Intelligent remote follow-up visit method and system for tumor patients

Abstract

The invention discloses an intelligent remote follow-up visit method and system for tumor patients, and relates to the technical field of medical intellectualization, and the method comprises the steps: collecting follow-up visit data through a remote medical terminal, carrying out the processing of the follow-up visit data, obtaining a physical sign index sequence and a symptom feature vector, and generating a dynamic follow-up visit evaluation score through a time sequence prediction model; performing semantic analysis on the electronic medical record data of the patient based on the medical knowledge map, extracting tumor stage classification data, chemoradiotherapy measurement data, weight nutrition indexes and laboratory inspection data, determining a basic risk coefficient, a cumulative dose ratio and a body state score, and generating a baseline follow-up assessment score through a prediction model; and according to the dynamic follow-up evaluation score and the baseline follow-up evaluation score, calculating a comprehensive follow-up grade and generating a next follow-up plan. According to the invention, the accuracy of patient state evaluation is improved, accurate configuration of medical resources is realized, and the quality and efficiency of remote follow-up visit are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent medical technology, and in particular to an intelligent remote follow-up method and system for tumor patients. Background Art

[0002] Remote follow-up management of cancer patients is an indispensable part of the contemporary cancer diagnosis and treatment system. Its development process has undergone a major transformation from traditional telephone follow-up to intelligent remote monitoring. Telemedicine systems based on Internet of Things technology and wearable devices have realized real-time collection of patients' physiological indicators, and the introduction of artificial intelligence and big data analysis technology has made the assessment of patients' status more accurate. However, the existing remote follow-up system still has problems such as insufficient data integration, single evaluation model, and rigid follow-up strategy. Specifically, the current system often separates real-time monitoring data from electronic medical record information, lacks an effective multi-source heterogeneous data fusion mechanism; the evaluation model is mostly based on static indicators at a single time point, which makes it difficult to capture the dynamic changes in the patient's status; and the preset fixed-period follow-up mode cannot adjust the follow-up frequency in time according to individual differences and changes in the patient's condition, resulting in inefficient allocation of medical resources.

[0003] More importantly, existing technologies have obvious limitations when dealing with the complex clinical characteristics of cancer patients. Traditional assessment methods are difficult to effectively integrate multi-dimensional clinical characteristics such as tumor staging information, treatment measurement data, nutritional status and test indicators, resulting in insufficient accuracy in risk assessment. At the same time, due to the lack of semantic analysis capabilities based on medical knowledge graphs, the system cannot fully mine the implicit information in electronic medical records, affecting the reliability of the assessment results. In addition, the existing follow-up plan formulation method is relatively mechanical, and has failed to establish an effective correlation mechanism between dynamic assessment and baseline assessment, and cannot achieve intelligent early warning and timely intervention of changes in patient status. Summary of the invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, the present invention provides an intelligent remote follow-up method and system for tumor patients, which can solve the problems mentioned in the background technology.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: an intelligent remote follow-up method for tumor patients, comprising: collecting follow-up data through a remote medical terminal, the follow-up data including real-time physical sign monitoring data and symptom quantification data, standardizing the real-time physical sign monitoring data to obtain a physical sign indicator sequence, structurally processing the symptom quantification data to obtain a symptom feature vector, inputting the physical sign indicator sequence and the symptom feature vector into a time series prediction model to generate a dynamic follow-up evaluation score; semantically analyzing the electronic medical record data of the patient based on a medical knowledge graph, extracting tumor staging classification data, radiotherapy and chemotherapy measurement data, weight and nutrition indicators, and laboratory test data, determining a basic risk coefficient according to the tumor staging classification data, calculating a cumulative dose ratio according to the radiotherapy and chemotherapy measurement data, calculating a body state score according to the weight and nutrition indicators and the laboratory test data, inputting the basic risk coefficient, the cumulative dose ratio and the body state score into a prediction model to generate a baseline follow-up evaluation score; calculating a comprehensive follow-up level and generating a next follow-up plan according to the dynamic follow-up evaluation score and the baseline follow-up evaluation score.

[0007] As a preferred solution of the intelligent remote follow-up method for tumor patients described in the present invention, the standardized processing formula for the real-time vital sign monitoring data is expressed as: ; in, is the historical mean of the jth physical sign indicator; is the historical standard deviation of the jth physical sign indicator; n is the length of the time series; m is the number of physical sign indicators; is the value of the i-th time point in the physical sign index sequence obtained after standardization; It is the original value of the jth vital sign indicator at the ith time point in the real-time vital sign monitoring data.

[0008] The structured processing formula for the symptom quantification data is expressed as: ; in, is the weight coefficient of the kth symptom; p is the number of symptom indicators; T is the length of the observation time window; represents the value of the kth symptom in the symptom quantification data at time t; represents the historical maximum reference value of the kth symptom; Represents the value at time t in the symptom feature vector obtained after structured processing.

[0009] As a preferred solution of the intelligent remote follow-up method for tumor patients described in the present invention, the time series prediction model is constructed based on a convolutional neural network and a long short-term memory network.

[0010] Inputting the physical sign indicator sequence and the symptom feature vector into a time series prediction model to generate a dynamic follow-up evaluation score includes the following steps: in the time series prediction model, fusing the physical sign indicator sequence and the symptom feature vector with time series features: ; in, It represents the fused feature vector at time t in the time series prediction model, and the superscript f represents the fused feature dimension; represents the value of the vital sign index sequence at time t; represents the value of the symptom feature vector at time t; the symbol || represents a vector concatenation operation.

