Drug clinical trial subject management system and method

Abstract

This invention provides a drug clinical trial subject management system and method. The system includes: a learnable model construction module for building a learnable model; a subject screening module for inputting multi-dimensional characteristics of subjects into the learnable model for analysis and screening to obtain target subjects; an informed consent management module for managing the informed consent signing information of target subjects; a subject grouping module for grouping target subject data; a trial process management module for real-time monitoring and management of post-trial observation data of grouped target subjects; and a learnable model relearning module for retraining and updating the learnable model after the completion of the current clinical trial. This provides more efficient and accurate subject screening and grouping services for clinical trials, helping to improve the success rate and research value of clinical trials, and enhancing the accuracy of model predictions.

Description

Technical Field

[0001] This invention relates to the field of information management technology, and in particular to a drug clinical trial subject management system and method. Background Technology

[0002] Drug clinical trials are a crucial stage in the new drug development process. Their primary purpose is to scientifically and objectively evaluate the safety, efficacy, and adverse reactions of a new drug, providing essential evidence for drug registration. Clinical trials involve human subjects; from the perspective of clinical trial project management and quality, the most critical and challenging aspect is subject management, including subject recruitment, grouping, compliance, dropout, exclusion, and ensuring subject safety. Subject management permeates the entire clinical trial process; effective subject management is a core element, especially subject screening and grouping, as well as subject compliance, which are crucial for ensuring the smooth progress of the trial and the scientific reliability of the results.

[0003] However, the current process of recording informed consent and the management of informed consent forms are mainly based on paper materials and manual management, which is inefficient. The screening and randomization methods are simple and cannot completely eliminate systematic errors.

[0004] Therefore, in order to overcome the above-mentioned technical problems, the present invention provides a drug clinical trial subject management system and method. Summary of the Invention

[0005] This invention provides a drug clinical trial subject management system and method. It constructs a learnable model, inputs multi-dimensional characteristics of subjects into the model for analysis and screening, obtains target subjects, manages informed consent information of target subjects, groups target subject data, and manages post-trial observation data of grouped target subjects in real time. After the completion of each clinical trial, the learnable model is retrained and updated. On the one hand, this provides more efficient and accurate subject screening and grouping services for clinical trials, helping to improve the success rate and research value of clinical trials. On the other hand, the learnable model constructed by this invention has relearning capabilities; after a clinical trial ends, the generated data can be used to further train the learnable model, improving the accuracy of model predictions.

[0006] A drug clinical trial subject management system, comprising:

[0007] The learnable model building module is used to build learnable models;

[0008] The subject screening module is used to input the multi-dimensional characteristics of subjects into a learnable model for analysis and screening to obtain target subjects.

[0009] The informed consent management module is used to manage the informed consent signing information of the target subjects;

[0010] The subject grouping module is used to group the target subject data;

[0011] The trial process management module is used to monitor and manage the post-trial observation data of the target subjects after grouping in real time.

[0012] The learnable model relearning module is used to retrain and update the learnable model after the completion of the current clinical trial.

[0013] Preferably, a drug clinical trial subject management system includes a learnable model building module, comprising:

[0014] The data collection unit is used to collect basic personal information data of participants in historical clinical trials;

[0015] Quantitative scoring units are used for:

[0016] The basic personal information data of participants in historical clinical trials were analyzed, and each characteristic of the clinical trial participants was quantitatively scored based on the analysis results.

[0017] The input vector for constructing a learnable model is based on the first quantization score.

[0018] The clinical trial performance of participants in historical clinical trials is given a second quantitative score, and the results of the second quantitative score are used as the output value of the learnable model.

[0019] The learning unit is used to acquire a preset neural network and learn from the preset neural network based on the input vector and output value to obtain a learnable model.

[0020] Preferably, in a drug clinical trial subject management system, the quantitative scoring unit analyzes the basic personal information data of historical clinical trial participants, including:

[0021] The feature dimension determination subunit is used to analyze the personal basic information data of historical clinical trial participants, and to obtain multiple feature dimensions corresponding to the personal basic information data and the information representation under each feature dimension.

[0022] The first quantitative scoring unit is used to obtain the scoring reference range under each feature dimension, compare the information representation under the corresponding feature dimension with the scoring reference range, and output the first quantitative score corresponding to each feature dimension.

[0023] The input vector determination sub-unit is used to integrate the first quantization score corresponding to each feature dimension to obtain a multi-dimensional feature vector, and use the multi-dimensional feature vector as the input vector of the learnable model.

[0024] Preferably, a drug clinical trial subject management system includes a subject screening module, comprising:

[0025] The input vector determination unit is used to collect and process the basic personal information data of the first few subjects currently recruited, and obtain the multi-dimensional feature input vector of each subject;

[0026] The model analysis unit is used to input multi-dimensional feature input vectors into a learnable model for analysis and output a target score for each subject based on clinical trial performance.

[0027] Filtering unit, used for:

[0028] The target scores of each subject are sorted from highest to lowest to obtain the first ranking sequence of a certain number of subjects;

[0029] Obtain a preset number of participants, and select a second group of participants with high scores in the first sorting sequence based on the preset number of participants. The second group of participants are the target participants.

[0030] Preferably, a drug clinical trial subject management system includes a subject grouping module, comprising:

[0031] The sorting unit is used to read the target score S corresponding to the target subject based on the output of the learnable model, and sort the target scores S of the target subject in descending order to obtain the second sorting sequence.

[0032] Interval partitioning units, used for:

[0033] Divide the target score S into m target intervals, and simultaneously read the minimum target score S. max With the maximum target score S min ;

[0034] The width ΔS of each target interval is calculated based on the minimum and maximum target scores.

