Colorectal cancer postoperative symptom group trajectory prediction method and system based on multi-source data

Abstract

The invention discloses a colorectal cancer postoperative symptom group trajectory prediction method and system based on multi-source data, and relates to the technical field of colorectal cancer postoperative management, and the method comprises the steps: obtaining a unique identity label and a static physiological feature of a target patient, querying a historical evolution trend and a historical intervention track of the postoperative symptom group of the target patient according to the unique identity identifier; and acquiring a target intervention measure, and inputting the static physiological features, the historical evolution trend, the historical intervention trajectory and the target intervention measure into a pre-trained trajectory prediction model to generate an expected evolution trend of the target symptom, thereby realizing technical transformation of auxiliary decision making from single-moment risk prediction to continuous trajectory prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of postoperative management of colorectal cancer, in particular to a postoperative symptom cluster trajectory prediction method and system based on multi-source data. BACKGROUND

[0002] Colorectal cancer is one of the common malignant tumors of the digestive system. Patients often have multiple symptoms during the postoperative recovery process, such as abdominal pain, abdominal distension, constipation, fatigue, loss of appetite, etc. Accurate prediction of the occurrence and evolution of postoperative syndromes is beneficial to the evaluation of patient recovery progress and the customization of intervention programs.

[0003] Existing postoperative symptom prediction systems mainly predict the occurrence probability of postoperative symptoms based on the physiological characteristics of patients, such as a postoperative anastomotic fistula prediction method and system for colorectal cancer with application number CN202411463054.1, which screens the input features of the prediction model through multi-factor regression analysis, and constructs an anastomotic fistula occurrence probability prediction model based on a logistic regression model. However, for clinical rehabilitation management, patients and medical staff need to focus on the evolution process of symptoms during the entire recovery period, and the single-time symptom risk is difficult to provide effective support for the continuity management of postoperative symptom clusters. SUMMARY

[0004] The purpose of the present application is to solve the technical problem that the existing technology for predicting single-time symptom risk cannot provide effective support for the continuity management of postoperative symptom clusters, and to provide a postoperative symptom cluster trajectory prediction method and system for colorectal cancer based on multi-source data, which provides the prediction results of the postoperative symptom cluster change trajectory of colorectal cancer patients to achieve the purpose of guiding individualized rehabilitation decision-making.

[0005] According to the first aspect of the present application, a postoperative symptom cluster trajectory prediction method for colorectal cancer based on multi-source data is claimed, comprising: obtaining the unique identity and static physiological characteristics of the target patient, querying the historical evolution trend and historical intervention trajectory of the postoperative symptom cluster of the target patient according to the unique identity; Obtaining the target intervention measure, inputting the static physiological characteristics, historical evolution trend, historical intervention trajectory and target intervention measure into the pre-trained trajectory prediction model to generate the expected evolution trend of the target symptom.

[0006] Preferably, in the step of generating the expected evolution trend of the target symptom, it further comprises: Based on the generalized dynamic Bayesian network, the stage of the patient after the operation is mapped to the hidden state, the symptom evolution trend is mapped to the observation vector, the historical intervention trajectory is taken as the dynamic intervention variable emitted from the hidden state to the observation value, and the static physiological characteristics are taken as the static intervention variable; According to the expectation maximization algorithm, the model parameters of the trajectory prediction model are estimated. inputting the static physiological characteristics, the historical evolution trend, the historical intervention trajectory and the target intervention measure of the target patient into the trajectory prediction model based on recursion; obtaining the expected evolution trend according to the observation vectors of all symptoms at each time point.

[0007] Preferably, in the step of obtaining the target intervention measure, the method further comprises: judging the number of candidate intervention measures in response to the input operation of the candidate intervention measures; if the number of candidate intervention measures is greater than 1, encoding all candidate intervention measures into initial gene sequences, constructing an initial population by crossover and mutation of the initial gene sequences, calculating the risk value of the target patient using the intervention measure corresponding to each individual in the initial population, generating the corresponding fitness according to the risk value, preferentially selecting individuals with higher fitness into the next generation population, generating the remaining individuals by crossover and mutation, and continuing iteration after replacing the initial population with the new population until the iteration times reach a threshold value, and selecting the intervention measure corresponding to the individual with the highest fitness as the target intervention measure.

[0008] Preferably, in the step of calculating the risk value of the individual, the method further comprises: generating the occurrence probability and the expected recovery time length of the candidate symptom group according to the static physiological characteristics, the historical evolution trend, the historical intervention trajectory and the intervention measure corresponding to the individual of the target patient; generating the risk component according to the expected recovery time length of each symptom, weighting and summing the risk components with the occurrence probability of the symptom as the weighting coefficient to generate the risk value of the corresponding individual.

[0009] Preferably, in the step of generating the risk component of the individual, the method further comprises: obtaining the corresponding risk component according to the time length difference between the expected recovery time length and the average recovery time length corresponding to the symptom of the target patient matched with the patient portrait of the target patient.

[0010] Preferably, the value of the risk component of each symptom is obtained from the ratio between the time length difference and a preset maximum difference value, and the time length difference takes the preset maximum difference value only when the symptom type is irreversible, and the corresponding time length difference is less than the preset maximum difference value when the symptom type is reversible.

