Lung cancer immunotherapy adverse reaction patient report management system
By constructing a phase coordinate system and a multidimensional signal decoupling model, the problem of misjudging the source of symptoms in lung cancer immunotherapy was solved, and the accurate identification of immune-related adverse reactions and tumor progression was achieved, improving the accuracy and continuity of treatment decisions.
Patent Information
- Application Number
- CN202511955700.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-23
- Publication Date
- 2026-02-10
AI Technical Summary
Current technologies lack the ability to accurately distinguish between immune-related adverse reactions and tumor progression in lung cancer immunotherapy, leading to misjudgments and incorrect treatment decisions, and failing to effectively utilize the inherent logical relationship between symptom fluctuations and dosing rhythm.
A phase coordinate system based on the drug cycle is constructed. Through a multidimensional signal decoupling model and pathological topological constraints, the symptom signals are accurately decoupled. The symptom components of immune-related adverse reactions and tumor progression are separated by calculating the phase offset and characteristic parameters.
This achieves precise decoupling of immune-related adverse reactions from tumor progression, avoids misjudgment, improves the accuracy of treatment decisions, corrects underestimation of risk, and ensures the continuity and effectiveness of immunotherapy.
Smart Images
Figure CN121506536A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of medical information technology, specifically to a patient reporting and management system for adverse reactions to lung cancer immunotherapy. Background Technology
[0002] In the field of lung cancer treatment, immunotherapy, represented by immune checkpoint inhibitors, has become an important clinical approach, greatly prolonging the survival of patients. However, the resulting immune-related adverse reactions are characterized by being systemic, delayed, and insidious. If they are not identified and intervened in a timely manner, they may lead to serious clinical consequences. Collecting subjective symptom data of outpatients using electronic patient reporting systems has become an important means of achieving full-course disease management.
[0003] However, in the complex physiological environment of immunotherapy, the symptoms exhibited by patients are actually a nonlinear superposition of the dual factors of drug-induced immune response and tumor progression itself. From a pathological perspective, immune-related adverse reactions, as products of drug stimulation, often have a specific biological coupling relationship with the dosing cycle, exhibiting a pulse-like characteristic that follows the dosing rhythm. On the other hand, the symptom exacerbation caused by tumor progression is more often manifested as linear drift or random fluctuations unrelated to the dosing rhythm. Current technologies lack the ability to perceive this deep temporal logic, focusing only on the current snapshot intensity of symptoms while ignoring their fluctuating rhythm characteristics. This blindness leads to the misjudgment of non-drug-related symptoms caused by tumor progression as severe immune adverse reactions in clinical practice, resulting in unnecessary treatment interruptions; or some early weak immune signals that strictly follow the dosing schedule are ignored as background noise.
[0004] Therefore, how to accurately identify the true pathological source of symptoms by utilizing the inherent logical relationship between symptom fluctuations and dosing rhythm in mixed signals is a core technical challenge that urgently needs to be solved. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a patient reporting and management system for adverse reactions to lung cancer immunotherapy.
[0006] To achieve the above objectives, the technical solution of the present invention is as follows:
[0007] In a first aspect, the present invention discloses a patient reporting management system for adverse reactions to lung cancer immunotherapy, comprising:
[0008] The data acquisition module is used to acquire the first symptom data set and drug administration time series of the target patient; the first symptom data set includes multiple symptom feature vectors and their corresponding acquisition times within multiple drug administration cycles defined by the drug administration time series, and the symptom feature vectors contain symptom values of multiple different clinical dimensions;
[0009] The phase offset calculation module is used to determine the phase offset of the acquisition time of each symptom feature vector in the first symptom data set relative to the start time of its respective dosing cycle, based on the drug administration time series.
[0010] The feature parameter calculation module is used to construct a periodic phase coordinate system based on the phase offset, map the symptom feature vector to the corresponding phase coordinate point in the periodic phase coordinate system, and calculate the first feature parameter and the second feature parameter based on the mapping result.
[0011] The real-time data processing module is used to acquire the second symptom data of the target patient at the current acquisition time and determine the current phase offset based on the position of the current acquisition time in the drug administration time series.
[0012] The feature decomposition module is used to construct a multidimensional signal decoupling model based on the first feature parameter and the second feature parameter, and to perform vector decomposition on the second symptom data based on the current phase offset through the multidimensional signal decoupling model to obtain the first component vector and the second component vector.
[0013] The evaluation generation module is used to generate corresponding hierarchical evaluation results based on the magnitude values of the first component vector and the second component vector.
[0014] Secondly, this invention discloses a method for managing patient reports of adverse reactions to lung cancer immunotherapy, comprising the following steps:
[0015] Acquire the first symptom data set and drug administration time series of the target patient; the first symptom data set includes multiple symptom feature vectors and their corresponding collection times within multiple drug administration cycles defined by the drug administration time series, and the symptom feature vectors contain symptom values of multiple different clinical dimensions;
[0016] Based on the drug administration time series, the phase offset of the acquisition time of each symptom feature vector in the first symptom data set relative to the start time of its respective drug administration cycle is determined.