[0035]

[0036] Based on the width ΔS of each target interval, the i-th target interval is [S min +(i-1)ΔS,S min +iΔS], where i = 1, 2, ..., m;

[0037] Grouping units, used for:

[0038] Obtain the number of subjects n in each target interval i Within each target interval, a preset number of k samples are extracted according to a preset extraction method. i Target subjects, i.e. Where, k i Pre-set;

[0039] Based on the sampling results, target subjects drawn from different target intervals will be grouped into m combinations.

[0040] Preferably, a drug clinical trial subject management system includes a learnable model relearning module, comprising:

[0041] Retraining units are used for:

[0042] After the completion of the current clinical trial, the target subjects will be manually scored based on their clinical trial performance.

[0043] The system reads the input vector corresponding to the multi-dimensional features of the target subject, and at the same time, the result of the manual scoring of the target subject is used as the output result.

[0044] The learnable model is retrained based on the input vector and output results. At the same time, the weights and biases of the learnable model are updated and saved based on the retraining results.

[0045] Preferably, a drug clinical trial subject management system includes an informed consent management module, comprising:

[0046] The informed consent form generation module is used to retrieve the informed consent form template and improve the subject's content, risks and rights, and subject's rights in the informed consent form template to obtain the target informed consent form.

[0047] The consent form issuance and feedback unit is used for:

[0048] The obtained informed consent forms are sent to the subjects' terminals, and consultation data from the subjects is received in real time. Online interactive consultations and answers are provided to the subjects based on the consultation data.

[0049] Based on the results of online interactive consultations, the signing status of the subject's informed consent form is obtained in real time, and the signed informed consent form is locked after successful signing;

[0050] Informed consent signing management unit, used for:

[0051] Make a backup of the signed target informed consent form, obtain the backup consent form, and record and archive the original signed target informed consent form on the management terminal;

[0052] At the same time, a backup consent form will be sent to the subject's terminal, and the subject will be reminded on the terminal to record and archive it.

[0053] Preferably, a drug clinical trial subject management system further includes a full-process management module, comprising:

[0054] The node determination unit is used to obtain the full-process parameters of drug clinical trials based on the management terminal, divide the entire process into nodes based on the node data aggregation degree of the full-process parameters, and obtain the management node sequence based on the node division results.

[0055] The end-to-end management unit is used for:

[0056] Determine the business execution objectives and business attributes of each management node in the management node sequence, and determine the node business representation of each management node based on the business execution objectives and business attributes;

[0057] Based on the node business representation, a business association identifier is generated between each management node, and based on the business association identifier and the node business representation, a continuous configurable management space is allocated to the management node sequence in the blockchain.

[0058] In the blockchain, data interface interfaces are allocated to configurable management spaces, and the interface parameters of the data interface are unified based on the node business representation of each management node.

[0059] Based on the unified results of interface parameters, the real-time business data of each management node is uploaded to the corresponding configurable management space. At the same time, based on the node business representation, the business subordination relationship between adjacent management nodes in the management node sequence is determined, and based on the business subordination relationship, the data utilization habits of the lower management node to the upper management node are determined.

[0060] Based on data utilization habits, data limitation parameters between adjacent management nodes are generated, and based on the real-time business data upload results, the data limitation parameters of the upper management node are added to the header of the real-time business data of the lower management node.

[0061] Based on the added results, the real-time business data in the configurable management space is dynamically updated according to the data upload timestamp, and the data limitation parameters between adjacent management nodes in each configurable management space are synchronously adjusted.

[0062] The report generation unit is used for:

[0063] Based on the results of the synchronization adjustment, the drug clinical trial is continuously monitored, and a business management report for each node is generated based on the real-time business data in each management node when the entire process is completed.

[0064] Business management reports are fed back to user terminals for visual display.

[0065] Preferably, a drug clinical trial subject management system includes a trial process management module, comprising:

[0066] The prediction unit is used to analyze and predict based on the post-trial observation data of the target subjects, determine whether the current trial process is qualified based on the analysis and prediction results, and generate an alarm report when it is unqualified and transmit the alarm report to the management terminal.

[0067] A method for managing subjects in a drug clinical trial, based on a drug clinical trial subject management system, includes the following steps:

[0068] Step 1: Build a learnable model;

[0069] Step 2: Input the multi-dimensional characteristics of the subjects into the learnable model for analysis and screening to obtain the target subjects;

[0070] Step 3: Manage the informed consent signing information of the target subjects;

[0071] Step 4: Group the target subject data;

[0072] Step 5: Monitor the post-trial observation data of the target subjects after grouping in real time, and manage the observation data;

[0073] Step 6: After completing the current clinical trial, retrain the learnable model and update it.

[0074] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in this application.

[0075] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0076] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0077] Figure 1 This is a structural diagram of a drug clinical trial subject management system according to an embodiment of the present invention;

[0078] Figure 2 This is a diagram of the learnable model construction module in an embodiment of the present invention;

[0079] Figure 3This is a flowchart of the subject screening module in an embodiment of the present invention;

[0080] Figure 4 This is a flowchart of the subject grouping module in an embodiment of the present invention;

[0081] Figure 5 This is a flowchart of the test process management module in an embodiment of the present invention;

[0082] Figure 6 This is a flowchart of the informed consent management module in an embodiment of the present invention;

[0083] Figure 7 This is a flowchart of the learnable model relearning module in an embodiment of the present invention. Detailed Implementation

[0084] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0085] Example 1:

[0086] This embodiment provides a drug clinical trial subject management system, such as... Figure 1 Shown, including:

[0087] The learnable model building module is used to build learnable models;

[0088] The subject screening module is used to input the multi-dimensional characteristics of subjects into a learnable model for analysis and screening to obtain target subjects.

[0089] The informed consent management module is used to manage the informed consent signing information of the target subjects;

[0090] The subject grouping module is used to group the target subject data;

[0091] The trial process management module is used to monitor and manage the post-trial observation data of the target subjects after grouping in real time.