[0011] The convolutional neural network in the time series prediction model is used to extract time series features from the fusion features: ; in, represents the f-dimensional fusion feature sequence from tk to time t; Represents the kth convolution kernel parameter; Represents the convolution layer bias term; K represents the convolution time window size; Represents the c-dimensional features output by the convolutional layer.

[0012] The output features of the convolutional neural network are input into the long short-term memory network to generate the dynamic follow-up assessment score: ; in, represents the dynamic follow-up assessment score; represents the feature mapping function of the lth layer of the long short-term memory network; represents the feature weight coefficient of the lth layer at time t; L represents the number of neural network layers; represents the risk volatility factor; Represents the normalization function.

[0013] If the change rate of the physical sign index sequence is greater than the change rate of the symptom feature vector, then: ; otherwise: ; in, represents the gradient norm of the vital sign index sequence; Represents the gradient norm of the symptom feature vector.

[0014] If the change trends of the physical sign index sequence and the symptom feature vector are consistent, then: ; otherwise: ; Where cov represents the covariance function; Represents the standard deviation of the physical sign indicator sequence; represents the standard deviation of the symptom feature vector.

[0015] As a preferred solution of the intelligent remote follow-up method for tumor patients described in the present invention, the basic risk coefficient is determined according to the tumor staging classification data, and is publicly expressed as: ; in, represents the basic risk coefficient at time t; S represents the number of tumor stage characteristics; represents the weight coefficient of the s-th stage feature; represents the degree index of the s-th stage feature; Represents the time progression factor of the s-th stage characteristic.

[0016] The calculation formula of the cumulative dose ratio is: ; in: It represents the cumulative dose ratio at time t; represents the actual cumulative dose of the kth therapeutic drug; represents the standard cumulative dose of the kth therapeutic drug; M represents the number of drug types; λ represents the time decay coefficient; Indicates the time interval since the last dose.

[0017] The calculation formula of the body status score is: ; in, represents the body state score at time t; N represents the number of state indicators; Indicates the weight of the nth state indicator; represents the normalized value of the nth weight nutrition index; It represents the normalized value of the nth laboratory test index; Represents the indicator volatility variance.

[0018] As a preferred solution of the intelligent remote follow-up method for tumor patients described in the present invention, the basic risk coefficient, the cumulative dose ratio and the body state score are input into a prediction model to generate a baseline follow-up assessment score, which is expressed as follows: ; in, represents the baseline follow-up assessment score at time t; represents the normalization function; ε represents the smoothing factor; γ represents the time adjustment coefficient.

[0019] If the basic risk factor Greater than the basic risk factor of the previous moment , and the cumulative dose ratio Greater than the cumulative dose ratio of the previous moment ,but: ; otherwise .

[0020] If the body status score Less than the body status score at the previous moment ,but: ; otherwise .

[0021] in, Indicates the basic time adjustment coefficient; represents the basic smoothing factor; δ represents the state change threshold.

[0022] As a preferred solution of the intelligent remote follow-up method for tumor patients described in the present invention, wherein: the comprehensive follow-up grade is calculated according to the dynamic follow-up evaluation score and the baseline follow-up evaluation score: ; in, represents the comprehensive follow-up grade at time t; represents the dynamic follow-up assessment score; represents the baseline follow-up assessment score; Represents the dynamic follow-up assessment score sequence from the initial moment to moment t; represents the baseline follow-up assessment score sequence from the initial moment to moment t; ω represents the dynamic assessment weight coefficient; α represents the assessment deviation adjustment coefficient.

[0023] As a preferred solution of the intelligent remote follow-up method for tumor patients described in the present invention, the next follow-up plan is generated based on the comprehensive follow-up level: ; in, represents the next follow-up time interval calculated at time t; represents the basic follow-up time interval; β represents the time adjustment coefficient of the comprehensive follow-up level; λ represents the dynamic change adjustment coefficient; Indicates the dynamic follow-up assessment score at the previous moment.

[0024] To further solve the above technical problems, the present invention provides the following technical solutions: an intelligent remote follow-up system for tumor patients, comprising: a data processing module, used to collect follow-up data through a remote medical terminal, the follow-up data including real-time physical sign monitoring data and symptom quantification data, standardizing the real-time physical sign monitoring data to obtain a physical sign indicator sequence, and structurally processing the symptom quantification data to obtain a symptom feature vector, inputting the physical sign indicator sequence and the symptom feature vector into a time series prediction model to generate a dynamic follow-up evaluation score; an analysis and evaluation module, used to analyze the electronic medical records of the patient based on a medical knowledge graph. The data is semantically analyzed to extract tumor staging classification data, radiotherapy and chemotherapy measurement data, weight and nutrition indicators, and laboratory test data; the basic risk coefficient is determined according to the tumor staging classification data; the cumulative dose ratio is calculated according to the radiotherapy and chemotherapy measurement data; the body state score is calculated according to the weight and nutrition indicators and the laboratory test data; the basic risk coefficient, the cumulative dose ratio and the body state score are input into a prediction model to generate a baseline follow-up assessment score; a follow-up planning module is used to calculate a comprehensive follow-up level and generate a next follow-up plan according to the dynamic follow-up assessment score and the baseline follow-up assessment score.

[0025] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the above-mentioned intelligent remote follow-up method for cancer patients when executing the computer program.