[0011] Preferably, in the step of weighting and summing all risk components, the target symptom group is generated by taking all candidate symptoms with an occurrence probability not less than a probability threshold, and the weighting and summing of all risk components corresponding to the target symptom group is performed.

[0012] According to the second aspect of the present application, the present application claims to protect a colorectal cancer postoperative symptom group trajectory prediction system based on multi-source data, comprising: The acquisition module is configured to acquire a unique identity of a target patient, static physiological characteristics, and a target intervention measure; The query module is configured to query a historical evolution trend of a postoperative symptom group and a historical intervention trajectory of the target patient according to the unique identity. The generation module is configured to input the static physiological characteristics, the historical evolution trend, the historical intervention trajectory, and the target intervention measure into a pre-trained trajectory prediction model to generate an expected evolution trend of a target symptom.

[0013] Preferably, the generation module further comprises: mapping a stage in which the patient is located after surgery as a hidden state, a symptom evolution trend as an observation vector, and a historical intervention trajectory as a dynamic intervention variable emitted from the hidden state to an observation value, and a static physiological characteristic as a static intervention variable based on a generalized dynamic Bayesian network; estimating model parameters of the trajectory prediction model according to an expectation maximization algorithm; inputting the static physiological characteristics, the historical evolution trend, the historical intervention trajectory, and the target intervention measure of the target patient into the trajectory prediction model based on a recursive formula; obtaining the expected evolution trend according to observation vectors of all symptoms at each time.

[0014] Preferably, the acquisition module further comprises: in response to an input operation of a candidate intervention measure, determining a number of the candidate intervention measures; if the number of the candidate intervention measures is greater than 1, encoding all the candidate intervention measures as an initial gene sequence, constructing an initial population from the initial gene sequence through crossover and mutation, calculating a risk value of the target patient using each individual in the initial population corresponding to an intervention measure, generating a corresponding fitness according to the risk value, preferentially selecting individuals with higher fitness to enter a next generation population, generating remaining individuals through crossover and mutation, replacing the initial population with the new population, and continuing iteration until the number of iterations reaches a threshold value, and then stopping, and selecting an intervention measure corresponding to an individual with the highest fitness as the target intervention measure.

[0015] Preferably, the acquisition module further comprises: generating an occurrence probability and an expected recovery duration of a candidate symptom group according to the static physiological characteristics, the historical evolution trend, the historical intervention trajectory, and the intervention measure corresponding to the individual of the target patient; generating a risk component according to the expected recovery duration of each symptom, weighting and summing the risk components with the occurrence probability of the symptom as a weighting coefficient to generate a risk value of the corresponding individual.

[0016] Preferably, the acquisition module further comprises: According to the patient portrait of the target patient, the patient population is matched with the average recovery time length of corresponding symptoms, and a corresponding risk component is obtained according to a time length difference between the expected recovery time length of each symptom and the corresponding average recovery time length.

[0017] Preferably, the value of the risk component of each symptom is obtained from the ratio between the time length difference and a preset maximum difference value, and the time length difference takes the preset maximum difference value only when the symptom type is an irreversible symptom, and the corresponding time length difference is less than the preset maximum difference value when the symptom type is a reversible symptom.

[0018] Preferably, the obtaining module further comprises generating a target symptom group by taking all candidate symptoms with an occurrence probability not less than a probability threshold, and performing a weighted summation on all risk components corresponding to the target symptom group.

[0019] According to a third aspect of the present application, the present application claims to protect a colorectal cancer postoperative symptom group trajectory prediction device based on multi-source data, comprising a processor and a memory, the memory storing computer readable instructions, when the computer readable instructions are executed by the processor, the steps in the method as described in the first aspect above are executed.

[0020] The present application has the following beneficial effects: 1. By using unique identity and static physiological characteristics, the prediction input is accurately matched with the individual characteristics of the target patient, and by introducing historical symptom evolution trends and historical intervention trajectories, time series context information is provided, so that the model can capture the dynamic rules of the symptom group at different recovery stages, and the static physiological characteristics, target intervention measures and symptom evolution trends are modeled as explicit features, realizing the technical transformation from single-time risk prediction to continuous trajectory prediction, and improving the auxiliary ability of continuous management of the colorectal cancer postoperative symptom group.

[0021] 2. The trajectory prediction model is based on the modeling method of generalized dynamic Bayesian network, which can describe the hidden changes of the patient's postoperative recovery stage and the action mechanism of the target intervention measures under the condition of multi-source data. The symptom group of the colorectal cancer patient after surgery has dynamic characteristics, and the traditional time series prediction is difficult to describe the implicit rules at different recovery stages, and the explanation is insufficient. By introducing hidden state nodes and directed conditional dependence structures, the recovery stage of the patient, the intervention measures and the symptom performance are unified into a probability modeling framework, realizing the dynamic representation of "state transfer-intervention response-symptom emission", so as to realize the prediction of the future symptom trend of the target patient, and also simulate the recovery evolution path, providing an interpretable basis for clinical decision-making. In addition, the model estimates the model parameters by EM algorithm, keeps stable convergence under the condition of limited sample size, has high robustness, and significantly improves the description ability of the postoperative recovery process and the practical application value in the clinical scene.

