Gastric cancer chemotherapy benefit analysis method and system based on multi-index calculation

Abstract

The invention relates to the field of chemotherapy benefit analysis, in particular to a gastric cancer chemotherapy benefit analysis method and system based on multi-index calculation. The method comprises the following steps: extracting multi-temporal biomarkers of a patient in different treatment cycles, and carrying out nonlinear change trend analysis to obtain trend characteristics of the biomarkers; collecting drug monitoring parameters in a chemotherapy cycle, performing drug metabolism simulation according to the trend characteristics of the biomarkers, and constructing a drug synchronous kinetic model; the method comprises the following steps: extracting vital sign information and chemotherapy toxic reaction information of a patient based on clinical detection, performing chemotherapy safety boundary prediction, and generating chemotherapy tolerance assessment data; and performing chemotherapy benefit hierarchical calculation based on the chemotherapy tolerance evaluation data and the drug synchronous kinetic model to generate multiple levels of chemotherapy benefit indexes. According to the method, accurate and efficient gastric cancer chemotherapy benefit calculation is realized, and the cost performance and efficiency of treatment are improved.

Description

Technical Field

[0001] This invention relates to the field of chemotherapy benefit analysis, and in particular to a method and system for analyzing the benefit of gastric cancer chemotherapy based on multi-index calculation. Background Technology

[0002] With the continuous advancement of medical technology and the ongoing development of precision medicine, the treatment model for gastric cancer, a globally prevalent malignant tumor of the digestive system, is gradually shifting from "experience-driven" to "data-driven" and "personalized precision intervention." Chemotherapy, as one of the core treatment methods for gastric cancer, plays a crucial role in preoperative neoadjuvant therapy, postoperative adjuvant therapy, and treatment of advanced metastases. However, due to significant individual differences among gastric cancer patients in terms of genetic background, physical condition, tumor molecular characteristics, and drug response, traditional "one-size-fits-all" chemotherapy regimens often fail to achieve the expected therapeutic effects and may even cause severe toxic side effects, impacting patients' quality of life and treatment adherence.

[0003] In clinical practice, the main criteria for evaluating the effectiveness of chemotherapy still focus on single or static indicators such as tumor shrinkage rate, imaging changes, and tumor marker levels. This assessment method is difficult to comprehensively reflect the dynamic response process of patients to treatment. At the same time, the lack of systematic modeling of drug metabolism behavior, toxicity tolerance boundaries, and biomarker trends also makes the adjustment of current chemotherapy regimens dependent on physician experience, making it difficult to achieve quantitative management and intelligent decision-making assistance. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a method and system for analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation, thereby resolving at least one of the aforementioned technical problems.

[0005] To achieve the above objectives, this invention provides a method for analyzing the efficacy of chemotherapy for gastric cancer based on multi-index calculation, comprising the following steps: Step S1: Extract multi-temporal biomarkers from patients at different treatment cycles, perform nonlinear trend analysis, and obtain biomarker trend characteristics; Step S2: Collect drug monitoring parameters during the chemotherapy cycle, perform drug metabolism simulation based on biomarker trend characteristics, and construct a drug synchronous kinetic model; Step S3: Based on clinical tests, extract patients' vital signs and chemotherapy toxicity information, predict the safety boundary of chemotherapy, and generate chemotherapy tolerance assessment data; Step S4: Based on chemotherapy tolerance assessment data and pharmacokinetic model, perform chemotherapy benefit stratification calculation to generate multiple levels of chemotherapy benefit index.

[0006] This specification provides a system for analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation, used to perform the gastric cancer chemotherapy efficacy analysis method based on multi-index calculation as described above, including: The trend analysis module is used to extract multi-temporal biomarkers from patients at different treatment cycles, perform non-linear trend analysis, and obtain biomarker trend characteristics. The drug metabolism simulation module is used to collect drug monitoring parameters during chemotherapy cycles, perform drug metabolism simulation based on biomarker trend characteristics, and construct a drug synchronous kinetic model. The chemotherapy tolerance assessment module is used to extract patients' vital signs and chemotherapy toxicity information based on clinical tests, predict the safety boundaries of chemotherapy, and generate chemotherapy tolerance assessment data. The benefit stratification calculation module is used to perform chemotherapy benefit stratification calculations based on chemotherapy tolerance assessment data and drug kinetic models, generating chemotherapy benefit indices at multiple levels.

[0007] The specific benefits of this invention are as follows: By analyzing the nonlinear changes in biomarkers (such as CEA, CA19-9, and blood indicators) at different time phases during multiple treatment cycles, the dynamic trends of tumor biological behavior can be revealed. Extraction of trend features helps to identify efficacy inflection points or drug resistance signals early, providing important clues for clinical decision-making and avoiding the continuation of ineffective treatment. Compared to static detection values, trend features better reflect inter-individual heterogeneity, providing highly discriminative input features for subsequent pharmacokinetic simulation and benefit assessment. Integrating drug concentration monitoring data with biomarker trend features enables personalized simulation of drug behavior in vivo (absorption, distribution, metabolism, and clearance), constructing patient-specific pharmacokinetic curves. Synchronous drug pharmacokinetic models can be used to optimize dosing time and dosage, improving treatment efficacy while reducing the risk of adverse reactions. Real-time simulation of the patient's response to drugs provides technical support for developing flexible treatment strategies, embodying the core concept of precision medicine. By modeling and analyzing vital signs (such as heart rate, body temperature, and blood pressure) and toxic reactions (such as bone marrow suppression, gastrointestinal reactions, and abnormal liver and kidney function), the tolerance boundary of individual chemotherapy can be quantified, and its safety threshold can be assessed. This step can predict high-risk patients in advance, prevent serious toxic events, and improve treatment safety. Tolerance assessment data can provide a basis for patient stratification, allowing those with high tolerance and those with low tolerance to receive different intensities of treatment, improving the overall cost-effectiveness of treatment. By fusing pharmacokinetic and tolerance data, benefit indices across multiple dimensions (such as efficacy, risk, and persistence) can be calculated, providing a basis for comprehensively judging the value of treatment. Based on different levels of chemotherapy benefit indices, patients can be divided into high-benefit, intermediate-benefit, and low-benefit groups, facilitating the development of differentiated treatment strategies by physicians. This stratification system can serve as an important decision-making tool for clinical pathway optimization and treatment resource allocation, especially suitable for highly heterogeneous diseases such as gastric cancer, improving the cost-effectiveness and system efficiency of treatment. Attached Figure Description

[0008] Figure 1 This is a schematic diagram of the steps in the present invention for a method of analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation; Figure 2 This is a detailed flowchart illustrating the implementation steps of step S1. Figure 3 This is a flowchart illustrating the detailed implementation steps of step S2. Detailed Implementation

[0009] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0010] This application provides a method and system for analyzing the efficacy of chemotherapy for gastric cancer based on multi-index calculation. The executing entities of the method and system for analyzing the efficacy of chemotherapy for gastric cancer based on multi-index calculation include, but are not limited to, mechanical equipment, data processing platforms, cloud server nodes, and network upload devices that can be considered as general computing nodes in this application. The data processing platform includes, but is not limited to, at least one of an audio-visual management system, an information management system, and a cloud-based data management system.

[0011] Please see Figures 1 to 3 This invention provides a method for analyzing the efficacy of chemotherapy for gastric cancer based on multi-index calculation, comprising the following steps: Step S1: Extract multi-temporal biomarkers from patients at different treatment cycles, perform nonlinear trend analysis, and obtain biomarker trend characteristics; Step S2: Collect drug monitoring parameters during the chemotherapy cycle, perform drug metabolism simulation based on biomarker trend characteristics, and construct a drug synchronous kinetic model; Step S3: Based on clinical tests, extract patients' vital signs and chemotherapy toxicity information, predict the safety boundary of chemotherapy, and generate chemotherapy tolerance assessment data; Step S4: Based on chemotherapy tolerance assessment data and pharmacokinetic model, perform chemotherapy benefit stratification calculation to generate multiple levels of chemotherapy benefit index.