[0017] A periodic phase coordinate system is constructed based on the phase offset, the symptom feature vector is mapped to the corresponding phase coordinate point in the periodic phase coordinate system, and the first feature parameter and the second feature parameter are calculated based on the mapping result.
[0018] Acquire the target patient's second symptom data at the current acquisition time, and determine the current phase offset based on the position of the current acquisition time in the drug administration time series;
[0019] A multidimensional signal decoupling model is constructed based on the first and second feature parameters. Based on the current phase offset, the second symptom data is decomposed into vectors through the multidimensional signal decoupling model to obtain the first component vector and the second component vector.
[0020] Based on the magnitude of the first component vector and the magnitude of the second component vector, the corresponding hierarchical evaluation results are generated.
[0021] Compared with the prior art, the beneficial effects of the present invention are as follows:
[0022] 1. Abandoning the traditional linear threshold determination, a phase coordinate system based on the drug cycle was constructed. By calculating the resonance intensity of symptoms and dosing rhythm and the temporal topological consistency between multiple dimensions, the precise decoupling of immune-related adverse reactions (irAE) and tumor progression (PD) at the mathematical level was achieved, effectively avoiding erroneous treatment decisions caused by misjudgment of the source of symptoms.
[0023] 2. Through vectorized normalized mapping and modulus analysis, the system can capture the synergistic resonance signal of weak symptoms in multiple organs; even if a single symptom does not reach the critical threshold, the system can identify systemic risks that meet the precursors of an immune storm through topological analysis, thereby correcting the risk underestimation caused by single-indicator monitoring.
[0024] 3. By converting physical time into biological phase time and introducing a set of pathological topological constraints based on clinical consensus, the data analysis process strictly follows the pharmacokinetic laws of immunotherapy and the evolution logic of complications, thereby upgrading traditional static numerical management to dynamic pathological tracing. Attached Figure Description
[0025] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0026] Figure 1 This is an overall block diagram of the system according to Embodiment 1 of the present invention;
[0027] Figure 2 This is an overall block diagram of the method in Embodiment 2 of the present invention;
[0028] Figure 3 This is a flowchart illustrating the overall execution process of the method in Embodiment 2 of the present invention. Detailed Implementation
[0029] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0030] Application Overview: In the field of full-course management of lung cancer immunotherapy, especially in continuous monitoring of outpatient-reported outcomes (PROs), accurate tracing of the source of symptoms is regarded as the core basis for making clinical decisions. At the pathophysiological level, this high-quality symptom interpretation is essentially a process of signal source separation. That is, immune-related adverse reactions (irAEs), as biological responses induced by drugs, should exhibit phase-locked characteristics on the time axis that are highly synchronized with the drug concentration metabolism rhythm through a strict pharmacokinetic coupling mechanism, using the dosing cycle as a reference clock, and following the temporal topological laws between specific organ systems.
[0031] However, existing technologies lack a mechanism to verify the spatiotemporal consistency between the input drug delivery pulse and the output symptom response, resulting in an inability to accurately identify pseudo-progression interference and occult immune activation issues in complex mixed signals. Pseudo-progression interference manifests as the system capturing the absolute increase in symptom values but ignoring the characteristic that this fluctuation occurs during the drug washout period (low concentration phase), which is actually driven by the linear deterioration of the tumor's baseline burden. Occult immune activation manifests as low-level non-specific symptoms in multiple organs, but because the system has severed the temporal connection between dimensions, it fails to identify that it conforms to the chain topological law of cytokine release-tissue damage. As a result, a strict physical correspondence cannot be established between the periodic rhythm of the drug and the vector changes in symptoms, causing the system to misjudge pathological attributes or flood the signal, thereby affecting the specificity of adverse reaction warnings and the continuity of anti-tumor treatment.
[0032] For example, in real-world patient follow-up scenarios, when a patient experiences worsening dyspnea at the end of the dosing cycle (around day 20), conventional systems can only detect symptom score deviations through linear threshold analysis, but cannot distinguish whether there is a significant phase orthogonality (i.e., irrelevance) to the dosing rhythm. Furthermore, when a patient develops a mild rash followed by diarrhea in the early stages, the system only records them as two independent low-risk events, failing to detect the systemic immune storm precursor implied by this skin-gut-induced timing. Specifically, the system misjudges baseline drift caused by tumor progression as immune pneumonia and triggers a discontinuation recommendation, or misclassifies early weak signals consistent with immune mechanisms as background noise. This leads to clinical decisions continuously deviating from the correct pathological logic, failing to form a dynamic management plan that conforms to the biological principles of immunotherapy.
[0033] If the above problems are not addressed, the management system will continue to lose its ability to objectively distinguish the pathological attributes of symptoms. Among these issues, failure to identify tumor progression signals will lead patients to receive incorrect immunosuppressive therapy, resulting in the loss of the optimal window for anti-tumor treatment. Simultaneously, failure to recognize the phase characteristics of the immune response will cause a lag in the early warning mechanism, causing early controllable toxicity to evolve into severe illness, ultimately rendering the monitoring data worthless in terms of decision support. Consequently, inaccurate teaching feedback will systematically hinder doctors from grasping the true immune tolerance status of patients, affecting the achievement of the survival benefit goal of immunotherapy.