[0022] 3、By introducing the adaptive optimization mechanism of genetic algorithm, the method can realize the automatic search of global optimal or approximate optimal solution in multiple potential intervention schemes, reduce the deviation caused by artificial experience, improve the intelligent level of multi-scheme intervention selection, and make the intervention measures automatically optimize the configuration of intervention variables according to the individual differences of patients, assisting medical staff to judge the best or approximate best symptom recovery trajectory of patients after operation.

[0023] 4、By introducing the expected recovery time in the fitness calculation method, the risk of the candidate symptom group of the patient is quantified across symptoms. For symptoms such as pain, constipation, diarrhea or fatigue, the system can use the recovery speed to evaluate the severity, and then convert it into a numerical risk component that can be compared across symptoms, so as to accurately reflect the overall symptom compliance and intervention response state of the patient, and solve the problem of comparison difference caused by dimensional difference between different symptoms, so that risk assessment and intervention optimization can be more accurately performed under the condition of cross-symptom group.

[0024] 5、By introducing the average recovery time of the patient group as a reference, the expected recovery time of the target patient is compared with the average recovery time of the group. Compared with the absolute expected recovery time, the system can more accurately identify abnormal recovery of symptoms, and provide more accurate cross-symptom severity comparison results for the genetic algorithm.

[0025] 6、For irreversible symptoms, the expected recovery time and the average recovery time are usually considered as +∞, which will cause the calculation of fitness to fail. By setting a preset maximum difference for irreversible symptoms, the genetic algorithm can optimize the intervention measures according to the occurrence of irreversible symptoms, expand the application scenarios, and improve the practicability.

[0026] 7、By the screening mechanism of probability threshold, only the candidate symptoms with occurrence probability not less than the preset threshold are included in the calculation object of risk value, which effectively reduces the low-probability symptoms caused by data noise, model uncertainty or abnormal input, guides medical staff to pay more attention to high-occurrence symptoms, and improves the robustness of intervention decision and trajectory prediction. BRIEF DESCRIPTION OF DRAWINGS

[0027] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0028] Figure 1 The flowchart of the colorectal cancer postoperative symptom group trajectory prediction method based on multi-source data involved in the embodiments of the present application; Figure 2 The flowchart of the generation step of the expected evolution trend involved in the embodiments of the present application; Figure 3A flowchart of a target intervention measure acquisition process involved in an embodiment of the present application; Figure 4 A structural schematic diagram of a colorectal cancer postoperative symptom cluster trajectory prediction system based on multi-source data involved in an embodiment of the present application; Figure 5 A structural schematic diagram of an electronic device involved in an embodiment of the present application. DETAILED DESCRIPTION

[0029] The present application provides a colorectal cancer postoperative symptom cluster trajectory prediction method and system based on multi-source data. In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. The components of the embodiments of the present application described and shown in the accompanying drawings can be arranged and designed in various different configurations. Therefore, with reference to the terms “one embodiment”, “some embodiments”, “embodiment”, “embodiment”, “illustrative embodiment”, “example”, “specific example” or “some examples”, the detailed description of the embodiments of the present application provided below in the accompanying drawings is not intended to limit the scope of the claimed present application, but only to represent that the specific features, structures or characteristics described in conjunction with the embodiment or example are contained in at least one embodiment or example of the present application. Moreover, the described specific features, structures or characteristics can be combined in any one or more embodiments or examples in a suitable manner. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0030] It should be noted that: similar reference numerals and letters represent similar items in the following drawings, therefore, once an item is defined in one drawing, it does not need to be further defined and explained in subsequent drawings. Meanwhile, in the description of the present application, the terms “first”, “second”, “S1”, “S2” and the like relationship terms are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between the entities or operations.

[0031] According to a first aspect of the present application, the present application claims to protect a colorectal cancer postoperative symptom cluster trajectory prediction method based on multi-source data, with reference to the accompanying drawings, Figure 1 as shown, comprising: S1: obtaining a unique identity of a target patient.

[0032] It should be noted that the unique identity of the target patient can be automatically obtained through a hospital information system (HIS), an electronic medical record system, or a follow-up management platform. The unique identity is used to correspond and associate patient information in different data sources. The unique identity can be a patient number, a medical record number, a hospitalization number, or other feasible encrypted identifiers.

[0033] S2: According to the identity, the evolution trend of the postoperative symptom group of the target patient is queried to obtain the historical evolution trend. The historical evolution trend refers to the change sequence of the characteristics of the symptom group at different time points after the operation. According to the identity, the evolution trend of the postoperative intervention measure of the target patient is queried to obtain the historical intervention trajectory. The historical intervention trajectory refers to the change sequence of the adopted intervention measures at different time points after the operation It should be noted that the historical evolution trend includes a plurality of historical evolution components, and each historical evolution component corresponds to an evolution trend of a characteristic of a postoperative symptom. For example, for the pain symptom, the historical evolution trend can be obtained through the time sequence record of the pain level; for the constipation symptom, the historical evolution trend can be obtained through the time sequence record of the defecation frequency, defecation duration, etc.; for the fatigue symptom, the historical evolution trend can be obtained based on the time sequence record of the evaluation results of the Brief Fatigue Inventory (BFI), and it can not be limited thereto.

[0034] S3: Obtain the static physiological characteristics of the target patient, such as gender, age, medical history, family history, disease name, etc., and it can not be limited thereto.