[0092] The learnable model relearning module is used to retrain and update the learnable model after the completion of the current clinical trial.

[0093] In this embodiment, the learning model construction module establishes a continuously trainable mathematical model based on data from previous clinical trial participants using a neural network approach; the participant screening module collects and scores information from recruited participants, generates a "compliance coefficient" score for each participant using the learning model, and screens participants based on the "compliance coefficient" score; the informed consent management module manages the informed consent process, signing, and storage of informed consent forms for trial participants; the participant grouping module groups participants based on their "compliance coefficient" scores using stratified random sampling; the trial process management module manages participants' medication adherence, daily life, visits, regular check-ups, and adverse reactions; and the learning model relearning module manually scores the participants' actual "compliance coefficients" after each clinical trial, using these actual "compliance coefficient" values ​​to retrain the learning model, thereby continuously improving the predictive accuracy of the learning model.

[0094] In this embodiment, post-trial observation data includes: the subject's medication adherence as required and daily life status, visits, regular follow-up examinations, and management of adverse reactions.

[0095] In this embodiment, the trial process management module includes: Step 1, Medication administration and recording: The research physician issues a detailed prescription according to the clinical trial protocol, specifying the subject's medication administration method, including dosage, time interval, and administration method (e.g., before meals, after meals, with warm water, etc.). Through face-to-face communication, distribution of written materials, and necessary video demonstrations, the prescription details precautions during medication administration, such as dietary restrictions and avoiding concurrent medications. This ensures the subject fully understands the medication administration method and precautions. Under the guidance of the research physician, the subject takes the medication on time as prescribed. Simultaneously, the subject needs to diligently record a subject diary card, including the time of each medication administration, post-medication physical sensations (e.g., nausea, dizziness, fatigue, etc.), and any special circumstances in daily life (e.g., strenuous exercise, alcohol consumption, etc.). During medication administration, if the subject discovers any problems or experiences discomfort, they can communicate with the research physician promptly through various channels, such as telephone, text message, and dedicated clinical trial communication software. After receiving feedback from the participants, the research physicians will promptly assess and address the issues. If a participant is found to be unable to continue the clinical trial (e.g., adverse reactions, changes in condition meeting exclusion criteria), that participant will be removed to ensure their safety. The second step involves scheduled follow-up visits: Participants are required to return to the hospital for visits strictly according to the clinical trial protocol. During these visits, the research physicians will issue comprehensive examination and testing orders based on the protocol, such as blood tests, urine tests, and imaging examinations. After the tests are completed, the research physicians will promptly conduct a professional analysis of the results, comparing and interpreting various indicators. If any potential adverse reactions are detected, the research physicians will immediately initiate the appropriate treatment procedures. Depending on the severity of the adverse reaction, different measures will be taken, such as adjusting drug dosage, suspending medication, or providing symptomatic treatment, ensuring that the participant's health status is promptly monitored and addressed. The third step involves recording case reports: During each visit, in addition to conducting routine examinations and analyzing results, the research physicians will inquire in detail about the participant's medication adherence and physical condition, and will check and confirm the participant's diary card. Throughout the process, the research physicians meticulously recorded case report forms, including basic information about the subjects, various data from the trial, examination and test results, problems encountered, and corresponding solutions. The fourth step is data processing: systematically processing the large amount of data generated during the trial. First, data management personnel questioned the data, identifying potential errors, omissions, or contradictions by establishing data logic verification rules and comparing them with original records, and promptly communicating with relevant researchers for verification. After ensuring the data's accuracy, a data locking operation was performed to seal the data and prevent unauthorized modification.After the data is locked, professional statistical analysts will use appropriate statistical analysis methods to conduct in-depth analysis of the data, such as descriptive statistics, correlation analysis, and difference tests, to uncover the underlying information and patterns. Finally, under the conditions specified in the trial protocol, the unblinding process will be performed to reveal the trial groupings. Combined with the statistical analysis results, a complete and accurate clinical trial summary report will be generated. This report will serve as an important basis for evaluating the efficacy and safety of the investigational new drug. The flowchart of the trial process management module is shown below. Figure 5 As shown.

[0096] In this embodiment, the flowchart of the informed consent management module is as follows: Figure 6 As shown, it includes: explanation of the trial protocol, questions from the subjects to the research physician, signing of the informed consent form, and recording and preservation of the informed consent form set and informed consent procedure.

[0097] The working principle and beneficial effects of the above technical solution are as follows: By constructing a learnable model and inputting the multi-dimensional characteristics of the subjects into the learnable model for analysis and screening, target subjects are obtained. The informed consent information of the target subjects is managed, and the target subject data is grouped. The post-trial observation data of the grouped target subjects is monitored and managed in real time. After the completion of the current clinical trial, the learnable model is retrained and updated. On the one hand, this provides more efficient and accurate subject screening and grouping services for clinical trials, helping to improve the success rate and research value of clinical trials. On the other hand, the learnable model constructed by this invention has relearning capabilities. After a clinical trial ends, the generated data can be used to provide the learnable model for further training to improve the accuracy of model predictions.

[0098] Example 2:

[0099] Based on Example 1, this example provides a drug clinical trial subject management system with a learnable model building module, including:

[0100] The data collection unit is used to collect basic personal information data of participants in historical clinical trials;

[0101] Quantitative scoring units are used for:

[0102] The basic personal information data of participants in historical clinical trials were analyzed, and each characteristic of the clinical trial participants was quantitatively scored based on the analysis results.

[0103] The input vector for constructing a learnable model is based on the first quantization score.

[0104] The clinical trial performance of participants in historical clinical trials is given a second quantitative score, and the results of the second quantitative score are used as the output value of the learnable model.

[0105] The learning unit is used to acquire a preset neural network and learn from the preset neural network based on the input vector and output value to obtain a learnable model.

