor characterization signal mapping based on the multimodal tumor-related data to generate weighted tumor characterization signal mapping data. The tumor candidate target stability domain analysis module is used to perform tumor candidate target stability domain analysis based on weighted tumor characterization signal mapping data and generate tumor candidate target stability domain data. The tumor early screening targeted panel configuration analysis module is used to perform tumor early screening targeted panel configuration analysis and optimization processing on tumor candidate target stability domain data, and generate optimized tumor early screening targeted panel configuration data. The Panel constraint recognition tumor characterization signal analysis module is used to perform Panel constraint recognition tumor characterization signal feature analysis on weighted tumor characterization signal mapping data based on optimized tumor early screening targeted Panel configuration data, and generate Panel constraint recognition tumor characterization signal feature mapping data. The dynamic tumor early screening targeted panel establishment module is used to establish tumor early screening targeted panels by dynamically iterating the tumor early screening targeted panel configuration data based on panel constraint recognition tumor characterization signal feature mapping data, and generating dynamic tumor early screening targeted panel data.

2. The tumor early screening targeted panel dynamic optimization and identification analysis system according to claim 1, characterized in that, The tumor characterization signal analysis module includes the following functions: Retrieve pre-stored multimodal tumor-related data from medical databases; Based on multimodal tumor-related data, logical space mapping analysis of tumor-related signals is performed to generate multimodal tumor-related signal mapping data; Signal stability analysis was performed on multimodal tumor-related signal mapping data under different measurement conditions for each modality to generate multimodal tumor-related signal stability data; We perform contribution discriminant analysis on the multimodal tumor-related signal mapping data for each modality to generate multimodal tumor-related signal contribution discriminant data. Based on the stability data and discriminative contribution data of multimodal tumor-related signals, a joint steady-state and discriminative analysis of tumor-related signal attribute features is performed to generate tumor-related signal attribute feature data, and tumor-related signal attribute weight parameters are designed based on the tumor-related signal attribute feature data. The multimodal tumor-related signal mapping data is weighted and fused using tumor-related signal attribute weight parameters to generate weighted tumor characterization signal mapping data.

3. The tumor early screening targeted panel dynamic optimization and identification analysis system according to claim 2, characterized in that, The multimodal tumor-related data includes tumor type data, cfDNA mutation site data, methylation region distribution data, DNA fragment length statistics, tumor structural variation data, and peripheral blood background noise data.

4. The tumor early screening targeted panel dynamic optimization and identification analysis system according to claim 2, characterized in that, The logical space mapping analysis of tumor-related signals based on multimodal tumor-related data includes: Modal heterogeneity analysis is performed based on multimodal tumor-related data to generate tumor-related modal heterogeneity data; Multimodal tumor-related signal analysis is performed based on multimodal tumor-related data to generate multimodal tumor-related signal data; Tumor-related modal heterogeneity mapping rules are designed based on tumor-related modal heterogeneity relation data, and then used to perform logical space mapping processing on multimodal tumor-related signal data to generate multimodal tumor-related signal mapping data.

5. The tumor early screening targeted panel dynamic optimization and identification analysis system according to claim 1, characterized in that, The tumor candidate target stability domain analysis module includes the following functions: Multi-scale signal fluctuation decomposition processing is performed on weighted tumor characterization signal mapping data to generate multi-scale signal fluctuation data of tumor characterization. The amplitude-noise ratio relationship of tumor characterization signals is analyzed using multi-scale signal fluctuation data of tumor characterization, and tumor characterization signal amplitude-noise ratio data is generated. Based on the amplitude-noise ratio data of tumor characterization signals, sample screening and optimization of the steady-state interval of tumor characterization are performed to obtain optimized steady-state interval sample data of tumor characterization. Sample specificity analysis was performed on the optimized tumor characterization steady-state interval sample data to generate optimized tumor characterization steady-state interval sample specificity data; Based on the specific data of the optimized steady-state interval of tumor characterization samples, the candidate tumor characterization samples are clustered to generate clustered tumor characterization sample data. By clustering tumor characterization sample data, tumor candidate target stability domain analysis is performed to generate tumor candidate target stability domain data.