[0035] S4: Obtain the target intervention measure. The target intervention measure can be manually input by medical staff, can be obtained by an existing intervention measure recommendation system, or can be other feasible implementation manners.

[0036] It should be noted that the target intervention measure includes a plurality of intervention variables, and the intervention variable represents a specific intervention content. For example, the target intervention measure of the first patient includes a first intervention variable, a second intervention variable, and a third intervention variable, the first intervention variable can be to use a first drug, 3 times a day, 2 pills each time, the second intervention variable can be to use a second drug, 2 times a day, 30 milliliters each time, and the third intervention variable can be to use a first device every day, 1 time each time, and each time for 1 hour. The corresponding target intervention measure is obtained by combining the first intervention variable, the second intervention variable, and the third intervention variable. The type of intervention variable includes drug intervention, device intervention, nutritional support, and exercise intervention, and it can not be limited thereto.

[0037] S5: encode and feature fusion operation are performed on the static physiological characteristics, the historical evolution trend of the postoperative symptom group, the historical intervention trajectory and the target intervention measure, and a pre-trained trajectory prediction model is input to generate the expected evolution trend of the target symptom.

[0038] It should be noted that the trajectory prediction model corresponding to each postoperative symptom can be constructed by selecting a recurrent neural network, a convolutional neural network or the like, and the historical patient data collected retrospectively is used as a corresponding data set, and is trained by supervised learning or the like.

[0039] In the embodiment, the trajectory prediction model is constructed based on a generalized dynamic Bayesian network (GDBN), the postoperative process of the patient is modeled as a state transition model at discrete time, i.e., the hidden state (recovery phase) drives the observation (symptom change trend), and the state transition is adjusted by the intervention measure and the static physiological characteristics, and the parameters are learned based on the historical data using the expectation maximization algorithm (EM algorithm), and when the target intervention measure is given, the future pain level distribution is predicted based on the inferred current hidden state, and the expected evolution trend corresponding to the pain level is output. The change trend of the postoperative pain level is described in detail, and the trajectory prediction model is constructed based on the generalized dynamic Bayesian network (GDBN) as shown in the accompanying drawings. Figure 2 Step S5 further includes the following sub-steps: S51, a multi-source data set of a past colorectal cancer patient after surgery is obtained from historical medical records, and the multi-source data set is preprocessed.

[0040] It should be noted that the multi-source data set at least includes static physiological characteristics, time series of symptom observation and time series of intervention measures, and of course it can not be limited thereto.

[0041] In the embodiment, the time step is denoted as t, t = 1, …, T. Wherein T represents the total number of time steps so far. The time difference between adjacent two time steps is the same, which can be obtained by presetting, for example, in units of days, when the time step t = 1, it represents the first day after surgery. The recovery phase of the patient is defined as a hidden state variable, denoted as S, and the recovery phase of the patient at time t is denoted as S t . The characteristics of the patient at time t are denoted as {Y t , A t}. Wherein Y t represents the pain level of the patient at time t, and A t represents the intervention measure adopted by the patient at time t.

[0042] It should be noted that for time t, the patient appears Y tThe corresponding pain level is based on the intervention scheme A adopted at t-1 and before t-1 .

[0043] It should be noted that the hidden state is obtained by pre-setting, which represents the stage of the target patient after the operation, including acute inflammation period, early recovery period, stable period, metabolic imbalance period, complication period, deterioration period, recurrence period and palliative period, of course, the period can not be limited to this. Among them, the acute inflammation period means that the immediate or early inflammatory response after the operation is dominant, which is manifested as the increase of tissue stress and inflammatory indicators; the early recovery period means that the inflammation is weakened, and the nutrition and tissue repair begin, which is manifested as the gradual improvement of symptoms and physiological indicators; the stable period means that the patient enters a relatively stable rehabilitation stage, which is manifested as slight symptoms or close to baseline; the metabolic imbalance period means metabolic abnormalities, such as insufficient nutritional support, affecting wound healing and functional recovery; the complication period means that postoperative infection or anastomotic fistula and other complications occur; the deterioration period means that organ dysfunction or systemic complications occur, and the disease develops in a serious direction; the recurrence period means that the symptoms related to the primary disease, recurrence or metastasis are worsened; the palliative period means that the disease enters a stage mainly for symptom relief and quality of life maintenance, such as home follow-up treatment, etc.

[0044] In the embodiment, the static physiological characteristics are recorded as Z, the z-score is used for completing the standardization operation for continuous class static physiological characteristics such as age, and the one-hot encoding is performed for discrete class static physiological characteristics such as gender and medical history. The pain level is mapped to the score result of 0-10, and the higher the score result is, the more serious the pain is. The category of the intervention measure is one-hot encoded, and the specific daily usage is converted into a standardized continuous value, so as to construct the intervention feature vector.

[0045] S52, a hidden state distribution P(S t ) is randomly initialized for each patient, and the model parameters are randomly valued.

[0046] It should be noted that the model parameters include initial state distribution parameters, state transition parameters and observation emission parameters. Among them, the initial state distribution parameters represent the probability distribution of the initial recovery stage of the patient after the operation for the patient with static physiological characteristics Z, that is, the probability of the patient being in each hidden state on the first day after the operation; the state transition parameters represent the probability of the patient being transferred from the hidden state S t to the next hidden state S t under the condition of performing the intervention measure A t+1 ; and the observation emission parameters represent the conditional probability of generating the pain level Y t under the hidden state S t .