[0106] In this embodiment, the basic personal information data of the clinical subjects includes: age, gender, education level, economic status, health status, and occupation.

[0107] In this embodiment, data related to 1000 previous clinical trial participants were collected, including information such as age, gender, education level, economic status, health status, and occupation. Detailed medical history records were also collected, such as past disease diagnoses, treatment processes, surgical history, and allergy history; as well as various physiological examination indicators, such as complete blood count, urinalysis, liver function, kidney function, electrocardiogram, and imaging examination results. Analysis of this clinical data provides a preliminary understanding of the participants' health status and disease background, offering important evidence for subsequent screening. For example, when collecting data from participants in cardiovascular disease clinical trials, data such as blood pressure, blood lipids, and electrocardiograms were obtained from hospital cardiology medical records; these data directly reflect the participants' cardiovascular health status.

[0108] In this embodiment, the second quantitative score refers to the "compliance coefficient" score of the participants in the historical clinical trials. The score is between 0 and 10, with higher scores for better compliance and lower scores for worse compliance. The "compliance coefficient" scores of these 1,000 participants in the previous clinical trials can be determined as the output value of the learnable model (k = 1, 2, ... 1000).

[0109] In this embodiment, a preset neural network is obtained, and the preset neural network is learned based on the input vector and output value to obtain a learnable model. This includes: establishing a learnable model using a neural network method to generate the "compliance coefficient" for each subject. The neural network-based learnable model includes an input layer, a hidden layer, and an output layer, which are described below:

[0110] a: Input Layer: The input layer contains 6 nodes, each corresponding to a score for one of the six features mentioned above. Each node's value ranges from 0 to 10. Let the input vector be a{x}={x1,x2,x3,x4,x5,x6}, where x1 is the age score, x2 is the gender score, x3 is the education level score, x4 is the income score, x5 is the health status score, and x6 is the occupation score.

[0111] b: Hidden layer: The hidden layer has 10 nodes. The transformation from the input layer to the hidden layer can be represented as:

[0112]

[0113] Where h j Let f be the output of the j-th node in the hidden layer, and f be the activation function. Let x be the weight from the i-th node in the input layer to the j-th node in the hidden layer. i Let x1 be the age score, x2 be the gender score, x3 be the education level score, x4 be the income score, x5 be the health status score, and x6 be the job nature score. This is the bias of the j-th node in the hidden layer.

[0114] In this study, the ReLU (Rectified Linear Unit) function was chosen as the activation function f, and its mathematical expression is as follows:

[0115] f(x) = max(0,x)

[0116] This expression indicates that when the input value x is greater than 0, the output value of the ReLU function is equal to the input value x, meaning the neuron linearly transmits the positive input signal. On one hand, the ReLU function calculation process is extremely simple, requiring only one comparison and selection operation, which greatly reduces computational costs and improves the training speed of the model. On the other hand, during model training, the derivative of the ReLU function is always 1 when the input is positive. This allows the gradient to be effectively propagated during backpropagation, avoiding the vanishing gradient phenomenon, helping the model converge faster, and also enabling better feature extraction from the data, enhancing the model's ability to process complex data.

[0117] c: Output Layer: The output layer has only one node, outputting a value y between 0 and 10. This value is the "compliance coefficient" that this invention needs to obtain. The transformation from the hidden layer to the output layer is as follows:

[0118]

[0119] Where y is the output of the output layer. Let b be the weight from the j-th node in the hidden layer to the output layer. 2 The output layer is biased. Similarly, the output layer also uses the ReLU function as the activation function. The output value y of the output layer ranges from 0 to 10. The ReLU function can ensure that the output value is within a reasonable range, and at the same time, it utilizes its non-linear characteristics to enable the model to better fit the complex relationship between the data and the true "compliance coefficient".

[0120] Learnable model training:

[0121] Collect relevant data from 1000 previous clinical trial participants, including information such as age, gender, education level, economic status, health status, and occupation. Using the method described in the first step, score each of these 1000 participants to establish the model's input vector a{x}k={x1,x2,x3,x4,x5,x6}k, (k=1,2,……1000).

[0122] The "compliance coefficient" of these 1000 previous clinical trial participants was manually scored, with scores between 0 and 10, and these scores were used to determine the true values ​​y that the learnable model could output. k , (k = 1, 2, ..., 1000).

[0123] The model training employs the backpropagation (BP) algorithm. This algorithm calculates the error between the predicted and actual values, then propagates this error back from the output layer to the input layer. This process adjusts the weights and biases of each layer, gradually reducing the error. A loss function is used to measure the error between the model's predictions and the actual values. Based on the application characteristics of this invention, mean squared error is used as the loss function. By minimizing the loss function, the model can continuously adjust its parameters to improve prediction accuracy.

[0124] For the predicted values ​​output by the model and the true value y k (k = 1, 2, ... 1000), the formula for calculating the mean square error is: Where N is the number of samples, and in this invention, the initial training sample size N = 1000. In the learnable model constructed in this invention, we expect the "compliance coefficient" output by the model to be as close as possible to the true "compliance coefficient". Mean squared error can effectively quantify this degree of closeness. By continuously minimizing the mean squared error, the model can learn the complex relationship between various factors and the comprehensive score, thereby achieving accurate screening of subjects.

[0125] During training, gradient descent is used to update the weights. and bias b 2 For weights and bias The updated formula is:

[0126] Where η is the learning rate, which determines the step size of each parameter update. In this invention, the learning rate η = 0.01 is set, which is a suitable value determined after multiple experiments and adjustments. and These represent the loss function L with respect to the weights. and bias The gradient of (l=1,2) is calculated by taking j=1 when calculating the weights and biases of the output layer, since the output layer has only one node. Therefore, j can be omitted in the expression. By calculating the gradient, we can determine in which direction the weights and biases can be adjusted to reduce the loss function value. During backpropagation, the chain rule is used to calculate the gradient of each layer from the output layer to the input layer. By repeatedly calculating the loss function, gradient, and updating the weights and biases, the performance of the learnable model of this invention gradually improves. Thus, using the collected data from 1000 previous clinical trial subjects, the learnable model of this invention is obtained through training and is ready for use. The learnable model construction module diagram is shown below. Figure 2 As shown.