6. The tumor early screening targeted panel dynamic optimization and identification analysis system according to claim 5, characterized in that, The sample screening and optimization process for the steady-state interval of tumor characterization based on tumor characterization signal amplitude-noise ratio data includes: Based on the amplitude-noise ratio data of tumor characterization signals, sample screening processing of the steady-state interval of tumor characterization is performed to obtain sample data of the steady-state interval of tumor characterization. The potential redundancy relationship data of tumor sample signals in the steady-state interval of tumor characterization is analyzed, and the mutual information optimization processing of the steady-state interval samples of tumor characterization is performed through the potential redundancy relationship data of tumor sample signals to generate optimized steady-state interval sample data of tumor characterization.

7. The tumor early screening targeted panel dynamic optimization and identification analysis system according to claim 3, characterized in that, The tumor early screening targeted panel configuration analysis module includes the following functions: Based on the stable domain data of tumor candidate targets, the combined unit analysis of stable domains of tumor candidate targets is performed to generate combined unit data of stable domains of tumor candidate targets. Tumor type characteristics are analyzed by tumor type data to obtain tumor type characteristic data, and multi-target early screening targeted panel efficacy decision is designed based on tumor type characteristic data; Using the efficacy decision of multi-target early screening targeted panel, tumor candidate target stable domain combination unit data were used to perform tumor early screening targeted panel configuration analysis to generate tumor early screening targeted panel configuration data. The panel configuration of the tumor early screening targeted panel configuration data is optimized by performing risk scenario simulation, thereby generating optimized tumor early screening targeted panel configuration data.

8. The tumor early screening targeted panel dynamic optimization and identification analysis system according to claim 7, characterized in that, The analysis of tumor candidate target stability domain combination units based on tumor candidate target stability domain data includes: Perform stability domain feature analysis on tumor candidate target stability domain data to generate tumor candidate target stability domain feature data; Based on the characteristic data of the stable domain of tumor candidate targets, the adjacency relationship analysis of the stable domain of candidate targets is performed to generate the adjacency relationship data of the stable domain of tumor candidate targets. Perform coexistence feature analysis on the adjacency relationship data of stable domains of tumor candidate targets to generate coexistence feature data of stable domains of tumor candidate targets; Based on the characteristic data of the stable domain of tumor candidate targets and the coexistence characteristic data of the stable domain of tumor candidate targets, the combined unit analysis of the stable domain of tumor candidate targets is performed to generate the combined unit data of the stable domain of tumor candidate targets.

9. The tumor early screening targeted panel dynamic optimization and identification analysis system according to claim 1, characterized in that, The Panel-constrained tumor characterization signal analysis module includes the following functions: Perform panel configuration measurable feature boundary analysis on optimized tumor early screening targeted panel configuration data to generate panel configuration measurable feature boundary data. The panel configuration measurable feature boundary data is transmitted to the weighted tumor characterization signal mapping data to perform preliminary tumor characterization signal feature analysis of the panel configuration measurable boundary, and generate preliminary constrained tumor characterization signal feature mapping data. Based on the measurable feature boundary data of the panel configuration, the feature mapping data of the preliminary constrained tumor characterization signal are analyzed to generate the feature response difference data of the tumor characterization signal of the panel combination. Based on the difference data of the characteristic response of the panel combined tumor characterization signal, equivalent correction processing of the characteristic response of the panel combined tumor characterization signal is performed to generate corrected characteristic response data of the panel combined tumor characterization signal. Based on the corrected panel combined tumor characterization signal feature response data, panel constraint recognition tumor characterization signal feature analysis is performed to generate panel constraint recognition tumor characterization signal feature mapping data.

10. The tumor early screening targeted panel dynamic optimization and identification analysis system according to claim 1, characterized in that, The dynamic tumor early screening targeted panel establishment module includes the following functions: Analyze the discrimination path data of the targeted panel for early tumor screening based on the optimized configuration data of the panel and the feature mapping data of the panel constraint recognition of tumor characterization signals. Obtain prior assessment data from the targeted panel for early cancer screening; The tumor early screening targeted panel discrimination path data is processed by prior evaluation data of tumor early screening targeted panel to perform multi-target panel parallel tumor early screening discrimination evaluation, generating multi-target panel parallel tumor early screening discrimination evaluation data. By using multi-target panel parallel tumor early screening discrimination evaluation data, the tumor early screening target panel configuration data is subjected to reverse dynamic iterative processing to generate dynamic tumor early screening target panel data.

Images