[0026] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of the intelligent remote follow-up method for tumor patients as described above are implemented.

[0027] Beneficial effects of the present invention: The present invention realizes the deep fusion and dynamic feature extraction of multi-source heterogeneous data by constructing a dual analysis engine of a time series prediction model and a medical knowledge graph, and effectively improves the accuracy and timeliness of remote follow-up by establishing feature association rules and a multi-level threshold judgment mechanism. Specifically, the beneficial effects of the present invention are embodied in: on the one hand, through the collaborative analysis of multi-dimensional features and dynamic weight adjustment, the accuracy of patient status assessment is improved; on the other hand, through the adaptive follow-up plan generation mechanism, the precise allocation of medical resources is achieved, effectively improving the quality and efficiency of remote follow-up. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0029] Figure 1 This is a schematic diagram of the overall process of an intelligent remote follow-up method for tumor patients proposed by the present invention; Figure 2 A diagram of computer equipment in an intelligent remote follow-up method for tumor patients proposed by the present invention. DETAILED DESCRIPTION

[0030] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0032] Example 1, reference Figure 1 , which is an embodiment of the present invention, provides an intelligent remote follow-up method for tumor patients.

[0033] Figure 1 The overall process diagram of an intelligent remote follow-up method for cancer patients is shown, including the following steps: S1: Follow-up data are collected through telemedicine terminals. The follow-up data include real-time vital sign monitoring data and symptom quantification data. The real-time vital sign monitoring data are standardized to obtain a vital sign indicator sequence. The symptom quantification data are structured to obtain a symptom feature vector. The vital sign indicator sequence and symptom feature vector are input into the time series prediction model to generate a dynamic follow-up evaluation score.

[0034] S1.1: Collect follow-up data through telemedicine terminals.

[0035] Among them, follow-up data include real-time vital sign monitoring data and symptom quantification data.

[0036] In the data collection phase of the telemedicine terminal, firstly, the real-time vital sign monitoring data, including ECG data, blood oxygen saturation, respiratory rate, body temperature, blood pressure and other physiological indicators, are collected through the intelligent medical device equipped with the physiological parameter sensor module. These data are transmitted to the telemedicine terminal in real time through the Bluetooth low energy protocol, and preliminary data verification and abnormal value screening are carried out in the terminal device to ensure the authenticity and validity of the data.

[0037] In the process of collecting symptom quantitative data, the telemedicine terminal uses the built-in medical-grade electronic scale system to record the patient's 24-hour pain intensity value, average daily sleep duration, daily food intake, daily activity and other objective indicators. A fixed time window data collection strategy is adopted, combined with information such as data collection timestamp and device number, to build a standardized data collection flow, and the collected data is uploaded to the medical data center in real time through an encrypted channel, providing a reliable data foundation for subsequent data processing and analysis.

[0038] S1.2: Standardize the real-time vital sign monitoring data to obtain a vital sign indicator sequence, and structure the symptom quantification data to obtain a symptom feature vector.

[0039] Standardized processing formula for real-time vital sign monitoring data: ; in, is the historical mean of the jth physical sign indicator; is the historical standard deviation of the jth physical sign indicator; n is the length of the time series; m is the number of physical sign indicators; is the value of the i-th time point in the physical sign index sequence obtained after standardization; It is the original value of the jth vital sign indicator at the ith time point in the real-time vital sign monitoring data.

[0040] The structured processing formula for symptom quantitative data is as follows: ; in, is the weight coefficient of the kth symptom; p is the number of symptom indicators; T is the length of the observation time window; represents the value of the kth symptom in the symptom quantification data at time t; represents the historical maximum reference value of the kth symptom; Represents the value at time t in the symptom feature vector obtained after structured processing.

[0041] S1.3: Input the physical sign indicator sequence and symptom feature vector into the time series prediction model to generate a dynamic follow-up assessment score.

[0042] Specifically, the time series prediction model is built based on convolutional neural networks and long short-term memory networks.

[0043] S1.3.1: In the time series prediction model, the time series features of the physical sign indicator sequence and the symptom feature vector are fused: ; in, It represents the fused feature vector at time t in the time series prediction model, and the superscript f represents the fused feature dimension; Represents the value of the vital sign index sequence at time t; Represents the value of the symptom feature vector at time t; the symbol || represents the vector concatenation operation.

[0044] S1.3.2: Convolutional neural network in time series prediction model extracts time series features from fusion features: ; in, represents the f-dimensional fusion feature sequence from tk to time t; Represents the kth convolution kernel parameter; Represents the convolution layer bias term; K represents the convolution time window size; Represents the c-dimensional features output by the convolutional layer.

[0045] S1.3.3: Input the output features of the convolutional neural network into the long short-term memory network to generate dynamic follow-up assessment scores: ; in, represents the dynamic follow-up assessment score; Represents the feature mapping function of the lth layer of the long short-term memory network; represents the feature weight coefficient of the lth layer at time t; L represents the number of neural network layers; represents the risk volatility factor; Represents the normalization function.

[0046] Feature relationship rules: 1. Dynamic adjustment rules for feature weights: If the rate of change of the sign index sequence is greater than the rate of change of the symptom feature vector, then: ; otherwise: ; in, Represents the gradient norm of the vital sign index sequence; Represents the gradient norm of the symptomatic eigenvector.