[0034] Example 1:
[0035] like Figure 1 As shown, the patient reporting and management system for adverse reactions to lung cancer immunotherapy includes:
[0036] The data acquisition module is used to acquire the first symptom data set and drug administration time series of the target patient; the first symptom data set includes multiple symptom feature vectors and their corresponding acquisition times within multiple drug administration cycles defined by the drug administration time series, and the symptom feature vectors contain symptom values of multiple different clinical dimensions;
[0037] In real-world clinical settings, symptom reports from lung cancer immunotherapy patients often exhibit heterogeneous data types (e.g., body temperature is a continuous variable, while cough frequency is a discrete rank) and non-uniform temporal distribution. Existing data acquisition methods typically simply record isolated numerical values, making it impossible for subsequent analysis to assess synergistic pathological changes in different organs under the same dimensionality. To address this data silo and dimensional barrier problem, the data acquisition module in this embodiment does not merely perform simple read operations. Instead, based on multimodal feature alignment logic, it reconstructs the heterogeneous raw data into standardized symptom feature vectors and uses drug administration time series data to construct a pathologically significant temporal index.
[0038] Specifically, the system first obtains the patient's medication time series from the hospital information system (HIS), denoted as... This clearly defines the time boundaries for multiple dosing cycles. Subsequently, based on the raw symptom data (such as body temperature) reported by patients via mobile devices... Area of rash Shortness of breath score The module internally performs vectorized normalization mapping. Considering that the physiological baseline and critical threshold differ for different symptoms, this module introduces a dynamic maximum-minimum normalization algorithm to map the original value of each dimension to... We construct standardized symptom feature vectors from dimensionless intervals.
[0039] Its core normalization logic follows the formula below:
[0040]
[0041] in, Indicates at time The collected symptom feature vectors Standardized symptom values for each clinical dimension (such as the respiratory dimension); This refers to the original reported value for this dimension; This is the physiological baseline value for this dimension (e.g., baseline body temperature is 36.5℃, baseline cough is 0). This refers to the peak clinical urgency level for this dimension (e.g., a critical temperature of 40.0℃, or a cough of level 10). To prevent tiny compensation terms with a denominator of zero.
[0042] Through this mapping, the symptom feature vector It characterizes the relative severity of each organ system's deviation from its normal physiological state, which transforms the originally incomparable fever of 38 degrees and shortness of breath of level 3 into standardized signals that can be weighted and topologically compared in vector space, laying a mathematical foundation for subsequent accurate calculations.
[0043] For example, assuming the target patient is undergoing PD-1 immunotherapy, the data acquisition module obtains the most recent dosing time from the dosing time series. It is October 1st.
[0044] On October 6th (i.e.) The patient reported two sets of raw data:
[0045] Body temperature (clinical dimension 1): 38.5℃. Baseline for this dimension is known. =$36.5, critical peak .
[0046] Cough score (clinical dimension 2): Grade 2 (0-10). Baseline score for this dimension is known. Critical peak .
[0047] The system substitutes the above formula into the calculation:
[0048] Standardized values for body temperature:
[0049] Standardized values for the cough dimension:
[0050] Finally, the data acquisition module outputs the symptom feature vector at that moment. This vector clearly reveals that the patient's "systemic inflammatory response (body temperature)" is significantly stronger than "local respiratory symptoms (cough)" at the current moment. This relative difference in intensity between dimensions will be used by subsequent modules to identify the presence of early signals of "cytokine release syndrome" (i.e., systemic response precedes local response), thus effectively avoiding intuitive misjudgments that may result from looking only at the raw values.
[0051] The phase offset calculation module is used to determine the phase offset of the acquisition time of each symptom feature vector in the first symptom data set relative to the start time of its respective dosing cycle, based on the drug administration time series.
[0052] Furthermore, the calculation process for the phase offset includes:
[0053] Obtain the dosing cycle duration parameter from the dosing time series;
[0054] For each collection time in the first symptom data set, the collection time is compared with each dosing time in the dosing time series to filter out all dosing times in the dosing time series that are earlier than or equal to the collection time.
[0055] The dosing time with the largest timestamp value is selected as the baseline dosing time, and the time difference between the collection time and the baseline dosing time is calculated.
[0056] In real-world clinical scenarios, because patient medication administration times are not absolutely regular and symptom reporting times are random, directly using physical calendar times (e.g., October 5th) to analyze symptom data leads to temporal misalignment and fails to reveal the intrinsic relationship between symptoms and drug metabolism. To address this temporal alignment challenge, discrete physical time is transformed into biologically meaningful relative time. The phase offset calculation module no longer relies on absolute calendar timestamps but instead calculates the relative phase at each acquisition moment based on the dosing cycle defined by the dosing time series through rigorous temporal backtracking logic.