[0012] In the embodiments of the present invention, see Figure 1 The diagram below illustrates the steps of a method for analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation according to the present invention. In this example, the steps of the method for analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation include: Step S1: Extract multi-temporal biomarkers from patients at different treatment cycles, perform nonlinear trend analysis, and obtain biomarker trend characteristics; In this embodiment, biological samples were collected at key points in each treatment cycle according to a predetermined follow-up plan (e.g., baseline, end of each treatment cycle, 24 hours and 72 hours after drug administration). Sample types included peripheral plasma, peripheral blood apheresis, and tumor puncture fluid when necessary. Circulating tumor DNA, several tumor-associated antigens, grouped cytokines, and several metabolites were quantitatively measured using high-throughput sequencing (targeted panel, sequencing depth example: 10,000× level to improve the detection of low-frequency mutations), multiplex immunoassay (Luminex or similar platform), and liquid chromatography-mass spectrometry (LC-MS / MS) for targeted metabolite detection. The obtained raw time series were first subjected to quality control: obvious technical anomalies were removed, readings below the detection limit were filled according to predetermined rules, and uncertainty was recorded. Subsequently, each biomarker sequence was normalized (e.g., z-score or min-max), and a smoothed spline or generalized additive model (GAM) was used to fit the nonlinear time trend; the smoothing parameter was selected through cross-validation to avoid overfitting. After fitting, the first and second derivatives are calculated to extract kinetic features (rate of ascent / descent, peak time, inflection point, and steady-state duration), and a change point detection method is used to identify significant trend inflection points. The final output is a set of trend feature vectors for each biomarker for each patient (e.g., initial slope, peak time, maximum rate of descent, location of significant inflection points, and their confidence intervals). These features will serve as key inputs for subsequent pharmacokinetic modeling and efficacy prediction.

[0013] Step S2: Collect drug monitoring parameters during the chemotherapy cycle, perform drug metabolism simulation based on biomarker trend characteristics, and construct a drug synchronous kinetic model; In this embodiment, plasma and (if feasible) tumor stroma samples are collected at predetermined time windows during chemotherapy administration for quantitative analysis of the drug and its active metabolites. Examples of sampling time points include: before administration, 0.5 hours, 1 hour, 2 hours, 4 hours, 24 hours after administration, and selected points in subsequent days; if microdialysis is used, higher temporal resolution stroma concentration curves can be obtained. Quantitative methods primarily use LC-MS / MS, with calibration curves covering the required concentration range and internal standards input to ensure accuracy. Using the obtained concentration-time data and the trend features extracted in step S1, a coupled pharmacokinetic-pharmacodynamic model is established: first, at the individual level, clearance, volume of distribution, and absorption and elimination rates are estimated using a non-compartmental or simplified PBPK model; then, the delayed response characteristics of biomarkers are used as input to the pharmacodynamic (PD) model to construct a synchronous kinetic description of the drug-tissue-response (including three levels: plasma exposure, tissue accessibility, and biological effects). The model fitting employed a nonlinear mixed-effects model (NLME) to characterize inter-individual variability. Parameter estimation was performed using least squares or maximum likelihood methods, and robustness was validated using residual diagnosis, prediction calibration plots, and bootstrap. Model outputs included individualized time-series exposure predictions, drug-dose-response curve parameters, and biomarker response simulations under different dosing strategies, providing a dynamic basis for subsequent efficacy-safety balance and regimen optimization.

[0014] Step S3: Based on clinical tests, extract patients' vital signs and chemotherapy toxicity information, predict the safety boundary of chemotherapy, and generate chemotherapy tolerance assessment data; In this embodiment, vital signs (heart rate, blood pressure, body temperature, respiratory rate, blood oxygen saturation) and clinical chemotherapy toxicity data (hematopoietic suppression, abnormal liver and kidney function, gastrointestinal reactions, neurotoxicity, etc., recorded according to the general adverse event classification) and necessary laboratory indicators (complete blood count, liver and kidney function, electrolytes) are collected and structured. Time window analysis is applied to continuous vital signs to extract short-term and cumulative stress characteristics (such as heart rate variability, blood pressure fluctuation amplitude, and cumulative duration of hypoxia events); the frequency, highest severity, and recovery time of toxic events are calculated. These indicators are converted into tolerance-related sub-scores through a pre-designed scoring framework: chemotherapy tolerance index, organ reserve score, and recovery potential score, etc., with a trigger threshold set for each (e.g., probability of serious adverse events or decline in organ function). Risk prediction models (such as logistic regression or survival analysis models) are used to correlate these sub-scores with the probability of serious adverse events or the risk of irreversible organ damage and to estimate the probability of exceeding safety boundaries in future treatment courses. The final output is a dataset assessing patients' chemotherapy tolerance, including numerical tolerance indices, organ-level safety thresholds, and corresponding uncertainty intervals, providing a quantitative basis for developing feasible dosing constraints and monitoring plans.

[0015] Step S4: Based on chemotherapy tolerance assessment data and pharmacokinetic model, perform chemotherapy benefit stratification calculation to generate multiple levels of chemotherapy benefit index.

[0016] In this embodiment, a multi-indicator benefit evaluation framework is constructed by combining the output of the individualized pharmacokinetic model (exposure-response prediction) obtained in step S2 with the tolerance boundary provided in step S3. First, benefit and safety are represented by a multi-dimensional objective vector: tumor response-related items (such as predicted tumor burden reduction and PFS probability), immune benefit items (if applicable immune markers are upregulated), quality of life protection items, and organ safety items. For each candidate dosing strategy (dose and interval combination), a pharmacokinetic-pharmacodynamic model is used to simulate the temporal response of several future treatment cycles, while simultaneously assessing the probability of reaching the safety boundary during the simulation. The simulation results are normalized and scored according to preset weights or multi-objective ranking rules, and the strategies are divided into multiple benefit levels (e.g., high benefit / high risk, moderate benefit / moderate risk, conservative strategy, etc.) based on quantile or clustering methods. To ensure the clinical usability of the protocol, a typical representative protocol, corresponding expected benefit indicators and risk probabilities, and recommended monitoring strategies are output for each level. This stratified chemotherapy efficacy index not only supports optimized decision-making for individuals, but also facilitates regimen comparison and resource allocation at the group level, ultimately forming an operational multi-level chemotherapy recommendation catalog with uncertainty assessment and sensitivity analysis results.

[0017] In this embodiment, see Figure 2 The diagram below illustrates the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Multi-temporal biomarkers of patients at different treatment cycles were extracted using a high-throughput gene sequencing platform, including circulating tumor DNA concentration, tumor-associated antigens, cytokine lineages, and metabolites. Dynamic concentration changes of the multi-temporal biomarkers are calculated to obtain time-series data streams of individual biomarkers. The time-series data stream of individual biomarkers is normalized to generate dynamic evolution curves; Spline interpolation was used to smooth the dynamic evolution curves, and nonlinear trend analysis was performed to obtain the trend characteristics of biomarkers.