[0047] 2. Calculation rules for risk volatility factor: If the change trends of the physical sign index sequence and the symptom feature vector are consistent, then: ; otherwise:

[0048] Where cov represents the covariance function; Represents the standard deviation of the physical sign indicator series; represents the standard deviation of the symptom feature vector.

[0049] In this embodiment, the specific implementation of the time series prediction model is described in detail. First, in the feature fusion link, a vector splicing method is used to construct a fusion feature, which maintains the complete time series information of the physical sign indicator sequence and the symptom feature vector. Compared with a simple weighted combination, vector splicing avoids mutual interference between features, making subsequent feature extraction more accurate. For example, when the patient's physical signs such as body temperature and blood pressure change, this fusion method can retain its independent change characteristics and will not be masked by the changes of other features. In the feature extraction link, the present invention realizes local pattern recognition of fusion features. Specifically, the convolution kernel size K=3 is set, that is, feature extraction is performed within a window of 3 time points. This setting can effectively capture the short-term change law of physical sign indicators and symptom characteristics. The nonlinear expression ability of the model is further enhanced by the ReLU activation function, and a more effective feature representation is extracted. In the evaluation score generation link, the hierarchical feature mapping function adopts a two-layer structure: the first layer dimension is 64, and the second layer dimension is 32. This design can extract deep features layer by layer. And the risk volatility factor is introduced so that the evaluation score can respond to abnormal changes in health status in a timely manner.

[0050] In addition, in this embodiment, the dynamic adjustment of feature weights is achieved through the gradient norm ratio. When , it means that the change of physical sign index is more significant, and the weight update formula is The weight of physical sign characteristics will be automatically increased, so that the model pays more attention to physical sign changes. Conversely, when the symptom characteristics change more significantly, the model will increase the weight of symptom characteristics accordingly. The calculation of the risk volatility factor is based on the covariance of physical sign indicators and symptom characteristics, through the formula When the trends of the two types of features are consistent, it indicates that the change in health status has a higher degree of credibility; when the trends are inconsistent, the model will increase the risk factor and increase the vigilance of the assessment. For example, when the physical signs are normal but the symptoms are abnormal, the increase in the risk factor will indicate that there may be potential health risks.

[0051] Through the above implementation scheme, the time series prediction model of the present invention realizes the effective integration and dynamic evaluation of physical signs and symptom characteristics, providing reliable technical support for remote follow-up. Tests show that the model can accurately capture the changing trend of health status and timely discover potential health risks. Compared with the traditional fixed weight model, it has significantly improved the evaluation accuracy and timeliness.

[0052] S2: Based on the medical knowledge graph, semantic analysis is performed on the patient's electronic medical record data to extract tumor staging classification data, radiotherapy and chemotherapy measurement data, weight and nutrition indicators, and laboratory test data. The basic risk coefficient is determined based on the tumor staging classification data, the cumulative dose ratio is calculated based on the radiotherapy and chemotherapy measurement data, and the body state score is calculated based on the weight and nutrition indicators and laboratory test data. The basic risk coefficient, cumulative dose ratio and body state score are input into the prediction model to generate a baseline follow-up assessment score.

[0053] S2.1: Based on the medical knowledge graph, the patient's electronic medical record data is semantically analyzed to extract tumor staging classification data, radiotherapy and chemotherapy measurement data, weight and nutrition indicators, and laboratory test data. The formula is expressed as: ; in, represents the d-dimensional structured medical data extracted at time t, including tumor staging classification data, radiotherapy and chemotherapy measurement data, weight and nutrition indicators, and laboratory test data; G represents the medical knowledge graph; represents the electronic medical record data of the target patient at time t; Represents the semantic analysis mapping function based on the knowledge graph.

[0054] It should be noted that this formula performs semantic analysis on electronic medical record data based on the medical knowledge graph, and can effectively convert unstructured electronic medical record data into standardized structured data. Through the semantic mapping capability of the medical knowledge graph, multi-dimensional clinical characteristics such as tumor staging classification data, radiotherapy and chemotherapy measurement data, weight and nutrition indicators, and laboratory test data can be accurately identified and extracted, avoiding subjective biases and omissions that may occur in traditional manual extraction methods, and improving the accuracy and comprehensiveness of data extraction. At the same time, the formula adopts a time-series processing method, marking data at different times with timestamps, which is helpful for subsequent analysis of the dynamic trend of the patient's condition and provides a reliable data basis for accurate assessment of the patient's condition.

[0055] S2.2: Determine the basic risk factor based on tumor stage classification data and publicly express it as: ; in, represents the basic risk coefficient at time t; S represents the number of stage features in the tumor stage classification data; Represents the weight coefficient of the sth stage feature in the tumor stage classification data; Represents the degree index of the sth stage feature in the tumor stage classification data; Represents the time progression factor of the s-th stage characteristic.

[0056] It should be noted that the calculation formula of the basic risk coefficient introduces the time progression factor on the basis of the traditional tumor staging assessment method. , converting the static staging characteristics into dynamic evaluation indicators. The introduction of It reflects the severity of each staging feature at a specific moment. This multi-dimensional comprehensive calculation method not only takes into account the static characteristics of tumor staging, but also incorporates the temporal characteristics of disease progression, realizing the transformation from discrete staging to continuous risk assessment, significantly improving the accuracy and timeliness of risk assessment. In addition, the use of weighted summation makes the evaluation method well interpretable and scalable, facilitating dynamic adjustment and optimization in clinical practice.