[0057] In practice, this module first loads the drug administration time series extracted from medical records. The sequence consists of multiple dosing times arranged in chronological order. The time span between two adjacent dosing times in the sequence is defined as a dosing cycle, and its standard duration is denoted as the dosing cycle duration parameter. (For example, for a certain PD-1 inhibitor) (Typically 21 days). For each symptom feature vector in the first symptom dataset, the module reads its corresponding acquisition time. And perform a baseline anchoring operation: the system traverses the dosing time series, filters out all dosing times that are earlier than or equal to the acquisition time on the time axis, that is, constructs a subset. Subsequently, the time stamp with the largest value from this subset is selected as the baseline dosing time for that acquisition time. This screening logic ensures that the system can accurately pinpoint the most recent drug intervention event that affected the patient, regardless of whether the patient reported the data on the day of administration or at the end of the cycle, thus avoiding cross-cycle attribution errors.
[0058] After determining the baseline anchor point, the module further calculates the time difference between the acquisition time and the baseline drug administration time, and normalizes and maps it to a preset period interval to obtain the phase offset. Its core calculation logic follows the phase mapping formula:
[0059]
[0060] in, This represents the phase offset corresponding to the symptom feature vector, and its value range is typically [value range missing]. This characterizes the relative position of symptom occurrence during the dosing cycle (e.g., On the corresponding day of administration, (Corresponding to the midpoint of the cycle); This refers to the actual time when the symptom data was collected; The baseline dosing time is determined through the above screening logic; This refers to the duration of the dosing cycle. Through this nonlinear mapping, data collected from different dosing cycles (e.g., day 3 of cycle 1 and day 3 of cycle 5) are projected onto the same phase coordinate point. This transformation eliminates the barrier of physical time, enabling the system to perform superposition and resonance analysis on symptom data scattered over months of treatment in subsequent steps, under a unified drug metabolism clock. This provides a crucial temporal benchmark for identifying phase-locked adverse immune reactions.
[0061] The feature parameter calculation module is used to construct a periodic phase coordinate system based on the phase offset, map the symptom feature vector to the corresponding phase coordinate point in the periodic phase coordinate system, and calculate the first feature parameter and the second feature parameter based on the mapping result.
[0062] After completing the unified mapping of phase shifts, in order to extract diagnostically valuable drug response features from complex mixed symptom signals, the feature parameter calculation module does not directly process the original discrete points. Instead, it constructs a data aggregation space based on phase alignment, namely a periodic phase coordinate system. This module first defines the granularity of the phase axis, dividing the continuous phase period... Divided into A continuous and non-overlapping discrete phase interval (e.g., ... The cycle is divided into 20 intervals, each representing approximately 5% of the time window within the dosing cycle. Subsequently, the system performs a full mapping of historical data, traversing each symptom feature vector in the first symptom data set, and projecting it into the corresponding discrete phase interval based on the phase offset calculated in the preceding module.
[0063] For each clinical dimension (such as body temperature or rash), the module aggregates all historical symptom values falling within each discrete phase interval, constructing a subset of sample values. The logic behind this aggregation process is that if a symptom is an involuntary immune response (irAE) induced by a drug, according to pharmacokinetic principles, it should recur at the same phase across multiple dosing cycles (e.g., always appearing on days 5-7 after administration), resulting in sample values within that phase interval exhibiting high mean and low dispersion. Conversely, if the symptom is caused by tumor progression, it typically exhibits a random distribution or linear drift throughout the entire cycle on the phase axis, showing high dispersion or no obvious phase clustering. Based on this, the module calculates the mean of each subset. and standard deviation Based on this, the first feature parameter is constructed. This is used to quantify the resonance intensity of this symptom dimension stimulated by the drug.
[0064] Its core calculation formula is as follows:
[0065]
[0066] in, Indicates the first The clinical dimension in the first The first characteristic parameter under each discrete phase interval; and These represent the mean and standard deviation of the subset of sample values constructed for this dimension within this phase interval, respectively; To preset a minimum value to prevent the denominator from being zero; This is the adjustment coefficient. This formula, by using the ratio of the mean to the standard deviation as the resonance intensity, effectively amplifies the weight of signals that are "high in intensity and highly regular," while suppressing occasional noise interference.
[0067] In particular, for patients in the early stages of treatment or with insufficient sample size due to sparse historical data (e.g.) In cases where the feature parameter calculation module introduces a Population Prior Model as a supplement, the system pre-sets a standard resonance baseline based on statistical analysis of a large number of previous cases of similar drugs. When the sample size for a specific phase interval of the target patient is insufficient, a Bayesian smoothing strategy is adopted, using the population prior mean and variance to weight and correct the calculation results for that patient, ensuring that in the early stages of individual data accumulation, It can still reflect the general resonance pattern of the drug, thus avoiding calculation divergence or zero denominator anomalies caused by data scarcity.