[0018] In this embodiment, peripheral blood samples are collected from subjects according to a predetermined follow-up schedule at specific time points in the treatment cycle design to obtain multi-phase samples. Sampling time points typically include baseline (T0), the end point of each treatment cycle (e.g., T1 at the end of the first treatment cycle, T2 at the end of the second treatment cycle), and necessary intermediate time points (e.g., 24 hours and 72 hours after drug administration). Each time, 10–20 mL of peripheral venous blood is collected and stored in tubes containing an anticoagulant (EDTA). After sampling, plasma should be obtained by centrifugation within 2 hours and immediately frozen. 80 °C is used to reduce the degradation rate of nucleic acids and metabolites. Circulating tumor DNA (ctDNA) is sequenced using a targeted capture panel with high-throughput sequencing. The sequencing depth target is typically set to 10,000× to 100,000× to improve the detection sensitivity for low-frequency variants. Key indicators measured include variant allele frequency (VAF%) and absolute copy number (copies / mL). Tumor-associated antigens are analyzed using immunoassay (ELISA or chemiluminescence) to obtain concentrations (ng / mL or U / mL). Cytokine profiling is performed using multiplex immunoassay chips or the Luminex platform to simultaneously quantify several cytokines (pg / mL). Metabolites are detected using LC-MS / MS for targeted or non-targeted analysis, and peak area or relative abundance is output. To ensure traceability and data quality, each sample must be recorded with batch number, sampling timestamp, personnel identification, and pretreatment method (e.g., whether protein precipitation or metabolic arrest treatment was performed). Internal controls or exogenous standards are added before sequencing to monitor batch deviations. The output of the test is a raw quantitative matrix for each time point, along with measurement uncertainty indicators (such as CV%, LOD / LOQ values) and quality control sample readings, providing a raw numerical basis and quality assurance information for subsequent dynamic change calculations and sequence construction. After obtaining the raw quantitative values ​​for each time point, the discrete points need to be converted into a time-series data stream with strict time identification and comparability. First, the dimensions of each biomarker are standardized: ctDNA can be expressed as VAF% or normalized to copies / mL; tumor antigens and cytokines maintain their commonly used concentration units; metabolites are converted to absolute content (nmol / mL) where possible; if absolute conversion is not possible, peak area is retained and internal control is recorded. Each measurement must be accompanied by a precise timestamp (year-month-day-hour:minute:second) to support accurate time alignment; if multiple samplings occur within a treatment course, all raw readings can be retained in the data table and the mean and standard error within that time window can be calculated, or the most recent measurement point can be selected to represent that treatment course according to the study design. Values ​​below the limit of detection (LOD) or the limit of quantitation (LOQ) must be clearly labeled. A common practice is to fill in the gaps using LOD / √2 or by selecting the lowest half of the detectable values, and record the filling method in the metadata for subsequent sensitivity analysis. To handle batch effects, the sampling process recommends simultaneously inserting pooled QC samples and calibrators, recording these QC readings alongside the sample readings for subsequent batch calibration. The final output is a multi-channel time-series matrix arranged in ascending order of time, with each row corresponding to the sampling time and each column corresponding to the marker, plus metadata for each cell including measurement error, whether it is below the LOD, and the sample batch number. This matrix lays the time-series foundation for subsequent normalization and trend modeling, and supports missing value handling and interpolation strategy selection.

[0019] To facilitate comparisons across biomarkers and time points, multi-level normalization and batch correction are required for the original time series matrix. For sequencing count data, sequencing depth correction methods (such as RPM / CPM or standardization procedures more suitable for low-frequency variations) are first applied, and VAF is converted to copies / mL when necessary to account for biological significance. For metabolite and cytokine data, QC-based normalization is preferred, using the median of pooled QC or internal reference substances as the normalization benchmark. Subsequently, batch effect correction algorithms (such as the ComBat method based on position adjustment or Bayesian framework) are applied to the entire batch of samples to reduce systematic bias between different sequencing or detection batches. After normalization, each biomarker time series is scaled for comparison within the same frame. z-score normalization (subtracting the mean and dividing by the standard deviation) or min–max normalization can be used to map the values ​​to the [0,1] interval, and measurement errors are simultaneously converted to form standardized confidence intervals. If the time-series sampling intervals are unequal, the sampling interval information must be retained in the data table, and time difference weighting or interpolation strategies (such as linear or spline interpolation) should be used when plotting the dynamic evolution curves. This step outputs a standardized dynamic evolution curve for each individual and each marker, including time coordinates, standardized values, confidence intervals, whether it is an interpolation point, and batch correction indicators. These normalized curves facilitate visualization and serve as direct input for subsequent smoothing and trend analysis, while also providing a common-scale input matrix for multi-marker joint modeling. After obtaining the standardized dynamic evolution curves, spline interpolation and nonlinear modeling are used to denoise and extract key dynamic features. First, a smoothing spline (e.g., cubic spline or smoothing spline) is used to fit the discrete standardized points. The smoothing penalty coefficient (λ) is optimized through cross-validation (e.g., k-fold or leave-one-out method) to achieve a balance between avoiding overfitting and preserving key inflection points. If time points are sparse, it is recommended to increase the penalty strength or use a piecewise linear approximation to avoid overinterpreting single-point fluctuations; if time points are dense, more flexible splines can be used to capture short-term rapid changes. After fitting, the first and second derivatives of the smoothed curve are calculated to identify the rate of increase / decrease and inflection point. Change-point detection or significance testing based on spline derivatives is then used to determine significant trend segments. To quantify trend characteristics, several key indicators should be extracted and recorded: initial slope (units / day or units / treatment course), maximum rate of increase / decrease, peak time point, time constant required to reach steady state, and curve curvature or fluctuation amplitude. Simultaneously, confidence intervals for each indicator are calculated to assess statistical stability. For the analysis of the combined behavior of multiple biomarkers, cross-correlation and dynamic time warping (DTW) techniques can be used to assess the time lag relationships and synchronicity between different biomarkers.The final output is a smooth dynamic curve for each individual and each biomarker, a trend label (e.g., significant decrease, significant increase, oscillation, or plateau), and a set of structured trend feature vectors. These features can be used for multi-index efficacy scoring, individual treatment response stratification, or as feature inputs for predictive models to assess chemotherapy efficacy or predict short-term clinical outcomes.

[0020] In this embodiment, see Figure 3 The diagram below illustrates the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Collect drug monitoring parameters during chemotherapy cycles; the drug monitoring parameters include serum drug concentration and tissue concentration of active drug metabolites. Based on the drug monitoring parameters, drug clearance rate, time to peak concentration, and area under the curve are calculated, and pharmacokinetic parameters are extracted. Based on the trend characteristics of biomarkers, the local environment of the tumor is analyzed to generate a tumor environment characterization. Coupled analysis was performed on pharmacokinetic parameters and tumor environment characterization, and the actual effective concentration distribution of the drug was mined to obtain the actual effective concentration distribution index. Drug metabolism simulation was performed on the actual effective concentration distribution index to construct a drug synchronous kinetic model.

[0021] In this embodiment, blood samples are collected at predetermined time points during each treatment cycle to determine the parent drug and its active metabolites in serum. Simultaneously, tumor tissue or interstitial fluid is collected when feasible to assess local exposure. Blood samples are separated using routine plasma separation and cryopreservation to preserve drug stability. Tissue exposure samples are obtained through puncture sampling or microdialysis. Quantitative analysis employs highly selective mass spectrometry or specific immunoassay methods. All measurements include internal standard calibration, quality control samples, and blank controls to ensure quantitative accuracy. Sample processing procedures and metadata such as batch information, sample timestamps, and sampler information must be fully recorded for subsequent data cleaning and batch correction. The output is a time-series concentration table: total serum drug concentration, metabolite concentration, and (if applicable) interstitial fluid concentration, along with methodological accuracy and repeatability assessment results, providing a reliable raw data foundation for pharmacokinetic calculations. Using concentration-time data as input, key pharmacokinetic parameters are calculated using non-compartmental (non-separate) analysis methods or simplified compartmental models. First, the concentration curves undergo quality checks and necessary low-value processing. Then, the area under the curve (AUC) is obtained using numerical integration methods to identify peak concentrations and their corresponding times (C_max, T_max). Elimination rates and half-lives are estimated based on the elimination phase. Clearance and volume of distribution can be inferred from the relationship between administration route and AUC, and inter-individual variability can be analyzed using a population pharmacokinetic framework when necessary. If interstitial concentration data are available, the tissue / plasma ratio is calculated in parallel to estimate tissue accessibility. All parameters include uncertainty assessments and fit diagnostics for subsequent coupling and simulation. The method selection ensures both analytical robustness and avoids introducing excessive unnecessary complexity into the initial modeling.