[0057] S2.3: Calculate the cumulative dose ratio based on the chemoradiotherapy data, and calculate the physical status score based on the weight and nutritional indicators and laboratory test data.

[0058] Specifically, the calculation formula for the cumulative dose ratio is: ; in: It represents the cumulative dose ratio at time t; Represents the actual cumulative dose of the kth therapeutic drug in the chemoradiotherapy measurement data; represents the standard cumulative dose of the kth therapeutic drug in the chemoradiotherapy measurement data; M represents the number of drug types in the chemoradiotherapy measurement data; λ represents the time attenuation coefficient of the cumulative dose ratio; Indicates the time interval since the last dose.

[0059] The cumulative dose ratio formula combines the normalized dose ratio and the time decay characteristics, and uses the numerator With the denominator The ratio structure of can realize the unified quantification of the cumulative dose of multiple therapeutic drugs. In particular, the time decay term is introduced , scientifically simulates the metabolic process of drugs in the body, avoiding the limitation of treating drugs administered at different time points as equivalent in the traditional cumulative dose calculation method. Through the dynamic adjustment of the time decay coefficient λ, it not only ensures the continuity of long-term medication history, but also highlights the timeliness of recent medication, making the dose assessment more in line with clinical reality. This time-weighted cumulative dose assessment method provides a more accurate reference for drug administration, helps to adjust the treatment plan in a timely manner, and reduces the risk of drug accumulation.

[0060] The calculation formula of the body status score is: ; in, represents the body status score at time t; N represents the number of status indicators in the weight and nutrition indicators and laboratory test data; Indicates the weight of the nth state indicator; represents the normalized value of the nth weight nutrition index; It represents the normalized value of the nth laboratory test index; Represents the indicator volatility variance.

[0061] It should be noted that the body state scoring formula proposes a new comprehensive evaluation method for body state. The product form of tactfully realizes the coordinated evaluation of weight nutrition index and laboratory test index, overcoming the limitation of independent evaluation of each index in traditional methods. This enables the evaluation system to differentiate weights according to the clinical importance of different indicators, thus improving the pertinence of the evaluation. As a dynamic correction factor, it effectively smooths abnormal fluctuations in indicators and improves the stability of evaluation results. This multi-dimensional indicator fusion evaluation method not only improves the comprehensiveness of body state evaluation, but also ensures the reliability of evaluation results through adaptive adjustment of variance terms, providing a more accurate quantitative reference for understanding the overall condition of patients.

[0062] S2.4: Input the basic risk factor, cumulative dose ratio and body status score into the prediction model to generate the baseline follow-up assessment score, which is expressed as follows: ; in, represents the baseline follow-up assessment score at time t; It represents the normalization function of the time series prediction model; ε represents the smoothing factor of the time series prediction model; γ represents the time adjustment coefficient of the time series prediction model.

[0063] It should be noted that the baseline follow-up assessment score formula is based on the fraction structure The dynamic balance mechanism between risk factors and protective factors is cleverly established, in which the molecular Represents the cumulative effect of disease risk and treatment burden, the denominator reflects the patient's body tolerance. The introduction of ε effectively avoids the calculation singularity when the denominator approaches zero. In particular, through the time adjustment term The evaluation score is gradually adjusted with the follow-up time, avoiding the drastic fluctuation of the evaluation results in the early stage of follow-up. The application of ensures the numerical comparability of the evaluation scores, so that the evaluation results of different patients and different periods have a unified measurement standard. This multi-factor comprehensive and dynamically balanced evaluation mechanism provides a more scientific quantitative basis for the formulation of follow-up plans, and effectively improves the accuracy and reliability of follow-up evaluations.

[0064] If the basic risk factor Greater than the basic risk factor of the previous moment , and the cumulative dose ratio Greater than the cumulative dose ratio of the previous moment ,but: ; otherwise .

[0065] If the body status score Less than the body status score at the previous moment ,but: ; otherwise ; in, Indicates the basic time adjustment coefficient; represents the basic smoothing factor; δ represents the state change threshold.

[0066] Preferably, the present invention realizes accurate capture and timely response to changes in patient status through the above two feature association rules. In the association rule between the basic risk coefficient and the dose ratio, when it is detected that the risk and treatment load increase at the same time, the formula Dynamically adjust the time adjustment coefficient to make the evaluation system respond faster to high-risk conditions. At the same time, in the dynamic adjustment rule of the body state, the threshold δ is introduced as the judgment standard for the deterioration of the body state. When the body state is detected to be significantly deteriorated, the formula Adaptively adjust the smoothing factor to enhance the sensitivity of the evaluation model to changes in the body's state. This dual feature association mechanism not only improves the intelligence level of remote follow-up evaluation, but also realizes the early warning function of changes in the patient's condition, providing reliable decision-making support for clinicians to adjust follow-up strategies in a timely manner, and effectively improving the accuracy and timeliness of remote follow-up of cancer patients. Especially in the case of a sharp deterioration in the body's condition or a sudden increase in treatment risks, the mechanism can quickly adjust the evaluation parameters to ensure that the follow-up evaluation results can accurately reflect the patient's actual condition, providing a scientific basis for medical staff to take timely intervention measures.

[0067] S3: Based on the dynamic follow-up assessment scores and the baseline follow-up assessment scores, calculate the comprehensive follow-up level and generate the next follow-up plan.