[0127] The working principle and beneficial effects of the above technical solution are as follows: by collecting personal basic information data of historical clinical subjects, the input vector of the learnable model can be effectively obtained; the output value of the learnable model can be effectively obtained through the second quantization scoring; and then the learning of the neural network of the language language can be effectively realized through the input vector and output value, thereby ensuring the accuracy of building the learnable model and helping the learnable model to process complex data.

[0128] Example 3:

[0129] Based on Example 2, this example provides a drug clinical trial subject management system. In the quantitative scoring unit, the system analyzes the basic personal information data of historical clinical trial participants, including:

[0130] The feature dimension determination subunit is used to analyze the personal basic information data of historical clinical trial participants, and to obtain multiple feature dimensions corresponding to the personal basic information data and the information representation under each feature dimension.

[0131] The first quantitative scoring unit is used to obtain the scoring reference range under each feature dimension, compare the information representation under the corresponding feature dimension with the scoring reference range, and output the first quantitative score corresponding to each feature dimension.

[0132] The input vector determination sub-unit is used to integrate the first quantization score corresponding to each feature dimension to obtain a multi-dimensional feature vector, and use the multi-dimensional feature vector as the input vector of the learnable model.

[0133] In this embodiment, the information representation under each feature dimension is the data content corresponding to each feature dimension. The feature dimensions include: age, gender, education level, income, health status, and job nature.

[0134] In this embodiment, a first quantitative score is performed on each feature dimension of the clinical trial subjects. For example, each feature of these 1000 previous clinical trial subjects is quantitatively scored, with a score between 0 and 10. 1) x1 is the age score: age is divided into corresponding scores according to different age groups (e.g., 8 points for 18-30 years old, 6 points for 31-50 years old, etc.); 2) x2 is the gender score: male is set to 5 points, female is set to 5 points (the weight can be adjusted according to actual needs); 3) x3 is the education level score: education level is divided into educational level (2 points for primary school and below, 4 points for junior high school, and 4 points for senior high school / vocational school). 6 points, junior college 8 points, bachelor's degree and above 10 points); 4) x4 is the income score. Income (10 points for annual income of over 1 million, 8 points for 500,000 to 1 million, 6 points for 300,000 to 500,000, 4 points for 100,000 to 300,000, 2 points for 50,000 to 100,000, etc.); 5) x5 is the health score. Health is assessed and scored based on detailed medical examination reports and medical history (10 points for good health, 6 points for minor chronic diseases, etc.); 6) x6 is the job nature score. Job nature is scored based on factors such as work intensity and environment (8 points for easy and good environment, 4 points for high intensity or harsh environment, etc.).

[0135] The working principle and beneficial effects of the above technical solution are as follows: by determining the information representation and scoring reference range corresponding to each feature dimension, it is possible to effectively perform the first quantitative scoring on each feature dimension, and to effectively obtain the input vector based on the first quantitative scoring result, thus ensuring the accuracy of building a learnable model.

[0136] Example 4:

[0137] Based on Example 1, this example provides a drug clinical trial subject management system, including a subject screening module, comprising:

[0138] The input vector determination unit is used to collect and process the basic personal information data of the first few subjects currently recruited, and obtain the multi-dimensional feature input vector of each subject;

[0139] The model analysis unit is used to input multi-dimensional feature input vectors into a learnable model for analysis and output a target score for each subject based on clinical trial performance.

[0140] Filtering unit, used for:

[0141] The target scores of each subject are sorted from highest to lowest to obtain the first ranking sequence of a certain number of subjects;

[0142] Obtain a preset number of participants, and select a second group of participants with high scores in the first sorting sequence based on the preset number of participants. The second group of participants are the target participants.

[0143] In this embodiment, the first number of subjects is greater than or equal to the second number of subjects.

[0144] The working principle of the above technical solution is as follows: The subject screening module consists of the following steps: First, subject information collection and processing: Collect information on all recruited subjects, including age, gender, education level, economic status, health status, and occupation. Then, following the method described in the first step of the "learnable model construction module," obtain the input vector a{x}={x1,x2,x3,x4,x5,x6} of the learnable model for each subject. Second, subject "compliance coefficient" calculation: Input the input vector a{x}={x1,x2,x3,x4,x5,x6} of the learnable model for each subject into the model already constructed in the "learnable model construction module." The "learnable model" calculates its output value y, which is the subject's "compliance coefficient (i.e., target score)," thereby calculating the "compliance coefficient" for all subjects. Third, subject "compliance coefficient" ranking: Rank all subjects according to the magnitude of their "compliance coefficient" values. The fourth step is participant selection: Based on the number of participants determined in the clinical trial plan, participants with high compliance scores are selected according to the ranking results of the "compliance coefficient" from the previous step. Specifically, as follows... Figure 3 As shown.

[0145] The beneficial effects of the above technical solution are: it effectively screens subjects, ensuring the efficiency and accuracy of subject screening, thereby providing reliable data support for subsequent drug clinical trials.

[0146] Example 5:

[0147] Based on Example 1, this example provides a drug clinical trial subject management system, including a subject grouping module, comprising:

[0148] The sorting unit is used to read the target score S corresponding to the target subject based on the output of the learnable model, and sort the target scores S of the target subject in descending order to obtain the second sorting sequence.