[0022] Previously obtained biomarker temporal trends were used to infer multidimensional characteristics of the local tumor environment. A local environment description matrix was constructed based on representative indicators reflecting immune activity, metabolic state, and vascular permeability (e.g., immune-related cytokines, metabolite ratios, and vascular-related markers) and their temporal dynamics. Multiple biomarkers were integrated into several interpretable dimensions—e.g., immune activation, metabolic acidification, and blood flow / permeability index—using regularization or data-driven methods (e.g., weighted synthesis or principal component analysis), and each dimension was normalized for comparison. Local validation of the characterization was performed using microdialysis or imaging data when necessary. The resulting tumor environment characterization reflects temporal evolution and can be used to indicate microenvironmental factors influencing drug distribution and action, providing environmental correction factors for coupled pharmacokinetic analysis. Plasma and tissue concentration information was coupled with the tumor environment characterization to estimate the spatial-temporal exposure distribution of drugs actually exerting their effects within the tumor microenvironment. Methodologically, a multi-compartment or simplified physiological model is employed. Environment-dependent flux and diffusion correction factors (provided by the characterization matrix) are introduced into the tumor region. The measured tissue / plasma ratio is correlated with environmental factors using regression or machine learning to infer tissue exposure at unmeasured sites. An actual effective concentration distribution index is defined as a time-weighted indicator of local effective concentration. Weights for key microenvironmental regions are incorporated into the index construction to highlight exposure levels in highly sensitive areas such as low pH or hypoxia. The impact of population differences on the index is assessed using Monte Carlo or similar uncertainty analyses, resulting in an effective exposure metric that can be used to compare different dosage regimens or patient groups. Using individualized exposure information and pharmacokinetic parameters obtained from coupling analysis as input, a synchronous kinetic model incorporating key compartments such as plasma and tumor stroma is constructed to simulate the temporal evolution of the drug in the overall and microenvironmental contexts. The model can employ a simplified PBPK or multi-compartment form, incorporating metabolic and binding kinetic terms in the tumor region to reflect local clearance and retention. Parameter fitting employs numerical optimization or nonlinear mixed-effects methods to simultaneously explain longitudinal individual data and population differences. Model validation assesses prediction errors and adjusts model structure or parameters by comparing predicted tissue concentration curves with actual measured microdialysis or biopsy data. After fitting, scenario simulations—such as tissue exposure prediction under different dosages, formulations, or dosing intervals—can be performed to provide quantitative evidence for treatment regimen optimization and individualized dosing recommendations. The final output includes an individualized kinetic model, effective concentration distribution prediction, and uncertainty assessment report, providing kinetic support for multi-index-based chemotherapy benefit analysis.

[0023] In this embodiment, step S3 includes the following steps: Based on clinical testing, extract patients' vital signs information and chemotherapy toxicity information; Based on information on chemotherapy toxicity, the chemotherapy tolerance index, reserve function score, and recovery potential score are calculated to obtain the tolerance index. Chemotherapy stress load analysis was performed based on the patient's vital signs information to generate chemotherapy stress load values; The physiological stress response capacity was assessed based on the chemotherapy stress load value, and a physiological stress response capacity curve was generated. Predict the safety boundary of chemotherapy by analyzing physiological stress response curves and tolerance indices, and generate chemotherapy tolerance assessment data. In this embodiment, vital signs and toxicity data for subsequent assessment are extracted in a structured manner from clinical monitoring and routine testing. Vital signs data include heart rate, blood pressure (systolic / diastolic), respiratory rate, body temperature, oxygen saturation (SpO2), and continuous ECG (single-lead or multi-lead) event recordings. The recommended sampling frequency is continuous monitoring or at least once per minute for detecting suspected short-term changes. Resting measurements should also be recorded at outpatient / inpatient locations for cross-day comparisons. Chemotherapy toxicity information is derived from standardized clinical scores and laboratory indicators: gastrointestinal toxicity, hematopoietic toxicity (neutrophils, platelets, hemoglobin), abnormal liver and kidney function (ALT / AST, total bilirubin, creatinine, eGFR), neurotoxicity, and skin reactions are recorded according to the general adverse event classification (clinical symptom score), as well as patient subjective evaluations (pain, nausea scores, fatigue scale). Each sampling should record the timestamp, measuring device identification, measuring posture (supine / sitting position), medication time, and dosage information. The raw data undergoes consistency verification: duplicate records are removed, missing timestamps are added, abnormal measurements are labeled (such as lead detachment or sample hemolysis), and quality control notes are recorded. The output is a time-series dataset—a continuous stream of vital signs and a table of toxicity grading and test values ​​organized by visit, where each record includes the measurement source and confidence level label, providing a traceable input basis for subsequent tolerability and stress analysis. A multidimensional tolerability assessment vector is constructed based on clinical toxicity manifestations and organ function indicators. First, three sub-scores are defined: First, the Chemotherapy Tolerance Index (TI), primarily based on the severity and frequency of toxicity, quantified according to toxicity grade (e.g., 0–4 points) and weighted by organ importance; second, the Organ Reserve Function Score (RFS), using baseline functional values ​​of major organs such as the heart, lungs, kidneys, liver, and hematopoietic system as input (e.g., cardiac ejection fraction, eGFR, liver function enzymes, baseline neutrophil count), assigning weights to each organ according to clinical reversibility and replacement capacity, and normalizing them; third, the Recovery Potential Score (RP), which integrates data from age, physiological status indicators (e.g., body mass index, serum albumin), past comorbidities, and performance evaluation scales (e.g., activity level or frailty score), reflecting the patient's ability to recover from toxic events. Each sub-score is first standardized in scale (e.g., z-score or normalized to the 0–1 range), and then a weighted composite method is used to obtain the Overall Tolerance Index (ATI). The weights can be based on clinical expert consensus or obtained through regression / machine learning training using historical cohorts. The calculation process takes into account the toxicity time window: recent high-grade toxicity should result in a larger deduction; if an organ index is in the reversible range, it will be reflected as a moderate deduction in the RFS. The output is a numerical tolerance index value and decomposition of each sub-score, accompanied by uncertainty estimates and explanations of key contribution factors, which facilitates clinical judgment on whether to adjust the dosing intensity or initiate protective intervention.

[0024] Continuous vital signs and events are mapped to indicators representing the body's immediate and cumulative stress load. Time window analysis (e.g., a short window of 1–5 minutes for acute responses and a long window of 24–72 hours for cumulative load) is used to extract features from heart rate variability (HRV), blood pressure fluctuation amplitude, duration of abnormal body temperature, number of hypoxic events, and abnormal ECG events (such as premature ventricular contractions and ventricular tachycardia). Temporal (SDNN, RMSSD) and frequency domain indices of heart rate variability reflect sympathetic / parasympathetic balance, short-term fluctuations in blood pressure and respiratory parameters reflect circulatory and compensatory capacity, and repeated or continuous physiological abnormalities are weighted and summed to generate a cumulative stress score. These features are weighted according to physiological importance and clinical relevance, and then integrated or exponentially decayed over time to reflect the stronger impact of recent events (e.g., exponential weight decay coefficient). Furthermore, the timing of chemotherapy dosing peaks and troughs is used to correlate abnormal vital signs with the drug exposure window to determine whether the physiological stress is pharmacologically related. The final output is the "Chemotherapy Stress Load Value" (CSL), which provides a numerical description of the stress curve within a single dosing cycle and across cycles, including its components (cardiovascular load, respiratory load, metabolic thermal stress, etc.), for subsequent safety boundary assessment and individualized monitoring decisions. Using the tolerance index and stress load time series as inputs, the patient's physiological coping ability under different load levels is assessed, and response curves are constructed. Specifically, a dose-response framework is used: the chemotherapy stress load value is considered the "impact," and the tolerance index is considered the "buffer capacity." By simulating short-term and long-term load inputs, the output physiological response (e.g., the magnitude of heart rate increase, blood pressure decrease time, and oxygen saturation recovery time) is calculated. On the time axis, the instantaneous recovery rate and maximum deviation are calculated for each impact event, and a curve representing the recovery ability per unit load—the Physiological Stress Rating (PSR) curve—is constructed. The horizontal axis represents stress intensity or energy (CSL value), and the vertical axis represents the corresponding recovery efficiency or conservatism index. To quantify reversibility and the threshold effect, nonlinear fitting (e.g., sigmoid or exponential decay models) is used to characterize the transition from complete compensation to compensation failure and to estimate the critical load point—the point after which the recovery rate significantly decreases. Risk confidence intervals can also be embedded in the curve to reflect measurement and individual variability uncertainties. PSR curves allow clinicians to visually observe patients' immediate responses and recovery potential under different intensities of stress, thus providing a basis for determining the need for dosing rhythm, monitoring frequency, or supportive care.