[0068] S3.1: Calculate the comprehensive follow-up level based on the dynamic follow-up assessment score and the baseline follow-up assessment score: ; in, represents the comprehensive follow-up grade at time t; represents the dynamic follow-up assessment score; represents the baseline follow-up assessment score; Represents the dynamic follow-up assessment score sequence from the initial moment to moment t; represents the baseline follow-up assessment score sequence from the initial moment to moment t; ω represents the dynamic assessment weight coefficient; α represents the assessment deviation adjustment coefficient.

[0069] If the comprehensive follow-up level If the calculation result meets the following conditions, the corresponding follow-up level and corresponding measures are determined: 1. If the comprehensive follow-up level is The rate of change is less than the stable threshold, and the dynamic follow-up assessment score Baseline follow-up assessment scores If the ratio is close to 1, it is determined to be a stable level; 2. If the comprehensive follow-up level Greater than the comprehensive follow-up level of the previous moment , and the dynamic follow-up assessment score The growth rate exceeded the baseline follow-up assessment score If the growth rate is lower than , it is determined to be at the observation level; 3. If any of the following conditions are met, it is determined to be a warning level: Dynamic follow-up assessment score Two consecutive follow-up assessment scores greater than the baseline The warning threshold; Comprehensive follow-up level The rising rate exceeds the warning threshold and the duration exceeds the warning time window.

[0070] 4. If any of the following conditions are met, it is determined to be an emergency level: Dynamic follow-up assessment score Baseline follow-up assessment scores The deviation term Exceeding the emergency threshold; Three consecutive comprehensive follow-up grades All of them show an accelerating upward trend, and the most recent change rate exceeds the emergency change threshold.

[0071] Among them, the stability threshold is a key parameter for determining the stability of cancer patients' condition. Its calculation fully considers the historical statistical characteristics of the patient's physical sign indicator sequence and symptom feature vector. Specifically, the threshold is based on the dynamic follow-up assessment score. Historical fluctuation range and baseline follow-up assessment scores In practical applications, the standard deviation of the evaluation score within the sliding time window can be used as a benchmark, and dynamic calibration can be performed in combination with the disease progression law in the medical knowledge graph. Preferably, the threshold can be set to 1.5 to 2 times the standard deviation, which can ensure sensitivity to fluctuations in the disease and avoid misjudgment caused by random fluctuations.

[0072] The warning threshold mainly serves the early risk identification during remote follow-up, and its setting is based on the quantitative evaluation criteria for tumor disease progression in clinical medicine. The threshold is determined by analyzing the mutation characteristics in the sign indicator sequence and the clustering characteristics of the symptom feature vector, combined with the comprehensive performance of radiotherapy and chemotherapy measurement data and body status score. The system preferentially uses machine learning methods to analyze the characteristics of warning events in historical patient data, extract the key turning points of risk evolution, and thus establish quantitative standards for warning judgments. In the follow-up scenario of specific cancer types, the threshold can be personalized according to the characteristics of the disease and the treatment plan.

[0073] The emergency change threshold focuses on the early signs of rapid deterioration of the disease, and its setting is based on the second-order derivative characteristics of the dynamic follow-up assessment score. This threshold is mainly determined based on the acceleration characteristics in the time series prediction model. Combined with the development law of clinical critical and severe diseases, it is preferred to set the threshold to 0.15~0.25. Specifically, when it is detected that the acceleration of the change in the assessment score exceeds the threshold range, the emergency response mechanism is triggered. In practical applications, for different types of tumors, fine adjustments can be made within this range according to their progression speed: for relatively slow-progressing tumors (such as thyroid cancer), it can be around 0.15, and for faster-progressing tumors (such as pancreatic cancer), it can be around 0.25 to ensure that the system can capture signs of rapid deterioration of the disease in a timely manner.

[0074] The emergency threshold is the most stringent monitoring parameter in the entire remote follow-up system, and its setting is based on the characteristic patterns of critical events extracted from medical big data analysis. Through retrospective analysis of critical events in historical data, it is preferred to set the threshold to 0.35~0.45. When the deviation between the dynamic follow-up assessment score and the baseline follow-up assessment score exceeds this interval, it indicates that the patient's status has significantly deviated from the expected trajectory. Specifically, 0.40 can be taken as the standard value for general cancer patients, which can be reduced to 0.35 for elderly patients or patients with serious underlying diseases, and can be appropriately increased to 0.45 for patients with better physical fitness and strong tolerance, so as to achieve stratified and precise management.

[0075] It should be noted that the calculation formula for the comprehensive follow-up level proposes a three-dimensional integrated comprehensive follow-up level calculation method. and The normalization of dynamic evaluation and baseline evaluation is realized, overcoming the problem of inconsistent dimensions of the two evaluation scores. Specifically, the deviation term is introduced It is used to capture the difference between dynamic assessment and baseline assessment, and ensures the equivalent treatment of positive and negative deviations through absolute value calculation. This multi-dimensional comprehensive calculation method not only balances the impact of the immediate state and the baseline state, but also realizes the flexible configuration of the evaluation strategy through the adjustment of the weight coefficient ω, providing a more comprehensive and accurate quantitative basis for the hierarchical evaluation in the remote follow-up process.

[0076] S3.2: Generate the next follow-up plan based on the comprehensive follow-up level: ; in, represents the next follow-up time interval calculated at time t; represents the basic follow-up time interval; β represents the time adjustment coefficient of the comprehensive follow-up level; λ represents the dynamic change adjustment coefficient; Indicates the dynamic follow-up assessment score at the previous moment.