[0047] S53, calculate the posterior probability distribution of each patient in each hidden state under the current parameter value, thereby completing the soft classification of the hidden state, i.e. the probability of each hidden state appearing at each time under the observation of the pain level and the intervention measure, the posterior probability of the ith hidden state under s i can be expressed as: ; where Y 1：T represents the pain level observation sequence from the first day after the surgery to the Tth day; A 1：T-1 represents the intervention measure sequence adopted from the first day after the surgery to the T-1th day; represents the parameter estimation value at the kth iteration of the current iteration.

[0048] It should be noted that the numerator represents the joint probability of observing the current pain trajectory under the hidden state s i , and the denominator represents the total probability of the observation trajectory under all hidden states. The calculation can be obtained by using a forward algorithm, a backward algorithm or a forward-backward algorithm.

[0049] S54, update the initial state distribution parameter by maximizing the expected likelihood.

[0050] In the embodiment, since the body states of different patients are different after the completion of the surgery, the patients present different recovery stages after the surgery, for example, the speed of recovery, the occurrence of complications or the deterioration of the condition at the beginning, etc. Therefore, the initial state distribution parameter π (Z) is modeled in the form of softmax, i.e. ; where π i (Z) represents the value of the initial state distribution parameter in the ith hidden state, i.e. the probability of the patient being in the ith hidden state on the first day after the surgery; exp represents an exponential function; and represents the transposed result of the weight coefficient of the static physiological feature, and K represents the number of hidden states.

[0051] The update is converted into a weighted multinomial logistic regression, i.e. ; where ω i represents the weight coefficient of the static physiological feature; N represents the number of patients in the multi-source data set; represents the posterior probability of the nth patient belonging to the hidden state s i at the initial time; represents the static physiological feature of the nth patient.

[0052] It should be noted that the weight coefficient of the static physiological feature corresponding to the maximum expected likelihood can be obtained by methods such as gradient ascent or standard softmax regression, and then the specific value of the updated initial state distribution parameter can be calculated according to modeling.

[0053] S55, updating the transition probability parameter by maximizing the expected likelihood.

[0054] In the embodiment, to consider the intervention effect and individual difference, the transition probability is modeled as: wherein φ represents a joint feature vector of the intervention measure and the static physiological feature, β ij represents a weight vector of transition from state s i to s j .

[0055] In the embodiment, the state transition parameter is updated based on the maximum expected likelihood, that is: wherein, represents the posterior probability of the n th patient at time t from state s i to state s j . represents the intervention measure of the n th patient at time t.

[0056] It should be noted that the weight coefficient of the state transition corresponding to the maximum expected likelihood can be obtained by methods such as gradient ascent or standard softmax regression, and then the specific value of the updated state transition parameter can be calculated according to modeling.

[0057] S56, updating the observation emission parameter.

[0058] In the embodiment, the pain level is preset as a continuous variable and is modeled by Gaussian distribution. Meanwhile, to reduce the increase of parameter dimension caused by repeated modeling due to the influence of the static physiological feature, when the patient is in the same hidden state, the pain performance distribution is similar, that is, the modeling of the pain level is: wherein, represents the expectation of the j th hidden state, represents the variance of the j th hidden state.

[0059] ​​​​​S57, determine whether the current iteration meets the stopping rule, if yes, complete the parameter training of the trajectory prediction model, if not, return to step S53 and continue to execute.

[0060] It should be noted that the stopping rule can be obtained by pre-setting, for example, the parameter change is lower than the preset change threshold or the iteration number is greater than the preset iteration threshold, of course, it can not be limited to this.

[0061] S58, input the static physiological characteristics of the target patient, the historical evolution trajectory of the pain level, the historical intervention trajectory and the target intervention measure into the trajectory prediction model to output the expected value of the pain level And the corresponding hidden state . And through the recursive method, the expected value and hidden state corresponding to each time in the future preset time period are calculated, that is, the sequence data of the pain level changing over time in the future preset time period is obtained, that is, the expected evolution trend.

[0062] It should be noted that the evolution trend of other symptoms is similar to that of the pain level. After quantifying the evolution of the symptoms, similar steps to step S5 are performed, and again, too much description is not repeated.

[0063] In a feasible embodiment, referring to the accompanying drawings Figure 3 It is shown that in step S4, the method further comprises: S41, in response to the input operation of the candidate intervention measure, determine the number of candidate intervention measures; if the number of candidate intervention measures is 1, continue to execute step S5, if the number of candidate intervention measures is greater than 1, continue to execute from step S42.

[0064] In this embodiment, the candidate intervention measure can be input by the medical staff according to the characteristics of the target patient. After judging and / or optimizing based on the candidate intervention measure input by the medical staff, the target intervention measure is obtained.

[0065] S42, encode all candidate intervention measures into initial gene sequences.

[0066] In this embodiment, all optional intervention variables are used as standard variables, and the corresponding initial gene sequence is generated according to the daily implementation amount of each standard variable in each candidate measure. The initial gene sequence is denoted as G, G=[g1, …, g i , …, g M] , where M represents the number of standard variables, g i represents the daily implementation amount of the candidate intervention measure in the i-th standard variable.