[0068] After calculating the resonance features of a single dimension, the module continues to calculate the second feature parameter to further capture the temporal correlation logic between different symptoms. The system first identifies the characteristic phase coordinates of each clinical dimension in the periodic phase coordinate system, that is, the phase point where the first feature parameter of that dimension reaches its peak or the average symptom value is highest, which represents the typical outbreak time of that symptom within the dosing cycle. Next, the module calculates the resonance parameters of any two different clinical dimensions (such as the rash dimension). With diarrhea dimension The difference between the characteristic phase coordinates of the symptoms is used to construct a set of relative phase differences. This set constitutes the second characteristic parameter, which digitally represents the temporal topological relationships such as the occurrence of rash before diarrhea or the simultaneous occurrence of fever and shortness of breath. In this way, the module not only quantifies the drug relevance of individual symptoms, but also extracts the co-evolutionary patterns among multiple symptoms, providing key topological evidence for the subsequent accurate identification of specific adverse reaction patterns that conform to immune mechanisms.
[0069] The real-time data processing module is used to acquire the second symptom data of the target patient at the current acquisition time and determine the current phase offset based on the position of the current acquisition time in the drug administration time series.
[0070] After completing pattern learning and model building based on historical data, the real-time data processing module undertakes the crucial task of integrating the latest discrete monitoring data into the system's analysis logic to achieve immediate assessment of the patient's current state. Unlike the preceding modules, which focus on mining historical patterns, the core of this module lies in precise positioning—that is, rapidly determining the specific physiological phase within the medication cycle at the moment the patient reports symptoms, thereby providing accurate spatiotemporal coordinates for subsequent model inference.
[0071] In practice, this module first obtains the target patient's information at the current data collection time. The second symptom data was reported. To eliminate dimensional differences, the system reused the normalization logic in the data acquisition module to transform it into a standardized current symptom vector. However, this vector only represents a snapshot of symptoms in physical time, and its clinical significance is ambiguous without placing it within the context of drug metabolism. Therefore, the module must perform phase calibration of the current moment based on the dosing time series. Dosing time series Serving as a time reference axis here, it not only records historical dosing points but also defines the specific dosing cycle at the current moment.
[0072] The module executes a time-series retrieval algorithm to search for the dosing time series corresponding to the current acquisition time. The most recent dosing event that occurred before it is used as the baseline start point for the current cycle. The purpose of this step is that the biological effects of immunotherapy are reset and evolve with each dosing pulse; only by using the most recent dosing as the zero point can the cumulative effect or metabolic attenuation of the drug in the body be accurately measured. Based on the locked reference point, the module uses linear interpolation logic to calculate the current phase offset. The calculation formula is as follows:
[0073]
[0074] in, This is a parameter for the duration of the dosing cycle. The formula seamlessly converts the physical time difference into a phase angle. For example, if a patient reports data on day 10.5 of the dosing cycle, and the cycle is 21 days, the calculated... Will be This calculation result In essence, it acts as a pointer for the system to perform an index lookup in the periodic phase coordinate system. It indicates to subsequent modules which set of historical characteristic parameters (i.e., the resonance intensity and topological relationship of which phase interval) should be called to interpret the current symptom vector, thereby ensuring the accuracy of the analysis. The analysis at each time point is based on the specific pharmacokinetic context of that time point, rather than a general static assessment.
[0075] The feature decomposition module is used to construct a multidimensional signal decoupling model based on the first feature parameter and the second feature parameter, and to perform vector decomposition on the second symptom data based on the current phase offset through the multidimensional signal decoupling model to obtain the first component vector and the second component vector.
[0076] After the real-time data processing module accurately locates the phase coordinates at the current acquisition moment, the feature decomposition module constructs a multi-dimensional signal decoupling model to decouple the mixed and superimposed current symptom vector. Decomposed into the first component vector characterizing the immune response And the second component vector representing tumor progression. Existing technologies often struggle to distinguish between drug-induced symptom fluctuations and disease progression itself, leading to frequent misdiagnosis. The root cause lies in the failure to effectively utilize the temporal dependencies between symptoms for logical verification. To address this pain point, this module no longer considers the numerical magnitude of individual symptoms in isolation. Instead, it introduces a set of pathological topological constraints as a priori knowledge base. By verifying whether the current symptom combination conforms to the unique temporal occurrence pattern of the immune response (i.e., topological consistency), the weights of the decoupled model are dynamically adjusted.
[0077] In practice, the module first loads a pre-defined set of pathological topological constraints, which is a digital rule base built on clinical immunology consensus. This set is instantiated as a data table containing several sets of related dimension pairs. Each set of entries defines two clinical dimensions with pathological mechanism associations (e.g., rash and diarrhea) and the standard phase difference range they should satisfy under the immune response mechanism. For example, medical evidence suggests that skin toxicity usually occurs before gastrointestinal toxicity, and this temporal relationship is quantified as a specific phase difference in a periodic phase coordinate system (e.g., To facilitate computer processing, the pathological topological constraint set is concretely instantiated in engineering as a multidimensional associative data structure (or associative dataset). Each entry in this data structure represents a constraint rule and contains the following fields: 1. Precursor Dimension (Source Dim): e.g., rash; 2. Successor Dimension (Target Dim): e.g., diarrhea; 3. Expected Phase Difference (Exp Phase Diff): e.g., (Corresponding to a time lag of approximately 5-7 days); 4. Tolerance window: for example... 5. Rule Weight: Set based on the clinical incidence rate of this complication combination (e.g., 0.8). During system runtime, the actual phase difference is calculated by querying this associated data structure. The matching degree is scored by projecting the data onto the tolerance window mentioned above.