[0149] Interval partitioning units, used for:

[0150] Divide the target score S into m target intervals, and simultaneously read the minimum target score S. max With the maximum target score S min ;

[0151] The width ΔS of each target interval is calculated based on the minimum and maximum target scores.

[0152]

[0153] Based on the width ΔS of each target interval, the i-th target interval is [S min +(i-1)ΔS,S min +iΔS], where i = 1, 2, ..., m;

[0154] Grouping units, used for:

[0155] Obtain the number of subjects n in each target interval i Within each target interval, a preset number of k samples are extracted according to a preset extraction method. i Target subjects, i.e. Where, k i Pre-set;

[0156] Based on the sampling results, target subjects drawn from different target intervals will be grouped into m combinations.

[0157] In this embodiment, the preset sampling method is random sampling.

[0158] In this embodiment, k i For pre-set groups, the number of people in each group can be determined according to actual needs and a certain ratio. For example, if it is desired that the number of people in each group is as equal as possible, it can be determined according to... The number of people sampled in each interval is determined by rounding down. The remainder can be randomly assigned to different intervals to ensure that the final total number of people is n.

[0159] In this embodiment, the flowchart of the subject grouping module is as follows: Figure 4 As shown.

[0160] The working principle and beneficial effects of the above technical solution are as follows: By using stratified random sampling to group subjects, it is possible to ensure that the subject scores in each group are relatively evenly distributed, avoiding the situation where subjects with high or low "compliance coefficient" S (set target score) scores are concentrated in the same group, making the grouping results more reasonable and conducive to the subsequent clinical trials and results analysis.

[0161] Example 6:

[0162] Based on Example 1, this example provides a drug clinical trial subject management system with a learnable model relearning module, including:

[0163] Retraining units are used for:

[0164] After the completion of the current clinical trial, the target subjects will be manually scored based on their clinical trial performance.

[0165] The system reads the input vector corresponding to the multi-dimensional features of the target subject, and at the same time, the result of the manual scoring of the target subject is used as the output result.

[0166] The learnable model is retrained based on the input vector and output results. At the same time, the weights and biases of the learnable model are updated and saved based on the retraining results.

[0167] In this embodiment, the flowchart of the learnable model relearning module is as follows: Figure 7 As shown.

[0168] The working principle of the above technical solution is as follows: First, the true value of the subject's "compliance coefficient" is scored: After a clinical trial, each subject participating in the trial is manually scored on the "compliance coefficient", resulting in a true value y for each subject's "compliance coefficient". k (k = 1, 2, ..., total number of participants), and the score of participants who are eliminated or drop out during the trial is uniformly 1 point. The second step is retraining the learnable model: using the method described in "Step 3, Learning Model Training" of the "Learable Model Construction Module", the input vector a{x}k = {x1, x2, x3, x4, x5, x6}k (k = 1, 2, ..., total number of participants, excluding those who failed the screening) of the learnable model for each participant, obtained from "Step 1, Participant Information Collection and Processing" of the "Participant Screening Module", and the true value y of each participant's "compliance coefficient". k (k = 1, 2, ..., total number of participants), and train the model again. The third step is retraining the learnable model: after retraining, the weights in the learnable model can be updated. and bias b 2 The model was saved for use in the next clinical trial.

[0169] The beneficial effect of the above technical solution is that after a clinical trial is completed, the data generated can be used to provide the learnable model for further training, so as to improve the accuracy of the model's predictions.

[0170] Example 7:

[0171] Based on Example 1, this example provides a drug clinical trial subject management system, including an informed consent management module, comprising:

[0172] The informed consent form generation module is used to retrieve the informed consent form template and improve the subject's content, risks and rights, and subject's rights in the informed consent form template to obtain the target informed consent form.

[0173] The consent form issuance and feedback unit is used for:

[0174] The obtained informed consent forms are sent to the subjects' terminals, and consultation data from the subjects is received in real time. Online interactive consultations and answers are provided to the subjects based on the consultation data.

[0175] Based on the results of online interactive consultations, the signing status of the subject's informed consent form is obtained in real time, and the signed informed consent form is locked after successful signing;

[0176] Informed consent signing management unit, used for:

[0177] Make a backup of the signed target informed consent form, obtain the backup consent form, and record and archive the original signed target informed consent form on the management terminal;

[0178] At the same time, a backup consent form will be sent to the subject's terminal, and the subject will be reminded on the terminal to record and archive it.

[0179] In this embodiment, the informed consent form template is pre-built and contains the framework of the informed consent form.

[0180] In this embodiment, the target informed consent form refers to the result obtained after the subject has filled in the trial content, the risks and rights involved, and the rights of the subject in the informed consent form template, which is a report that can be directly signed by the subject.

[0181] In this embodiment, the consultation data refers to the questions asked by the subjects about the experiment, such as details of the experiment.

[0182] In this embodiment, the signing status refers to the subject's signing status of the informed consent form, including signed, unsigned, and the specific content of the signed form.

[0183] The working principle and beneficial effects of the above technical solution are as follows: by retrieving the informed consent form template, the subject content, the existing risks and rights, and the rights of the subject, a target informed consent form is generated. After the obtained target informed consent form is sent to the subject's terminal, the consultation data fed back by the subject is answered in real time. At the same time, after answering, the signing status of the subject's target informed consent form is monitored in real time, so as to determine the final subject based on the signing status. Meanwhile, the signed target informed consent forms are backed up and archived to ensure the reliability of the management of the informed consent signing information of the target subjects.

[0184] Example 8:

[0185] Based on Example 1, this example provides a drug clinical trial subject management system, which also includes a full-process management module, including:

[0186] The node determination unit is used to obtain the full-process parameters of drug clinical trials based on the management terminal, divide the entire process into nodes based on the node data aggregation degree of the full-process parameters, and obtain the management node sequence based on the node division results.