[0067] It should be noted that the standard variable can be preset according to all intervention variables that the colorectal cancer patient can use after surgery.

[0068] S43, construct an initial population by crossing and mutation of the initial gene sequence.

[0069] It should be noted that the crossover operation is used to exchange part of the gene fragments of two initial genes, and single-point crossover, multi-point crossover or uniform crossover can be selected, of course, which can not be limited to this. The mutation operation is used to randomly change some gene fragments in the initial gene sequence within a preset value range, and bit mutation or uniform mutation can be selected, of course, which can not be limited to this. The gene sequence generated by the crossing and mutation operation and the initial gene sequence together constitute the initial population.

[0070] S44, calculate the risk value of the target patient using the intervention measure corresponding to each individual in the initial population.

[0071] It should be noted that the risk value can be obtained according to the symptom aggravation probability and / or adverse reaction score, for example, according to a pre-constructed neural network model to predict the probability of the patient's symptom aggravation after using the intervention measure, or based on a preset drug safety rule to match whether the patient will have an adverse reaction, the probability of occurrence and the severity of adverse reaction, and the risk value is obtained by linear weighting operation.

[0072] S45, generate the corresponding fitness according to the risk value. The higher the risk value, the smaller the corresponding fitness, and the lower the risk value, the greater the corresponding fitness. Linear monotonic decreasing function, exponential decay function, logarithmic function, reciprocal function, etc. Decreasing function can be realized.

[0073] S46, according to the screening rule, select individuals with higher fitness from all the initial population to enter the next generation population. For example, roulette selection strategy or bidding selection strategy, of course, which can not be limited to this.

[0074] S47, generate the remaining individuals by crossing and mutation, replace the initial population with the new population, and continue iteration until the iteration times reach the threshold value. The threshold value of the iteration times can be preset.

[0075] S8, obtain the maximum value of the fitness in the current population, and take the corresponding intervention measure as the target intervention measure.

[0076] In one possible implementation, step S44 further includes: S441, generate the occurrence probability of the candidate symptom group according to the static physiological characteristics of the target patient, the historical evolution trend, the historical intervention trajectory and the intervention measure corresponding to the individual.

[0077] It should be noted that the candidate symptom cluster includes a set of various postoperative symptoms that patients may experience under the current intervention plan. For example, for patients after colorectal cancer surgery, the candidate symptom cluster may include diarrhea, constipation, abdominal pain, fatigue, and bowel dysfunction, but it is not limited to these. The candidate symptom cluster can be obtained by pre-setting all possible postoperative symptoms, matching them based on the records of patients who have visited the hospital in the past, or through other feasible implementation methods.

[0078] In this embodiment, an incidence prediction model for each symptom can be constructed based on a long short-term memory neural network model. The hidden layers capture the time-series dependencies, outputting the probability of occurrence for each candidate symptom. This incidence prediction model can be trained using data from historical patients' medical records.

[0079] S442. Based on the target patient's static physiological characteristics, historical evolution trend, historical intervention trajectory, and individual corresponding intervention measures, generate the expected recovery time of each candidate symptom cluster.

[0080] In this embodiment, a trajectory prediction model is constructed based on a generalized Bayesian network to obtain recovery evaluation rules for each candidate symptom. For example, for pain level, when the pain level is less than 3, the patient's symptoms can be considered recovered; when the pain level is not less than 3, the patient's symptoms are considered not recovered. Then, the patient's features are input into the trajectory prediction model in a recursive manner, and the output predicted value is compared with the recovery evaluation rules. If the recovery evaluation rules are met, the time step of the recursive operation, i.e., the number of iterations, is recorded and used as the expected recovery time. If the recovery evaluation rules are not met, the recursion continues, and the prediction result and the input features are combined to generate new input features, which are then re-input into the trajectory prediction model.

[0081] It should be noted that the specific implementation method of the trajectory prediction model is as described in step S5, and will not be described in detail here.

[0082] S44. Generate a risk component based on the expected recovery time for each symptom. The longer the expected recovery time, the larger the corresponding risk component value; the shorter the expected recovery time, the smaller the corresponding risk component value.

[0083] It should be noted that the expected recovery time can be directly used as the risk component of the symptom, or the normalized mapping result of the expected recovery time can be used as the component.

[0084] S45. Using the probability of symptom occurrence as a weighting coefficient, sum all risk components to generate the risk value for the corresponding individual.

[0085] In this embodiment, the risk value R is calculated as follows: ; Where m represents the number of candidate symptoms; α i r represents the incidence rate of the i-th candidate symptom; i denoted as the expected recovery time for the i-th candidate symptom; f represents the mapping function, where a larger input value results in a larger mapping result, and a smaller input value results in a smaller mapping result; R1 represents the penalty value, used to improve the safety of intervention measures. Adverse reaction matching rules are set according to clinical guidelines and other guiding rules. When a patient experiences an adverse reaction after using the corresponding treatment plan, R1 can accumulate the corresponding penalty value.

[0086] In one feasible implementation, the risk value The calculation method can also be: ; in, This represents the average recovery time for the target patient group at the i-th candidate symptom, which can be calculated based on historical medical records. The larger the difference between the expected recovery time and the average recovery time, the greater the corresponding risk component; the smaller the difference, the smaller the corresponding risk component.