[0078] The system iterates through these associated dimension pairs and extracts the actual phase difference value corresponding to the current patient from the second feature parameter (i.e., the set of phase difference values) obtained from the previous calculation. The matching degree between the two is then calculated to obtain the topological consistency coefficient. .
[0079] The calculation logic of this coefficient aims to quantify the degree of agreement between the currently observed symptom timeline pattern and the standard immune response pattern. The calculation formula is as follows:
[0080]
[0081] in, This is the topology consistency coefficient, and its value range is... The closer the value is to 1, the more it conforms to the characteristics of an immune response; This represents the set of indices for all associated dimension pairs in the pathological topological constraint set. Refers to the first The and the first One clinical dimension; This represents the patient's actual relative phase difference. The preset standard phase difference value; The weighting factor is used to adjust the importance of different rules. This formula utilizes the properties of the Gaussian radial basis function, so that the consistency coefficient decays exponentially as the actual phase difference deviates further from the standard value, thereby greatly suppressing the weights of spurious signals with temporal discrepancies (such as late symptoms appearing before early symptoms).
[0082] It is worth noting that the weighting factors in the above formula (such as adjustment coefficients) and rule weight The parameters are not fixed constants, but are obtained through offline training using a supervised learning algorithm. Specifically, during the research and development phase, a dataset of labeled historical cases diagnosed as immune adverse reactions (irAE) and tumor progression (PD) was collected. With classification accuracy as the optimization objective, these hyperparameters were iteratively optimized using gradient descent. This ensured that each weight value in the model represented the statistical significance of that feature in distinguishing between irAE and PD.
[0083] After obtaining the topological consistency coefficient, the module combines it with the first feature parameter to construct the final signal decoupling operator. (First feature parameter) While reflecting the resonance intensity of unidimensional symptoms over the course of medication, this may be affected by random noise. Therefore, this module employs a dual weighting strategy of resonance intensity and topological consistency to construct a resonance weight matrix. The modulated vector is then used to generate an immune response projection operator. This operator is then used to project the current symptom vector onto the immune response subspace, and the first component vector is calculated. The calculation formula is as follows:
[0084]
[0085] in, This is the current symptom vector; This is the current phase offset. It is a diagonal matrix whose diagonal elements are determined by each dimension in the current phase. The first characteristic parameter below The system retains only those signal components that historically resonate with the drug administration cycle and temporally conform to immune topological rules. Finally, the module calculates the vector difference between the current symptom vector and the first component vector to obtain the second component vector, thus achieving precise decoupling between drug side effects and tumor progression.
[0086] This vector difference calculation is based on the additive signal model assumption in signal processing. Based on fundamental principles of immunology, it assumes that the total symptom energy observed by the patient is mainly composed of a drug-induced component, a disease background component, and Gaussian white noise linearly superimposed. Because the first component vector... By employing resonant filtering and topological gating, signals with drug-related characteristics have been extracted to the maximum extent possible. Therefore, after removing this component from the total observation vector, the residual vector likely retains primarily the disease background signal (i.e., tumor progression) and random noise, which do not possess drug periodicity characteristics. Mathematically, this approach achieves effective separation of mixed pathological signals in orthogonal subspaces.
[0087] The evaluation generation module is used to generate corresponding hierarchical evaluation results based on the magnitude values of the first component vector and the second component vector.
[0088] After the feature decomposition module successfully decomposes the mixed symptom signals mathematically into a first component vector representing the immune response and a second component vector representing tumor progression, the assessment generation module undertakes the final task of transforming these abstract, high-dimensional mathematical features into concrete, actionable clinical decision recommendations. Since clinicians cannot directly and quickly determine the severity of a patient's condition by observing the numerical distribution of multidimensional vectors, and single-dimensional symptom values often fail to reflect the systemic biological burden, this module no longer relies on threshold alarms for single symptoms. Instead, it employs vector norm analysis to map the multidimensional components into scalar indicators representing pathological intensity, and generates grading assessment results accordingly.
[0089] In practice, the module first performs modulus calculation on the two input component vectors. Based on the Euclidean norm principle, the system calculates the first component vector... modulus Second component vector modulus Its calculation logic follows the formula below:
[0090]
[0091] in, This represents the calculated modulus value; The component vector representing the input (i.e., the first component vector or the second component vector). For the first vector The component values of each clinical dimension; The total number of dimensions of the vector. Modulus. In effect, it measures the total energy or systemic burden of the pathological component in the multidimensional symptom space. For example, a patient who presents with both a mild rash and mild diarrhea (values in both dimensions) may have a higher vector magnitude than a patient who presents with only a single moderate rash. This aligns with the clinical understanding that multi-organ involvement in immunotherapy often indicates a higher level of immune response, thus correcting the underestimation of systemic risk that may be caused by traditional single-indicator monitoring.