[0025] This study combines physiological stress capacity curves with tolerance indices to predict the safety boundary in future dosing scenarios. Using Monte Carlo simulation or scenario modeling, several dosing intensities / frequencies and potential concomitant event sequences are set to simulate the stress load trajectory of patients under these scenarios, and the corresponding recovery and failure probabilities are calculated using PSR curves. The safety boundary is defined as the maximum permissible stress load or dosing intensity that keeps the recovery probability above a preset threshold (e.g., ≥95% without severe toxic events in the short term). The output is structured chemotherapy tolerance assessment data, including recommended maximum single dose or adjacent dosing intervals, monitoring recommendations (e.g., when to provide enhanced continuous monitoring of vital signs or laboratory retesting), warning thresholds, and corresponding trigger actions (e.g., initiating supportive medication when CSL exceeds the warning value). Furthermore, the assessment data includes individualized risk maps (which time windows or types of stress are most likely to lead to boundary violations) and uncertainty quantification, facilitating evidence-based adjustments by clinicians when balancing efficacy and safety. The final output document can be used for individual clinical decisions or aggregated into cohort-level strategy guidance, supporting individualized optimization of chemotherapy regimens and dynamic risk management.

[0026] In this embodiment, step S4 includes the following steps: Establish a multi-indicator benchmark for evaluating the efficacy of chemotherapy, including the tumor response index, immune benefit index, quality of life protection index, and organ safety index; The drug synergistic model was used to evaluate the maximum drug effect in patients and extract the theoretical maximum efficacy value. Risk-efficacy benefit balance is calculated based on chemotherapy tolerance assessment data, and multi-objective optimization is performed to generate risk-efficacy benefit balance parameters. Based on the multi-index chemotherapy benefit evaluation benchmark, the chemotherapy benefit is calculated in a stratified manner using the risk-efficacy benefit balance parameter and the theoretical maximum efficacy value, generating a chemotherapy benefit index at multiple levels. The chemotherapy efficacy index is used to predict future chemotherapy treatment course index changes, generating an index prediction trajectory. Dynamic trajectory planning for chemotherapy regimens based on exponential prediction trajectories.

[0027] In this embodiment, the target domain for evaluation is first clearly defined, and a set of quantifiable indicators is defined for each dimension. The tumor response index uses the rate of change in tumor volume on imaging, the percentage decrease in tumor markers, and progression-free survival (PFS) as inputs; the immune benefit index uses the magnitude of cytokine upregulation, changes in tumor-infiltrating lymphocyte (TIL) density, and the benefit ratio of immune-related adverse reactions as inputs; the quality of life protection index uses patient-reported outcome scales (e.g., fatigue scale, pain score, functional score) and the frequency of hospitalization / supportive care needs as inputs; and the organ safety index uses changes in key organ function (e.g., liver and kidney function indicators, hematopoietic function values) and the incidence of serious adverse events as inputs. Each raw indicator is first scaled (e.g., normalized to the 0–1 range or z-score standardized), and a clinically significant threshold is defined for each item (e.g., a tumor volume decrease of >30% is considered a partial response contribution, and CTCAE ≥ grade 3 adverse events are considered safety deductions). In the indicator construction phase, a two-step approach of expert weighting and data-driven methods was adopted in parallel: initial weight allocation was given based on clinical expert consensus, while existing cohort data were used to assess the explanatory power of each indicator on the final outcome (overall survival, overall quality of life score, etc.) through multiple regression or principal component analysis (PCA) to adjust the weights. Finally, the four dimensions were unified into a multi-indicator benefit benchmark framework, forming a quantitative target vector that can be directly used to optimize the objective function, providing a clear and comparable metric for subsequent risk-benefit balancing. Based on individualized pharmacokinetic / pharmacodynamic models (including plasma and tumor stroma exposure, drug-target binding kinetics, and drug sensitivity response curves), the theoretically maximized efficacy scenario was solved under given tolerability constraints. First, the pharmacodynamic function was defined with model inputs (including individual AUC, peak-to-trough ratio, and tumor sensitivity parameters), for example, the pharmacodynamic response to tumor exposure exhibited an S-shaped response curve. Numerical scanning or optimization searches are performed on drug dosages and dosing intervals within permissible ranges (e.g., upper limits for single and cumulative doses). Global optimization methods (such as grid search combined with local quasi-Newton methods, or Bayesian optimization) are employed to find dosing strategies that maximize a predefined benefit vector (primarily targeting tumor response, supplemented by an immune benefit term). To ensure robustness, Monte Carlo sampling (e.g., 1000 individual parameter perturbations) is used to assess the model's sensitivity to individual differences, and the optimal expected efficacy is recorded under conditions that do not violate safety thresholds (e.g., organ safety index must not fall below a certain minimum or the probability of serious adverse events <5%). The final output is the theoretically maximized efficacy value—that is, the highest benefit estimate achievable under an ideal but constrained pharmacokinetic / pharmacodynamic model—accompanied by an uncertainty interval and constraints explaining why this value cannot be directly achieved clinically (e.g., tolerability boundaries or high-risk event probabilities).

[0028] Using tolerance assessment data (including tolerance index, chemotherapy safety boundary, and physiological stress curve) as constraints, the efficacy benefits (quantified values ​​from step two or clinical benefit benchmarks) and risk costs (serious adverse event rate, probability of organ dysfunction decompensation, and decline in quality of life) are jointly incorporated into a multi-objective optimization framework. Multi-objective optimization methods (such as the Pareto-front-based non-dominated sorting genetic algorithm NSGA-II, or multi-objective Bayesian optimization) are employed to solve for a set of Pareto optimal solutions within a given dose space and dosing strategy set, outputting a series of risk-efficacy trade-offs. During the calculation, risks are quantified in the form of probabilities or expected costs (e.g., weighted by the expected frequency of serious adverse events or potential hospital stays), and benefits are weighted or stratified (e.g., prioritizing tumor response while considering quality of life). To ensure stability, the number of simulation repetitions is set (e.g., 500–1000 Monte Carlo simulations per candidate protocol), and population sensitivity analysis is performed on the results to identify key driving factors. The final result is a risk-benefit balance parameter set, including candidate dosage regimens, corresponding risk probabilities, expected efficacy indicators, and applicable patient subgroup labels, allowing clinical decision-makers to choose between safety and benefit. The balanced candidate regimens from step one and the theoretical upper limits from step two are stratified and scored using the multi-indicator benchmark constructed in step one. First, the predicted performance of each regimen across four dimensions is mapped to a unified scale, and a comprehensive benefit score is calculated according to predefined clinical priorities (e.g., tumor response has the highest weight, followed by organ safety). The obtained scores are stratified into several levels (e.g., high-benefit, intermediate-benefit, conservative, and experimental levels), and the regimens are categorized using thresholding or clustering methods (e.g., K-means). To enhance clinical usability, typical representative regimens and key parameters (e.g., recommended dose range, monitoring frequency, and suitable patient characteristics) are provided for each level, along with a risk-benefit distribution map of the regimens within each level. This stratified chemotherapy benefit index presents the relative value of different regimens and facilitates the transformation of complex multi-objective results into clinical pathway selection (e.g., implementing high-risk but high-benefit regimens only in patients with adequate monitoring). In addition, the report includes uncertainty measures and sensitivity notes to help understand the likelihood of hierarchical shifts due to individual differences or changes in model assumptions.