[0077] It should be noted that this formula is based on the exponential function The core structure of the method realizes the nonlinear mapping relationship between the follow-up frequency and the comprehensive follow-up level. The present invention introduces the dynamic adjustment item , so that the follow-up plan can be adjusted in real time according to the rate of change of the patient's status, and the follow-up interval can be automatically shortened when a large fluctuation is detected. This adaptive follow-up mechanism based on the dynamic changes of the patient's status effectively solves the problem of insufficient pertinence of traditional fixed-period follow-up plans, avoiding excessive follow-up of stable patients and ensuring that high-risk patients receive timely medical intervention, significantly improving the timeliness of remote follow-up and the efficiency of medical resource utilization.

[0078] In summary, the present invention realizes the deep fusion and dynamic feature extraction of multi-source heterogeneous data by constructing a dual analysis engine of a time series prediction model and a medical knowledge graph, and effectively improves the accuracy and timeliness of remote follow-up by establishing feature association rules and a multi-level threshold judgment mechanism. Specifically, the beneficial effects of the present invention are embodied in: on the one hand, through the collaborative analysis of multi-dimensional features and dynamic weight adjustment, the accuracy of patient status assessment is significantly improved; on the other hand, through the adaptive follow-up plan generation mechanism, the precise allocation of medical resources is achieved, effectively improving the quality and efficiency of remote follow-up.

[0079] Embodiment 2 is an embodiment of the present invention, which provides an intelligent remote follow-up system for tumor patients, including: a data processing module, which is used to collect follow-up data through a remote medical terminal, the follow-up data including real-time physical sign monitoring data and symptom quantification data, standardize the real-time physical sign monitoring data to obtain a physical sign index sequence, structure the symptom quantification data to obtain a symptom feature vector, input the physical sign index sequence and the symptom feature vector into a time series prediction model, and generate a dynamic follow-up evaluation score; The analysis and evaluation module is used to perform semantic analysis on the electronic medical record data of patients based on the medical knowledge graph, extract tumor staging classification data, radiotherapy and chemotherapy measurement data, weight and nutrition indicators, and laboratory test data, determine the basic risk coefficient based on the tumor staging classification data, calculate the cumulative dose ratio based on the radiotherapy and chemotherapy measurement data, calculate the body state score based on the weight and nutrition indicators and laboratory test data, input the basic risk coefficient, cumulative dose ratio and body state score into the prediction model, and generate the baseline follow-up evaluation score; The follow-up planning module is used to calculate the comprehensive follow-up level and generate the next follow-up plan based on the dynamic follow-up assessment score and the baseline follow-up assessment score.

[0080] Example 3, reference Figure 2, is an embodiment of the present invention, which is different from the previous embodiment in that: if the function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0081] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute instructions), or in conjunction with such instruction execution systems, devices or apparatuses. For the purposes of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in conjunction with such instruction execution systems, devices or apparatuses.

[0082] More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk case (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be a paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, deciphering or, if necessary, processing in another suitable manner, and then stored in a computer memory.

[0083] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0084] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. An intelligent remote follow-up method for tumor patients, characterized in that: include: Collecting follow-up data through a remote medical terminal, the follow-up data including real-time physical sign monitoring data and symptom quantification data, performing standardization processing on the real-time physical sign monitoring data to obtain a physical sign index sequence, performing structured processing on the symptom quantification data to obtain a symptom feature vector, inputting the physical sign index sequence and the symptom feature vector into a time series prediction model to generate a dynamic follow-up evaluation score; Based on the medical knowledge graph, semantic analysis is performed on the electronic medical record data of the patient, tumor staging classification data, radiotherapy and chemotherapy measurement data, weight and nutrition indicators, and laboratory test data are extracted, a basic risk coefficient is determined according to the tumor staging classification data, a cumulative dose ratio is calculated according to the radiotherapy and chemotherapy measurement data, a body state score is calculated according to the weight and nutrition indicators and the laboratory test data, the basic risk coefficient, the cumulative dose ratio and the body state score are input into a prediction model, and a baseline follow-up assessment score is generated; According to the dynamic follow-up assessment score and the baseline follow-up assessment score, a comprehensive follow-up grade is calculated and a next follow-up plan is generated.

2. The intelligent remote follow-up method for tumor patients according to claim 1, characterized in that: The standardized processing formula for the real-time vital sign monitoring data is expressed as: ； in, is the historical mean of the jth physical sign indicator; is the historical standard deviation of the jth physical sign indicator; n is the length of the time series; m is the number of physical sign indicators; is the value of the i-th time point in the physical sign index sequence obtained after standardization; is the original value of the jth physical sign indicator at the i-th time point in the real-time physical sign monitoring data; The structured processing formula for the symptom quantification data is expressed as: ； in, is the weight coefficient of the kth symptom; p is the number of symptom indicators; T is the length of the observation time window; represents the value of the kth symptom in the symptom quantification data at time t; represents the historical maximum reference value of the kth symptom; Represents the value at time t in the symptom feature vector obtained after structured processing.