[0092] After obtaining two independent pathological intensity moduli, the module enters the differential interpretation stage, generating the final assessment using pre-set decision logic. The system will... and Compare each with its respective risk threshold: If The symptom intensity significantly exceeded the preset immune alert threshold and dominated the total symptom energy. The system determined that the current condition was primarily driven by drug-induced immune-related adverse reactions (irAEs), and then... The numerical range is mapped to the corresponding toxicity grade (e.g., CTCAE 1-4), and clinical recommendations are generated suggesting the suspension of immunotherapy or the initiation of hormonal intervention; conversely, if In a high position A low symptom level indicates that the current symptom fluctuations do not exhibit the resonance characteristics of the drug cycle or do not conform to immune topological laws. The system then classifies this as non-drug-related disease progression (PD) and generates a suggestion to conduct follow-up imaging studies, thereby effectively avoiding the erroneous implementation of immunosuppressive therapy for patients with tumor progression. Through this mechanism, the system ultimately achieves a closed loop from data monitoring to precise decision support, solving the clinical management dilemma caused by the one-size-fits-all approach to alarms in complex and mixed symptoms of existing technologies.
[0093] Example 2:
[0094] like Figures 2-3 As shown, the patient reporting and management methods for adverse reactions to lung cancer immunotherapy include the following steps:
[0095] Acquire the first symptom data set and drug administration time series of the target patient; the first symptom data set includes multiple symptom feature vectors and their corresponding collection times within multiple drug administration cycles defined by the drug administration time series, and the symptom feature vectors contain symptom values of multiple different clinical dimensions;
[0096] Based on the drug administration time series, the phase offset of the acquisition time of each symptom feature vector in the first symptom data set relative to the start time of its respective drug administration cycle is determined.
[0097] A periodic phase coordinate system is constructed based on the phase offset, the symptom feature vector is mapped to the corresponding phase coordinate point in the periodic phase coordinate system, and the first feature parameter and the second feature parameter are calculated based on the mapping result.
[0098] Acquire the target patient's second symptom data at the current acquisition time, and determine the current phase offset based on the position of the current acquisition time in the drug administration time series;
[0099] A multidimensional signal decoupling model is constructed based on the first and second feature parameters. Based on the current phase offset, the second symptom data is decomposed into vectors through the multidimensional signal decoupling model to obtain the first component vector and the second component vector.
[0100] Based on the magnitude of the first component vector and the magnitude of the second component vector, the corresponding hierarchical evaluation results are generated.
[0101] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.
[0102] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.
[0103] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.
Claims
1. A patient reporting and management system for adverse reactions to lung cancer immunotherapy, characterized in that, include: The data acquisition module is used to acquire a first symptom data set and a drug administration time series of the target patient; the first symptom data set includes multiple symptom feature vectors and their corresponding acquisition times within multiple drug administration cycles defined by the drug administration time series, and the symptom feature vectors contain symptom values of multiple different clinical dimensions; The phase offset calculation module is used to determine the phase offset of the acquisition time of each symptom feature vector in the first symptom data set relative to the start time of its respective dosing cycle, based on the drug administration time series. The feature parameter calculation module is used to construct a periodic phase coordinate system based on the phase offset, map the symptom feature vector to the corresponding phase coordinate point in the periodic phase coordinate system, and calculate the first feature parameter and the second feature parameter based on the mapping result. The real-time data processing module is used to acquire the second symptom data of the target patient at the current acquisition time, and determine the current phase offset based on the position of the current acquisition time in the drug administration time series; The feature decomposition module is used to construct a multidimensional signal decoupling model based on the first feature parameter and the second feature parameter, and to perform vector decomposition on the second symptom data based on the current phase offset through the multidimensional signal decoupling model to obtain a first component vector and a second component vector. The evaluation generation module is used to generate corresponding hierarchical evaluation results based on the magnitude of the first component vector and the magnitude of the second component vector.
2. The patient reporting and management system for adverse reactions to lung cancer immunotherapy according to claim 1, characterized in that: The calculation process for the phase offset includes: Obtain the dosing cycle duration parameter from the dosing time series; For each collection time in the first symptom data set, the collection time is compared with each administration time in the administration time series to filter out all administration times in the administration time series that are earlier than or equal to the collection time. The dosing time with the largest timestamp value is selected as the baseline dosing time, and the time difference between the acquisition time and the baseline dosing time is calculated. Based on the ratio of the time difference to the dosing cycle duration parameter, the acquisition time is mapped to a phase offset value within a preset cycle interval.