[0087] In one feasible implementation, the risk component of each symptom The value is obtained by the ratio between the duration difference and the preset maximum difference. The duration difference is the preset maximum difference only when the symptom type is irreversible; when the symptom type is reversible, the corresponding duration difference is less than the preset maximum difference. That is: ; Among them, R max This indicates that the preset maximum difference can be obtained through a pre-setting method.

[0088] In one feasible implementation, step S45 further includes generating a target symptom cluster by taking all candidate symptoms with an occurrence probability not less than a probability threshold, and weighting and summing all risk components corresponding to the target symptom cluster.

[0089] It should be noted that the probability threshold can be obtained by presetting, and the probability thresholds for different candidate symptoms can be the same or different.

[0090] According to a second aspect of the present invention, the present invention claims protection for a postoperative symptom cluster trajectory prediction system for colorectal cancer based on multi-source data, as detailed in the appendix. Figure 4 As shown, it includes: The acquisition module is used to acquire the unique identification, static physiological characteristics, and target intervention measures of the target patient; The query module is configured to query the historical evolution trend and the historical intervention trajectory of the postoperative symptom group of the target patient according to the identity. The generation module is configured to input the static physiological feature, the historical evolution trend, the historical intervention trajectory and the target intervention measure into the pre-trained trajectory prediction model to generate the expected evolution trend of the target symptom.

[0091] It should be noted that the acquisition module further includes an identity acquisition submodule, a feature acquisition submodule and an intervention acquisition submodule. The identity acquisition submodule is configured to acquire the unique identity of the target patient, the feature acquisition submodule is configured to acquire the static physiological feature of the target patient, and the intervention acquisition submodule is configured to acquire the target intervention measure of the target patient.

[0092] In a feasible implementation, the generation module further includes: Based on the generalized dynamic Bayesian network, the stage in which the patient is located after the operation is mapped to a hidden state, the symptom evolution trend is mapped to an observation vector, the historical intervention trajectory is taken as a dynamic intervention variable emitted from the hidden state to the observation value, and the static physiological feature is taken as a static intervention variable. The model parameters of the trajectory prediction model are estimated according to the expectation maximization algorithm. The static physiological feature, the historical evolution trend, the historical intervention trajectory and the target intervention measure of the target patient are input into the trajectory prediction model based on the recursive formula. The expected evolution trend is obtained according to the observation vectors of all symptoms at each time.

[0093] Preferably, the intervention acquisition submodule further includes: In response to the input operation of the candidate intervention measure, the number of candidate intervention measures is determined. If the number of candidate intervention measures is greater than 1, all candidate intervention measures are encoded into an initial gene sequence, the initial gene sequence is constructed into an initial population through crossover and mutation, the risk value of the target patient using each individual in the initial population corresponding to the intervention measure is calculated, the corresponding fitness is generated according to the risk value, the individual with higher fitness is preferentially selected into the next generation population, the remaining individuals are generated through crossover and mutation, the new population is used to replace the initial population, and the iteration is continued until the iteration times reach a threshold value, and the intervention measure corresponding to the individual with the highest fitness is selected as the target intervention measure.

[0094] Preferably, the intervention acquisition submodule further includes: According to the static physiological feature, the historical evolution trend, the historical intervention trajectory and the intervention measure corresponding to the individual of the target patient, the occurrence probability and the expected recovery time length of the candidate symptom group are generated, respectively. generate a risk component according to an expected recovery time length of each symptom; and generate a risk value of the individual by weighting and summing the risk components with the occurrence probability of the symptoms as weighting coefficients.

[0095] Preferably, the intervention obtaining submodule further comprises: According to the average recovery time length of the patient population in the corresponding symptom according to the patient portrait of the target patient, the corresponding risk component is obtained according to the time length difference between the expected recovery time length of each symptom and the corresponding average recovery time length.

[0096] Preferably, the value of the risk component of each symptom is obtained by the ratio between the time length difference and a preset maximum difference value, and the time length difference takes the preset maximum difference value only when the symptom type is irreversible, and the corresponding time length difference is less than the preset maximum difference value when the symptom type is reversible.

[0097] Preferably, the intervention obtaining submodule further comprises generating a target symptom group by taking all candidate symptoms whose occurrence probability is not less than a probability threshold, and weighting and summing all risk components corresponding to the target symptom group.

[0098] Referring to the accompanying Figure 5 The embodiments of the present application provide an electronic device, which comprises a processor and a memory, the processor and the memory are interconnected and communicate with each other through a communication bus and / or other forms of connection mechanism (not marked), the memory stores a computer program executable by the processor, when the computing device runs, the processor executes the computer program to execute the system in any optional embodiment of the above-mentioned embodiments.

[0099] The embodiments of the present application provide a storage medium, when the computer program is executed by the processor, the system in any optional embodiment of the above-mentioned embodiments is executed. Wherein, the storage medium can be realized by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk.

[0100] It should be understood that the disclosed system can be implemented in other ways. The above-described system embodiments are merely illustrative. For example, the division of the modules is merely a logical function division. In actual implementation, another division manner can be used. For example, a plurality of modules or components can be combined or integrated into another system, or some features can be omitted or not executed. In addition, the coupling or direct coupling or communication connection between the displayed or discussed systems or units can be indirect coupling or communication connection through some communication interfaces, and can be electrical, mechanical or other forms.