3. The intelligent remote follow-up method for tumor patients according to claim 2, characterized in that: The time series prediction model is constructed based on convolutional neural network and long short-term memory network; Inputting the physical sign indicator sequence and the symptom feature vector into a time series prediction model to generate a dynamic follow-up evaluation score includes the following steps: In the time series prediction model, the physical sign indicator sequence and the symptom feature vector are subjected to time series feature fusion: ； in, It represents the fused feature vector at time t in the time series prediction model, and the superscript f represents the fused feature dimension; represents the value of the vital sign index sequence at time t; represents the value of the symptom feature vector at time t; the symbol || represents a vector concatenation operation; The convolutional neural network in the time series prediction model is used to extract time series features from the fusion features: ； in, represents the f-dimensional fusion feature sequence from tk to time t; Represents the kth convolution kernel parameter; Represents the convolution layer bias term; K represents the convolution time window size; Represents the c-dimensional features output by the convolutional layer; The output features of the convolutional neural network are input into the long short-term memory network to generate the dynamic follow-up assessment score: ； in, represents the dynamic follow-up assessment score; represents the feature mapping function of the lth layer of the long short-term memory network; represents the feature weight coefficient of the lth layer at time t; L represents the number of neural network layers; represents the risk volatility factor; represents the normalization function; If the change rate of the physical sign index sequence is greater than the change rate of the symptom feature vector, then: ； otherwise: ； in, represents the gradient norm of the vital sign index sequence; represents the gradient norm of the symptom feature vector; If the change trends of the physical sign index sequence and the symptom feature vector are consistent, then: ； otherwise: Where cov represents the covariance function; Represents the standard deviation of the physical sign indicator sequence; represents the standard deviation of the symptom feature vector.

4. The intelligent remote follow-up method for tumor patients according to claim 3, characterized in that: The basic risk coefficient is determined according to the tumor staging classification data and is publicly expressed as: ； in, represents the basic risk coefficient at time t; S represents the number of tumor stage characteristics; represents the weight coefficient of the s-th stage feature; represents the degree index of the s-th stage feature; represents the time progression factor of the sth stage characteristics; The calculation formula of the cumulative dose ratio is: ； in: It represents the cumulative dose ratio at time t; represents the actual cumulative dose of the kth therapeutic drug; represents the standard cumulative dose of the kth therapeutic drug; M represents the number of drug types; λ represents the time decay coefficient; Indicates the time interval since the last administration of medication; The calculation formula of the body status score is: ； in, represents the body state score at time t; N represents the number of state indicators; Represents the weight of the nth state indicator; represents the normalized value of the nth weight nutrition index; It represents the normalized value of the nth laboratory test index; Represents the indicator volatility variance.

5. The intelligent remote follow-up method for tumor patients according to claim 4, characterized in that: The basic risk coefficient, the cumulative dose ratio and the body status score are input into the prediction model to generate a baseline follow-up assessment score, which is expressed as follows: ； in, represents the baseline follow-up assessment score at time t; represents the normalization function; ε represents the smoothing factor; γ represents the time adjustment coefficient; If the basic risk factor Greater than the basic risk factor of the previous moment , and the cumulative dose ratio Greater than the cumulative dose ratio of the previous moment ,but: ； otherwise ; If the body status score Less than the body status score at the previous moment ,but: ； otherwise ; in, Indicates the basic time adjustment coefficient; represents the basic smoothing factor; δ represents the state change threshold.

6. The intelligent remote follow-up method for tumor patients according to claim 5, characterized in that: The comprehensive follow-up grade is calculated according to the dynamic follow-up assessment score and the baseline follow-up assessment score: ； in, represents the comprehensive follow-up grade at time t; represents the dynamic follow-up assessment score; represents the baseline follow-up assessment score; Represents the dynamic follow-up assessment score sequence from the initial moment to moment t; represents the baseline follow-up assessment score sequence from the initial moment to moment t; ω represents the dynamic assessment weight coefficient; α represents the assessment deviation adjustment coefficient.

7. The intelligent remote follow-up method for tumor patients according to claim 6, characterized in that: Generate the next follow-up plan based on the comprehensive follow-up level: ； in, represents the next follow-up time interval calculated at time t; represents the basic follow-up time interval; β represents the time adjustment coefficient of the comprehensive follow-up level; λ represents the dynamic change adjustment coefficient; Indicates the dynamic follow-up assessment score at the previous moment.

8. An intelligent remote follow-up system for cancer patients, based on the intelligent remote follow-up method for cancer patients according to any one of claims 1 to 7, characterized in that: include, A data processing module is used to collect follow-up data through a remote medical terminal, wherein the follow-up data includes real-time physical sign monitoring data and symptom quantification data, perform standardization processing on the real-time physical sign monitoring data to obtain a physical sign index sequence, perform structured processing on the symptom quantification data to obtain a symptom feature vector, input the physical sign index sequence and the symptom feature vector into a time series prediction model, and generate a dynamic follow-up evaluation score; an analysis and evaluation module, for performing semantic analysis on the electronic medical record data of the patient based on a medical knowledge graph, extracting tumor staging classification data, radiotherapy and chemotherapy measurement data, weight and nutrition indicators, and laboratory test data, determining a basic risk coefficient according to the tumor staging classification data, calculating a cumulative dose ratio according to the radiotherapy and chemotherapy measurement data, calculating a body state score according to the weight and nutrition indicators and the laboratory test data, inputting the basic risk coefficient, the cumulative dose ratio, and the body state score into a prediction model, and generating a baseline follow-up evaluation score; The follow-up planning module is used to calculate the comprehensive follow-up level and generate the next follow-up plan based on the dynamic follow-up assessment score and the baseline follow-up assessment score.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the intelligent remote follow-up method for tumor patients described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the intelligent remote follow-up method for tumor patients described in any one of claims 1 to 7 are implemented.

Images