3. The patient reporting and management system for adverse reactions to lung cancer immunotherapy according to claim 1, characterized in that: The calculation process of the first feature parameter includes: The phase axis in the periodic phase coordinate system is divided into multiple continuous discrete phase intervals; For each clinical dimension contained in the symptom feature vector, traverse the first symptom data set and filter out all symptom feature vectors whose mapped coordinate points fall within the current discrete phase interval. From the selected symptom feature vectors, extract the symptom values corresponding to the clinical dimension and construct a subset of sample values under the phase interval; Calculate the numerical distribution characteristics within the subset of sample values, where the numerical distribution characteristics include the mean and standard deviation; The first feature parameter is calculated based on the mean and the standard deviation, and is used to represent the resonance intensity of the clinical dimension excited by the drug in the discrete phase interval.
4. The patient reporting management system for adverse reactions to lung cancer immunotherapy according to claim 3, characterized in that: The formula for calculating the first feature parameter is: in, Indicates the first The clinical dimension in the first The first characteristic parameter under each discrete phase interval; , They represent the first time. The mean and standard deviation of a subset of sample values constructed for this clinical dimension within a discrete phase interval; This is a preset minimum value to prevent the denominator from being zero; This is the preset weight adjustment coefficient.
5. The patient reporting and management system for adverse reactions to lung cancer immunotherapy according to claim 1, characterized in that: The calculation process of the second characteristic parameter includes: For each clinical dimension in the symptom feature vector, the characteristic phase coordinates of that clinical dimension are identified in the periodic phase coordinate system. The characteristic phase coordinates are the phase points where the symptom value distribution density is the highest or the symptom value is the highest in that clinical dimension. For any two different clinical dimensions, calculate the relative phase difference between their respective characteristic phase coordinates; A phase difference set containing the relative phase difference values of each group is constructed, and the phase difference set is defined as the second feature parameter. The second feature parameter is used to characterize the lag or lead relationship of the triggering time sequence between different symptom dimensions.
6. The patient reporting and management system for adverse reactions to lung cancer immunotherapy according to claim 5, characterized in that: The process of constructing the multidimensional signal decoupling model includes: Obtain a preset set of pathological topological constraints, in which several sets of associated dimension pairs with immune mechanism associations are defined, and the standard phase difference range that each set of associated dimension pairs should satisfy. Traverse the associated dimension pairs and extract the actual phase difference value corresponding to each associated dimension pair in the current monitoring data from the phase difference value set; The actual phase difference value is matched with the corresponding standard phase difference range to calculate the topological consistency coefficient, which is used to quantify the degree to which the temporal characteristics of the current symptom combination conform to the immune response mechanism.
7. The patient reporting and management system for adverse reactions to lung cancer immunotherapy according to claim 6, characterized in that: The formula for calculating the topology consistency coefficient is as follows: in, Represents the topology consistency coefficient; This represents the set of indexes consisting of all associated dimension pairs defined in the pathological topological constraint set; Indicates the first element belonging to the associated dimension pair The and the first One clinical dimension; This represents the value extracted from the set of phase difference values, corresponding to the first... The and the first The actual phase difference value of each clinical dimension; This represents the standard phase difference value defined in the pathological topological constraint set; These are preset weighting factors.
8. The patient reporting and management system for adverse reactions to lung cancer immunotherapy according to claim 6, characterized in that: The process of vector decomposition of the second symptom data includes: Map the second symptom data to the current symptom vector; A resonance weight matrix is constructed based on the first feature parameter, and the resonance weight matrix is modulated based on the topological consistency coefficient to generate an immune response projection operator. The first component vector is obtained by projecting the current symptom vector onto the immune response projection operator. Calculate the vector difference between the current symptom vector and the first component vector, and define the vector difference as the second component vector.
9. The patient reporting and management system for adverse reactions to lung cancer immunotherapy according to claim 8, characterized in that: The formula for calculating the first component vector is: in, Denotes the first component vector. This represents the current symptom vector. Indicates the current phase offset. Represents the topology consistency coefficient. This represents the resonance weight matrix, which is a diagonal matrix. The elements on the diagonal represent the phase intervals of each clinical dimension. The first characteristic parameter.
10. A method for managing patient reports of adverse reactions to lung cancer immunotherapy, characterized in that, Includes the following steps: Acquire a first set of symptom data and a drug administration time series of the target patient; the first set of symptom data includes multiple symptom feature vectors and their corresponding acquisition times within multiple drug administration cycles defined by the drug administration time series, and the symptom feature vectors contain symptom values of multiple different clinical dimensions; Based on the drug administration time series, determine the phase offset of the acquisition time of each symptom feature vector in the first symptom data set relative to the start time of its respective drug administration cycle; A periodic phase coordinate system is constructed based on the phase offset, the symptom feature vector is mapped to the corresponding phase coordinate point in the periodic phase coordinate system, and the first feature parameter and the second feature parameter are calculated based on the mapping result. Acquire the second symptom data of the target patient at the current acquisition time, and determine the current phase offset based on the position of the current acquisition time in the drug administration time series; A multidimensional signal decoupling model is constructed based on the first feature parameter and the second feature parameter. Based on the current phase offset, the second symptom data is vector decomposed through the multidimensional signal decoupling model to obtain the first component vector and the second component vector. Based on the magnitude of the first component vector and the magnitude of the second component vector, a corresponding hierarchical evaluation result is generated.