[0187] The end-to-end management unit is used for:

[0188] Determine the business execution objectives and business attributes of each management node in the management node sequence, and determine the node business representation of each management node based on the business execution objectives and business attributes;

[0189] Based on the node business representation, a business association identifier is generated between each management node, and based on the business association identifier and the node business representation, a continuous configurable management space is allocated to the management node sequence in the blockchain.

[0190] In the blockchain, data interface interfaces are allocated to configurable management spaces, and the interface parameters of the data interface are unified based on the node business representation of each management node.

[0191] Based on the unified results of interface parameters, the real-time business data of each management node is uploaded to the corresponding configurable management space. At the same time, based on the node business representation, the business subordination relationship between adjacent management nodes in the management node sequence is determined, and based on the business subordination relationship, the data utilization habits of the lower management node to the upper management node are determined.

[0192] Based on data utilization habits, data limitation parameters between adjacent management nodes are generated, and based on the real-time business data upload results, the data limitation parameters of the upper management node are added to the header of the real-time business data of the lower management node.

[0193] Based on the added results, the real-time business data in the configurable management space is dynamically updated according to the data upload timestamp, and the data limitation parameters between adjacent management nodes in each configurable management space are synchronously adjusted.

[0194] The report generation unit is used for:

[0195] Based on the results of the synchronization adjustment, the drug clinical trial is continuously monitored, and a business management report for each node is generated based on the real-time business data in each management node when the entire process is completed.

[0196] Business management reports are fed back to user terminals for visual display.

[0197] In this embodiment, the full-process parameters refer to all process data from the start to the end of a drug clinical trial.

[0198] In this embodiment, the node data aggregation degree refers to the similarity of data between different stages, that is, the data contained in each node has a high degree of similarity, thereby realizing the node division of the entire process.

[0199] In this embodiment, the management node sequence refers to a series of management nodes obtained by dividing the entire process into nodes according to the aggregation degree of node data, including management nodes for data collection, data transmission, data use, data sharing, and data destruction.

[0200] In this embodiment, the business execution objective refers to the final execution result and purpose that the business corresponding to each management node needs to obtain.

[0201] In this embodiment, the business attribute refers to the type of business executed by each management node.

[0202] In this embodiment, node service representation refers to the characteristics or features of the service presented by each management node.

[0203] In this embodiment, the business association identifier refers to the identifier generated based on the node business representation, which is used to represent the association relationship between different management nodes.

[0204] In this embodiment, configurable management space refers to allocating a corresponding data recording area for each management node in the blockchain, thereby facilitating the storage of the operational data of each management node in the blockchain.

[0205] In this embodiment, the business dependency relationship refers to the interaction or limitation relationship between adjacent management nodes during the execution of their business processes.

[0206] In this embodiment, the superior management node refers to the previous management node in the management node sequence.

[0207] In this embodiment, the lower-level management node refers to the current management node in the management node sequence, which is located after the upper-level management node and adjacent to the upper-level management node.

[0208] In this embodiment, data utilization habits refer to the calling situation or calling habits of the upper management node when the lower management node executes the corresponding data service.

[0209] In this embodiment, the data limiting parameter refers to the specific data that the data of the upper-level management node needs to be presented or play a related role in the lower-level management node.

[0210] In this embodiment, the data upload timestamp refers to the upload time information of real-time business data in each management node.

[0211] The working principle and beneficial effects of the above technical solution are as follows: By determining the full-process parameters of drug clinical trials and analyzing and dividing these parameters into nodes, the sequence of management nodes can be accurately and effectively determined. Simultaneously, each management node in the determined sequence is allocated a corresponding configurable management space in the blockchain. Next, a data interface is added to the configurable management space of each management node, enabling the uploading of real-time business data from each management node to the corresponding management node. Furthermore, the relationships between management nodes are analyzed, and the data between management nodes is linked and bound in the blockchain based on the analysis results, preventing data leakage and tampering. Finally, the drug clinical trials are continuously monitored, and upon completion of the entire process, a business management report is generated for each management node based on the real-time business data. This report is then visualized, facilitating effective viewing of the specific situation at different management nodes by both subjects and administrators, thus ensuring the reliability of subject management in drug clinical trials.

[0212] Example 9:

[0213] Based on Example 1, this example provides a drug clinical trial subject management system, including a trial process management module, comprising:

[0214] The prediction unit is used to analyze and predict based on the post-trial observation data of the target subjects, determine whether the current trial process is qualified based on the analysis and prediction results, and generate an alarm report when it is unqualified and transmit the alarm report to the management terminal.

[0215] The prediction unit includes:

[0216] The data reading subunit is used to read the post-trial observation data of the target subjects and determine the follow-up data in the post-trial observation data;

[0217] The first calculation subunit is used to calculate the follow-up compliance trend prediction value for each group based on the follow-up data;

[0218]

[0219] Among them, F j represents the predicted follow-up compliance trend value for group i; j represents the group number; N tj N represents the actual number of people who should have attended the t-th follow-up visit in group j; (t-1)j A represents the actual number of people who should have attended the (t-1)th follow-up visit in group j; tj A represents the number of people in group j who actually arrived on time during the t-th follow-up visit; (t-1)j S represents the number of people in group j who actually arrived on time during the (t-1)th follow-up visit; tjS represents the number of people in group j who submitted complete and valid data during the t-th follow-up visit; (t-1)j α, β, represent the number of people in group j who submitted complete and valid data at the (t-1)th follow-up visit; This represents the influence weighting coefficient, 0 < α, β, and

[0220] The second calculation subunit is used to calculate the overall compliance trend forecast value based on the follow-up compliance trend forecast value of each group;

[0221]

[0222] Where Z represents the overall compliance trend forecast; m represents the total number of groups;

[0223] The pass / fail determination unit is used for:

[0224] Obtain the preset prediction threshold and compare the comprehensive compliance trend prediction value with the preset prediction value to determine whether the current trial process is qualified.