[0029] Based on established individualized kinetic / response models and historical treatment data, time-series forecasting is performed to estimate the evolution of the benefit index in subsequent treatment cycles. Mixture models or state-space models (e.g., nonlinear mixed-effects models with time coupling or Kalman / particle filters) are used to model the changes in the benefit index over treatment cycles. Inputs include dose, interval, immediate toxicity, and biomarker response for each cycle. Model fitting and forward simulation of future treatment scenarios generate a predicted trajectory of the benefit index over time with confidence intervals. Comparable outputs are provided for scenario planning, simulating the index trajectory under several dosing strategies (e.g., maintenance dose, dose escalation, or dose reduction / discontinuation), and key inflection points (such as the expected significant increase in benefit or accumulation of toxicity risk in the nth cycle) are visualized in time series. Uncertainty propagation and sensitivity analyses are performed simultaneously to determine which covariates (age, baseline tumor burden, liver and kidney function) have the greatest impact on the trajectory, thus providing a basis for individualized adjustments. Dynamic dosing pathway planning is developed. The plan employs Model Predictive Control (MPC) or decision tree-based strategy generation: given an objective (e.g., maximizing short-term tumor shrinkage or long-term survival probability while ensuring safety boundaries) and constraints (tolerance threshold, organ safety baseline), it calculates the optimal sequence (including dose adjustments, interval adjustments, and time points for concurrent supportive therapy) over several future treatment cycles. To enhance real-time adaptability, triggering rules and a closed-loop feedback mechanism are established: after each treatment cycle, the individual model is updated based on actual observed tolerance and biomarker responses, and the trajectory is re-predicted (the recommended update frequency is per treatment cycle or at more granular key time points). When the predicted trajectory indicates a potential breach of safety boundaries or failure to achieve expected benefits, alternative strategies are triggered (e.g., temporary dose reduction, extended intervals, or addition of adjuvant therapy). The output is a time-based dynamic treatment recommendation (dose per cycle, monitoring level, and adjustment trigger conditions), accompanied by a comparison chart of decision rationale and expected trajectory, facilitating continuous evaluation and necessary clinical adjustments by the clinical team during implementation, achieving a dynamic balance and individualized optimization between efficacy and safety.

[0030] In this embodiment, the specific steps for predicting future chemotherapy course index changes and generating an index prediction trajectory for the chemotherapy efficacy index are as follows: Multiple time-series autocorrelation analyses were performed on the trend characteristics of biomarkers to obtain the lag period of indicator changes. Calculate the response delay value at the time of chemotherapy administration based on the lag period of indicator changes, and construct a multi-indicator response delay spectrum; The chemotherapy efficacy index is time-delayed and corrected based on the multi-index response delay spectrum to obtain the time-delayed chemotherapy efficacy index. Based on the time-delayed modified chemotherapy efficacy index, the future chemotherapy course index changes are predicted, and an index prediction trajectory is generated.

[0031] In this embodiment, the standardized time-series curves of each biomarker are preprocessed: obvious outliers are removed, short-term missing data are filled using spline interpolation or linear interpolation (if the missing data exceeds 20% of the total sequence length, it is marked as a low-confidence sequence and processed separately), and detrending is performed to eliminate long-term drift (e.g., using local regression loss or differencing methods). For data sampled at equal intervals, the autocorrelation function (ACF) and partial autocorrelation function (PACF) can be calculated to identify inherent periodicity and typical lag order; the conventional test selects the maximum lag window as 1 / 3 of the sequence length or no more than 12 sampling points (e.g., when sampling weekly, a maximum lag of 12 weeks is considered), and the significance threshold (usually a 95% confidence interval, p < 0.05) is used to determine the autocorrelation peak. For data sampled at unequal intervals or irregularly, Lomb-Scargle spectral analysis or time-series interpolation is used before calculating the ACF to avoid spectral distortion caused by sampling intervals. To identify lag relationships between different biomarkers, the cross correlation function (CCF) can be calculated and multiple tests can be statistically corrected (e.g., Benjamini–Hochberg FDR correction). The output includes the lag period set for each biomarker (e.g., primary lag L1, secondary lag L2, and their significance p-values), the corresponding autocorrelation magnitude, and the confidence interval. These lag periods reflect the delayed characteristics of the biological response (e.g., a cytokine peaks 2–3 weeks after administration) and provide quantitative evidence for aligning the administration time point with the biological response timeline. After obtaining the lag periods for each indicator, they are mapped to specific administration time points to calculate the response delay value. The specific approach involves treating each dosing event as a time anchor point and using this anchor point to search backwards for the response peak or significant change interval of the corresponding lag in the biomarker sequence: if the main lag of a biomarker is L weeks, the response delay value of the corresponding dosing event is recorded as L weeks (or L sampling intervals). To improve robustness, a tolerance window (e.g., ±1 sampling interval) can be set near the main lag, and local maximum detection or derivative thresholding (e.g., the first derivative exceeds twice the standard deviation of the baseline) can be used to confirm the response time point. For biomarkers with multiple significant lags, the main lag and secondary lags are recorded, and a delay vector is established for each dosing event. To handle inter-individual differences or the cumulative effect caused by repeated dosing, a short-time window deconvolution method or baseline-corrected event-related averaging is used to separate the contributions of different dosing events. All dosing events and delay values ​​of all biomarkers are aggregated to form a "multi-indicator response delay spectrum" (a matrix structure: rows represent biomarkers, columns represent dosing events, and cells contain delay values ​​and confidence levels). The statistical characteristics of the spectrum (median delay, delay variance, and long-tail proportion) are calculated. This spectrum can reveal which biomarkers respond to dosing immediately, delayed, or cumulatively, providing direct input for subsequent time correction.