[0101] In addition, the units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0102] In addition, the units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, that is, can be located in one place, or can be distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0103] Flowcharts herein are used to illustrate the steps of the method through the embodiments of the disclosure. It should be understood that the preceding or subsequent steps do not necessarily proceed in sequence. On the contrary, various steps can be evaluated in reverse order or simultaneously. Meanwhile, other operations can be added to these processes.

[0104] Unless otherwise defined, all terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this disclosure belongs. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and not in an idealized or overly formal sense unless expressly so defined herein.

[0105] The above provides a detailed description of the provided method and system for predicting the trajectory of the symptom group after colorectal cancer surgery based on multi-source data. The principles and implementation of the present application are described using specific examples. The above description of the embodiments is only for the embodiments of the present application and is used to help understand the method and system for predicting the trajectory of the symptom group after colorectal cancer surgery based on multi-source data, and does not limit the protection scope of the present application. Meanwhile, the present application can have various modifications and changes for those skilled in the art. Any modification, equivalent replacement made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting the trajectory of postoperative symptom clusters in colorectal cancer based on multi-source data, characterized in that, include: Obtain the unique identifier and static physiological characteristics of the target patient, and query the historical evolution trend and historical intervention trajectory of the postoperative symptom cluster of the target patient based on the unique identifier; To obtain the target intervention, static physiological characteristics, historical evolution trends, historical intervention trajectories, and the target intervention are input into a pre-trained trajectory prediction model to generate the expected evolution trend of the target symptom.

2. The method for predicting postoperative symptom clusters of colorectal cancer based on multi-source data according to claim 1, characterized in that, The steps for generating the desired evolution trend of the target symptom also include: Based on a generalized dynamic Bayesian network, the patient's postoperative stage is mapped to a hidden state, the symptom evolution trend is mapped to an observation vector, the historical intervention trajectory is used as a dynamic intervention variable emitted from the hidden state to the observation value, and the static physiological characteristics are used as a static intervention variable. The model parameters of the trajectory prediction model are estimated using the expectation-maximization algorithm; The trajectory prediction model is based on the recursive input of the target patient's static physiological characteristics, historical evolution trends, historical intervention trajectories, and target intervention measures into the trajectory prediction model. The expected evolution trend is obtained based on the observed vectors of all symptoms at each time point.

3. The method for predicting postoperative symptom cluster trajectory of colorectal cancer based on multi-source data according to claim 2, characterized in that, In obtaining targeted interventions, this also includes: In response to input from candidate interventions, determine the number of candidate interventions; If the number of candidate interventions is greater than 1, all candidate interventions are encoded into initial gene sequences. The initial gene sequences are then used to construct an initial population through crossover and mutation. The risk value of each individual in the initial population corresponding to the intervention is calculated for the target patient. Based on the risk value, the corresponding fitness is generated. Individuals with higher fitness are preferentially selected to enter the next generation population. The remaining individuals are generated through crossover and mutation. The new population replaces the initial population and the iteration continues until the number of iterations reaches a threshold. The intervention corresponding to the individual with the highest fitness is then selected as the target intervention.

4. The method for predicting the trajectory of postoperative symptom clusters in colorectal cancer based on multi-source data according to claim 3, characterized in that, The steps for calculating an individual risk value also include: Based on the target patient's static physiological characteristics, historical evolution trend, historical intervention trajectory, and individual corresponding intervention measures, the probability of occurrence and expected recovery time of candidate symptom clusters are generated respectively. A risk component is generated based on the expected recovery time for each symptom; the risk components are weighted and summed using the probability of symptom occurrence as a weighting coefficient to generate the risk value for the corresponding individual.

5. The method for predicting the trajectory of postoperative symptom clusters in colorectal cancer based on multi-source data according to claim 4, characterized in that, The steps involved in generating individual risk components also include: Based on the patient profile of the target patient, the average recovery time of the patient group for the corresponding symptoms is matched, and the corresponding risk component is obtained based on the difference between the expected recovery time and the corresponding average recovery time.

6. The method for predicting the trajectory of postoperative symptom clusters in colorectal cancer based on multi-source data according to claim 5, characterized in that, The risk component of each symptom is obtained by the ratio between the duration difference and the preset maximum difference. The duration difference is the preset maximum difference only when the symptom type is irreversible. When the symptom type is reversible, the corresponding duration difference is less than the preset maximum difference.

7. The method for predicting postoperative symptom cluster trajectory of colorectal cancer based on multi-source data according to claim 4, characterized in that, In the step of weighted summation of all risk components, all candidate symptoms with an occurrence probability not less than the probability threshold are selected to generate the target symptom cluster, and all risk components corresponding to the target symptom cluster are weighted summation.

8. A postoperative symptom cluster trajectory prediction system for colorectal cancer based on multi-source data, characterized in that, include: The acquisition module is used to acquire the unique identification, static physiological characteristics, and target intervention measures of the target patient; The query module is used to query the historical evolution trend of postoperative symptom clusters and historical intervention trajectory of target patients based on unique identification identifiers; The generation module is used to input static physiological characteristics, historical evolution trends, historical intervention trajectories, and target intervention measures into a pre-trained trajectory prediction model to generate the expected evolution trend of the target symptoms.

Images