[0032] Time correction of the original chemotherapy benefit index using response delay spectra can be categorized into two types: shift correction and convolution correction. Shift correction shifts the contribution of a biomarker in the benefit calculation forward or backward according to its delay value, aligning the time window for benefit calculation with the biological response window. For example, if the main response delay of a tumor biomarker to a certain drug is 3 weeks, then when assessing the immediate contribution of the drug to the tumor response, the biomarker curve should be shifted backward by 3 weeks, or the benefit assessment time point should be postponed accordingly. Convolution correction, based on a set of response kernels, convolves the drug administration time series with the response kernel of each biomarker to obtain the predicted temporal response. The original benefit index is then reweighted based on these predicted responses. The response kernels can be parametric (e.g., exponential decay or gamma distribution) and their peak values ​​and widths can be fitted using the statistical characteristics of the delay spectrum. To ensure the statistical robustness of the corrected index, bootstrap or Monte Carlo simulations are used to assess the impact of time delay uncertainty on the benefit index, and the difference range and significance of the index before and after correction are reported. The final output is a time-delayed adjusted chemotherapy efficacy index, which reflects the efficacy estimate after considering the delay in biomarker response at each moment, more closely approximating the true causal time relationship, thus avoiding misjudging delayed responses as immediate efficacy or considering delayed benefits as ineffective. Using the time-delayed efficacy index sequence as model input, a time series model that can be used to predict future treatment courses is constructed. First, the stationarity of the adjusted sequence is tested and differencing is performed (if the sequence is non-stationary, differencing or a state-space model is used directly). Model selection can be based on data density and complexity: if the sampling is dense and the nonlinear characteristics are significant, an autoregressive moving average model with a delay term (ARIMAX) or a generalized additive model (GAM) is used; for scenarios with uncertain sampling or requiring real-time updates, a state-space model combined with Kalman filtering or particle filtering is used to achieve sequential updates and uncertainty estimation. Covariates included in the model should include planned dosing time and dose, known response delay distribution (as an exogenous delay term), and key clinical covariates of the patient (such as baseline tumor burden and organ function scores). Scenario simulations of future treatment courses are performed: different dosing strategies (such as maintenance dose, dose increment, or interval adjustment) are input, and the model outputs the corresponding time-delay-corrected benefit index prediction trajectory, quantifying the prediction uncertainty using confidence intervals or probability bands. To verify the reliability of the predictions, hindcasting is employed—the model is trained using data from the first n treatment courses and predicted for the (n+1)th treatment course. The predicted values ​​are compared with actual observations, and prediction error indices (MAE, RMSE) are calculated to adjust model parameters or select a more suitable model form. The final generated prediction trajectory can be used for clinical decision support: visual time-series charts demonstrate the benefit evolution under different strategies, helping to develop more reasonable dosing plans and treatment course adjustment schemes that take into account the delay in biological response.

[0033] In this embodiment, the specific steps for dynamic trajectory planning of chemotherapy regimens based on exponential prediction trajectories are as follows: Based on the exponential prediction trajectory, dynamic trajectory planning of chemotherapy regimens is performed to obtain chemotherapy adjustment plans; Based on the chemotherapy regimen adjustment, the patient's chemotherapy tolerance was reassessed to obtain the latest chemotherapy tolerance boundary; The latest chemotherapy treatment intervals and dosing time windows are calculated based on the latest chemotherapy tolerance boundaries, and real-time chemotherapy improvements are made.

[0034] In this embodiment, the previously obtained future chemotherapy cycle index prediction trajectory is used as input to generate specific dynamic dosing adjustment suggestions according to pre-set clinical goals and safety constraints. First, the key outputs of the prediction trajectory are quantified into actionable objective functions, such as maximizing short-term tumor shrinkage rate and prolonging progression-free survival, while chemotherapy tolerance boundaries, organ safety thresholds, and quality of life preservation are used as hard constraints. Model predictive control (MPC) or scenario-based simulation-based optimization processes are employed to solve the dynamic trajectory planning problem: within a rolling time window (e.g., the next 3 cycles or 12 weeks), the current individualized kinetic / response model is used to perform forward simulations of candidate dosing sequences (including single dose, dosing frequency, interval adjustments, and supportive medication timing), and a multi-index benefit-risk score is calculated for each candidate trajectory. The candidate space can be limited to a clinically feasible range; for example, single doses can be gradually adjusted within ±30% of the baseline dose, intervals can vary between 0.5 and 2 times the original interval, and the presence or absence of supportive medications (anti-nausea, hematopoietic growth factors, etc.) is considered as a binary option.

[0035] In the simulation of each candidate strategy, several outputs are calculated simultaneously: changes in the time-delayed chemotherapy benefit index, the peak value of the chemotherapy stress load curve, the probability of expected decline in tolerability, and the probability of organ safety indicators exceeding limits. Multi-objective evaluation rules are applied to these outputs (e.g., firstly ensuring that organ safety thresholds are not breached, secondly maximizing tumor response, and thirdly considering immune benefit and quality of life) to screen strategies that meet clinical priorities. Screening can use Pareto optimality screening or a weighted sum method. If the Pareto method is used, a set of compromise solutions is output; if a single recommendation is required, the highest-ranked solution is generated according to preset weights. The final chemotherapy adjustment plan should include: the recommended single dose for each subsequent cycle, the dosing interval or the number of days of delay, whether supportive therapy is used concurrently, when to retest biomarkers and the frequency of vital sign monitoring, and rules to trigger further adjustments (e.g., if the next biomarker decline is less than expected or the peak chemotherapy stress load exceeds a set threshold, then plan B is activated). The protocol also includes uncertainty assessment based on Monte Carlo simulation (e.g., efficacy and risk distribution under 500–1000 parameter perturbations) to help clinical decision-making grasp the potential range of variation and prepare response plans.

[0036] Before implementing or planning to implement adjustment protocols, a real-time or short-term tolerability reassessment of the patient's current status is conducted to obtain the latest tolerance boundaries for subsequent calculations. The reassessment primarily uses two types of data: recent biomarkers and pharmacokinetic observations (reflecting drug exposure and biological response) and the latest clinical vital signs and laboratory results (reflecting organ function and stress response). First, key input data are collected or updated: the latest complete blood count, liver and kidney function, ECG and vital sign time series, the latest round of biomarker time points, and, if necessary, plasma drug concentrations. Rapid quality control is performed on these data (e.g., whether testing intervals are met, whether there are significant measurement anomalies or hemolysis), and the new data is incorporated into the established individual model for short-term refitting (e.g., updating model parameters using Kalman filtering or incremental least squares) so that the model reflects the latest pharmacokinetic behavior and biological response sensitivity.

[0037] At the reassessment level, the Tolerance Index (ATI) and its sub-items (chemotherapy tolerance index, reserve function score, and recovery potential score) are recalculated, and the physiological stress response curve is locally reassessed to reflect recent stress-recovery behavior. For key organs (heart, liver, kidney, and hematopoietic system), the latest test values ​​are converted into organ function scores and compared with the previous baseline; if any organ function declines beyond a preset trigger threshold (e.g., eGFR decrease ≥30% or platelet count drops to a certain lower limit), the tolerance boundary is immediately downgraded. These calculations are combined to determine the numerical form of the latest chemotherapy tolerance boundary: including the maximum permissible dose, the upper limit of cumulative dose, the minimum safe interval, and the required level of monitoring within different dosing windows. The reassessment process should also provide uncertainty indicators; for example, if measurement errors or model fit are not ideal, a conservative range for the tolerance boundary is given to reduce the risk of misjudgment. The final output is a clinically readable "Latest Tolerance Boundary Report," clearly stating the permissible range, necessary monitoring measures, and specific rules for trigger adjustments, providing constraints for immediate chemotherapy improvements.

[0038] Using the latest tolerance boundary as a constraint, and combining pharmacokinetics, time-delayed benefit prediction, and current patient exposure data, the optimal interval for the next treatment cycle and the dosing window are calculated, and immediate chemotherapy improvements are implemented. First, the latest tolerance boundary is input into the dosing timing optimization model: this model prioritizes safety, with the objective function being to maintain or improve the time-delayed chemotherapy benefit index as much as possible within the safety boundary, while considering drug clearance and observed changes in pharmacokinetic parameters. Short-term optimization (single-step or multi-step prediction) is used to determine the optimal time for the next dose. For example, it may be suggested to postpone the next dose by several days to allow organ function recovery, or to reduce the single dose by a certain percentage and shorten / extend the interval to balance drug exposure and toxicity risks. Common operational rules include: when the tolerance boundary indicates insufficient hematopoietic reserves, extending the next dosing interval by 1–2 weeks and providing hematopoietic support during the interval; when pharmacokinetic monitoring shows decreased plasma clearance and increased risk of tissue accumulation, it is recommended to reduce the single dose by 20–30% and increase monitoring frequency.

[0039] In calculating the time window, the biological response delay profile must also be considered—if the main lag of a key biomarker indicates that efficacy may not be observed for several weeks after administration, conservative adjustments should be prioritized in the short term to avoid over-intensity administration before efficacy is observed. The calculation results simultaneously generate specific implementation guidelines: the earliest and latest times for the next administration (e.g., "recommended to administer on day X, with an allowable window of day X–2 to X+4"), recommended single dose and cumulative dose limits, recommendations for supporting medications, and the immediate monitoring to be performed (which tests must be performed before / after administration at which time points). The implementation of immediate chemotherapy improvements should be accompanied by a rapid feedback pathway: key indicators should be collected at set time points after administration and an initial assessment completed within 24–72 hours. If unexpected toxicity or breaches of tolerance boundaries occur, adjustments should be made immediately according to pre-defined backup plans (e.g., temporary discontinuation or further dose reduction). All changes and monitoring results must be recorded and included in long-term archives for subsequent model calibration and population learning, supporting more precise individualized dosing management in the future.

[0040] In this embodiment, a gastric cancer chemotherapy benefit analysis system based on multi-index calculation is provided, used to perform the gastric cancer chemotherapy benefit analysis method based on multi-index calculation as described above, including: The trend analysis module is used to extract multi-temporal biomarkers from patients at different treatment cycles, perform non-linear trend analysis, and obtain biomarker trend characteristics. The drug metabolism simulation module is used to collect drug monitoring parameters during chemotherapy cycles, perform drug metabolism simulation based on biomarker trend characteristics, and construct a drug synchronous kinetic model. The chemotherapy tolerance assessment module is used to extract patients' vital signs and chemotherapy toxicity information based on clinical tests, predict the safety boundaries of chemotherapy, and generate chemotherapy tolerance assessment data. The benefit stratification calculation module is used to perform chemotherapy benefit stratification calculations based on chemotherapy tolerance assessment data and drug kinetic models, generating chemotherapy benefit indices at multiple levels.

[0041] 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.

[0042] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein are 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 method for analyzing the efficacy of chemotherapy for gastric cancer based on multi-index calculation, characterized in that, Includes the following steps: Step S1: Extract multi-temporal biomarkers from patients at different treatment cycles, perform nonlinear trend analysis, and obtain biomarker trend characteristics; Step S2: Collect drug monitoring parameters during the chemotherapy cycle, perform drug metabolism simulation based on biomarker trend characteristics, and construct a drug synchronous kinetic model; Step S3: Based on clinical tests, extract patients' vital signs and chemotherapy toxicity information, predict the safety boundary of chemotherapy, and generate chemotherapy tolerance assessment data; Step S4: Based on chemotherapy tolerance assessment data and pharmacokinetic model, perform chemotherapy benefit stratification calculation to generate multiple levels of chemotherapy benefit index.

2. The method for analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation according to claim 1, characterized in that, The specific steps of step S1 are as follows: Multi-temporal biomarkers of patients at different treatment cycles were extracted using a high-throughput gene sequencing platform, including circulating tumor DNA concentration, tumor-associated antigens, cytokine lineages, and metabolites. Dynamic concentration changes of the multi-temporal biomarkers are calculated to obtain time-series data streams of individual biomarkers. The time-series data stream of individual biomarkers is normalized to generate dynamic evolution curves; Spline interpolation was used to smooth the dynamic evolution curves, and nonlinear trend analysis was performed to obtain the trend characteristics of biomarkers.

3. The method for analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation according to claim 1, characterized in that, The specific steps of step S2 are as follows: Collect drug monitoring parameters during chemotherapy cycles; the drug monitoring parameters include serum drug concentration and tissue concentration of active drug metabolites. Based on the drug monitoring parameters, drug clearance rate, time to peak concentration, and area under the curve are calculated, and pharmacokinetic parameters are extracted. Based on the trend characteristics of biomarkers, the local environment of the tumor is analyzed to generate a tumor environment characterization. Coupled analysis was performed on pharmacokinetic parameters and tumor environment characterization, and the actual effective concentration distribution of the drug was mined to obtain the actual effective concentration distribution index. Drug metabolism simulation was performed on the actual effective concentration distribution index to construct a drug synchronous kinetic model.

4. The method for analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation according to claim 1, characterized in that, Step S3 is as follows: Based on clinical testing, extract patients' vital signs information and chemotherapy toxicity information; Based on information on chemotherapy toxicity, the chemotherapy tolerance index, reserve function score, and recovery potential score are calculated to obtain the tolerance index. Chemotherapy stress load analysis was performed based on the patient's vital signs information to generate chemotherapy stress load values; The physiological stress response capacity was assessed based on the chemotherapy stress load value, and a physiological stress response capacity curve was generated. Predict the safety boundary of chemotherapy by analyzing physiological stress response curves and tolerance indices, and generate chemotherapy tolerance assessment data.

5. The method for analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation according to claim 1, characterized in that, The specific steps of step S4 are as follows: Establish a multi-indicator benchmark for evaluating the efficacy of chemotherapy, including the tumor response index, immune benefit index, quality of life protection index, and organ safety index; The drug synergistic model was used to evaluate the maximum drug effect in patients and extract the theoretical maximum efficacy value. Risk-efficacy benefit balance is calculated based on chemotherapy tolerance assessment data, and multi-objective optimization is performed to generate risk-efficacy benefit balance parameters. Based on the multi-index chemotherapy benefit evaluation benchmark, the chemotherapy benefit is calculated in a stratified manner using the risk-efficacy benefit balance parameter and the theoretical maximum efficacy value, generating a chemotherapy benefit index at multiple levels. The chemotherapy efficacy index is used to predict future chemotherapy treatment course index changes, generating an index prediction trajectory. Dynamic trajectory planning for chemotherapy regimens based on exponential prediction trajectories.

6. The method for analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation according to claim 5, characterized in that, The specific steps for predicting future chemotherapy course index changes and generating an index prediction trajectory based on the chemotherapy efficacy index are as follows: Multiple time-series autocorrelation analyses were performed on the trend characteristics of biomarkers to obtain the lag period of indicator changes. Calculate the response delay value at the time of chemotherapy administration based on the lag period of indicator changes, and construct a multi-indicator response delay spectrum; The chemotherapy efficacy index is time-delayed and corrected based on the multi-index response delay spectrum to obtain the time-delayed chemotherapy efficacy index. Based on the time-delayed modified chemotherapy efficacy index, the future chemotherapy course index changes are predicted, and an index prediction trajectory is generated.

7. The method for analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation according to claim 5, characterized in that, The specific steps for dynamic trajectory planning of chemotherapy regimens based on exponential prediction trajectories are as follows: Based on the exponential prediction trajectory, dynamic trajectory planning of chemotherapy regimens is performed to obtain chemotherapy adjustment plans; Based on the chemotherapy regimen adjustment, the patient's chemotherapy tolerance was reassessed to obtain the latest chemotherapy tolerance boundary; The latest chemotherapy treatment intervals and dosing time windows are calculated based on the latest chemotherapy tolerance boundaries, and real-time chemotherapy improvements are made.

8. A system for analyzing the efficacy of gastric cancer chemotherapy based on multi-index calculation, characterized in that, The method for performing the gastric cancer chemotherapy benefit analysis based on multi-index calculation as described in claim 1 includes: The trend analysis module is used to extract multi-temporal biomarkers from patients at different treatment cycles, perform non-linear trend analysis, and obtain biomarker trend characteristics. The drug metabolism simulation module is used to collect drug monitoring parameters during chemotherapy cycles, perform drug metabolism simulation based on biomarker trend characteristics, and construct a drug synchronous kinetic model. The chemotherapy tolerance assessment module is used to extract patients' vital signs and chemotherapy toxicity information based on clinical tests, predict the safety boundaries of chemotherapy, and generate chemotherapy tolerance assessment data. The benefit stratification calculation module is used to perform chemotherapy benefit stratification calculations based on chemotherapy tolerance assessment data and drug kinetic models, generating chemotherapy benefit indices at multiple levels.

Images