[0225] When the overall compliance trend prediction is equal to or greater than the preset prediction threshold, the current test process is deemed qualified.

[0226] Otherwise, the current testing process is deemed unqualified;

[0227] The alarm subunit is used to generate an alarm report and transmit the alarm report to the management terminal when the current test process is determined to be unqualified.

[0228] In this embodiment, the preset prediction threshold is set in advance and used as a standard to measure whether the current test process is qualified.

[0229] In this embodiment, the alarm report refers to a report automatically generated to remind the administrator when the current test process fails. When the administrator receives the alarm report, he can intervene in advance to ensure the smooth progress of the test.

[0230] In this embodiment, the management terminal is the administrator's operating terminal, including but not limited to mobile phones, computers, etc.

[0231] The working principle and beneficial effects of the above technical solution are as follows: by reading follow-up data, the accurate calculation of the follow-up compliance prediction value for each group can be effectively achieved. The comprehensive compliance trend prediction value can be effectively calculated from the follow-up compliance trend prediction value for each group, thereby ensuring the effectiveness and completeness of the evaluation of the current trial process. Furthermore, when the trial process is unqualified, the alarm report can enable the administrator to intervene in advance, providing reliable data support and guarantee for the smooth completion of drug clinical trials.

[0232] Example 10:

[0233] A method for managing subjects in a drug clinical trial, based on the aforementioned drug clinical trial subject management system, implements the following steps:

[0234] Step 1: Build a learnable model;

[0235] Step 2: Input the multi-dimensional characteristics of the subjects into the learnable model for analysis and screening to obtain the target subjects;

[0236] Step 3: Manage the informed consent signing information of the target subjects;

[0237] Step 4: Group the target subject data;

[0238] Step 5: Monitor the post-trial observation data of the target subjects after grouping in real time, and manage the observation data;

[0239] Step 6: After completing the current clinical trial, retrain the learnable model and update it.

[0240] The working principle and beneficial effects of the above technical solution are as follows: By constructing a learnable model and inputting the multi-dimensional characteristics of the subjects into the learnable model for analysis and screening, target subjects are obtained. The informed consent information of the target subjects is managed, and the target subject data is grouped. The post-trial observation data of the grouped target subjects is monitored and managed in real time. After the completion of the current clinical trial, the learnable model is retrained and updated. On the one hand, this provides more efficient and accurate subject screening and grouping services for clinical trials, helping to improve the success rate and research value of clinical trials. On the other hand, the learnable model constructed by this invention has relearning capabilities. After a clinical trial ends, the generated data can be used to provide the learnable model for further training to improve the accuracy of model predictions.

[0241] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A drug clinical trial subject management system, characterized in that, include: The learnable model building module is used to build learnable models; The subject screening module is used to input the multi-dimensional characteristics of subjects into a learnable model for analysis and screening to obtain target subjects. The informed consent management module is used to manage the informed consent signing information of the target subjects; The subject grouping module is used to group the target subject data; The trial process management module is used to monitor and manage the post-trial observation data of the target subjects after grouping in real time. The learnable model relearning module is used to retrain and update the learnable model after the completion of the current clinical trial. It also includes a full-process management module, including: The node determination unit is used to obtain the full-process parameters of drug clinical trials based on the management terminal, divide the entire process into nodes based on the node data aggregation degree of the full-process parameters, and obtain the management node sequence based on the node division results. The end-to-end management unit is used for: Determine the business execution objectives and business attributes of each management node in the management node sequence, and determine the node business representation of each management node based on the business execution objectives and business attributes; Based on the node business representation, a business association identifier is generated between each management node, and based on the business association identifier and the node business representation, a continuous configurable management space is allocated to the management node sequence in the blockchain. In the blockchain, data interface interfaces are allocated to configurable management spaces, and the interface parameters of the data interface are unified based on the node business representation of each management node. Based on the unified results of interface parameters, the real-time business data of each management node is uploaded to the corresponding configurable management space. At the same time, based on the node business representation, the business subordination relationship between adjacent management nodes in the management node sequence is determined, and based on the business subordination relationship, the data utilization habits of the lower management node to the upper management node are determined. Based on data utilization habits, data limitation parameters between adjacent management nodes are generated, and based on the real-time business data upload results, the data limitation parameters of the upper management node are added to the header of the real-time business data of the lower management node. Based on the added results, the real-time business data in the configurable management space is dynamically updated according to the data upload timestamp, and the data limitation parameters between adjacent management nodes in each configurable management space are synchronously adjusted. The report generation unit is used for: Based on the results of the synchronization adjustment, the drug clinical trial is continuously monitored, and a business management report for each node is generated based on the real-time business data in each management node when the entire process is completed. Business management reports are fed back to user terminals for visual display.

2. The drug clinical trial subject management system according to claim 1, characterized in that, Learnable model building modules include: The data collection unit is used to collect basic personal information data of participants in historical clinical trials; Quantitative scoring units are used for: The basic personal information data of participants in historical clinical trials were analyzed, and each characteristic of the clinical trial participants was quantitatively scored based on the analysis results. The input vector for constructing a learnable model is based on the first quantization score. The clinical trial performance of participants in historical clinical trials is given a second quantitative score, and the results of the second quantitative score are used as the output value of the learnable model. The learning unit is used to acquire a preset neural network and learn from the preset neural network based on the input vector and output value to obtain a learnable model.

3. The drug clinical trial subject management system according to claim 2, characterized in that, The quantitative scoring unit analyzes the basic personal information data of participants in historical clinical trials, including: The feature dimension determination subunit is used to analyze the personal basic information data of historical clinical trial participants, and to obtain multiple feature dimensions corresponding to the personal basic information data and the information representation under each feature dimension. The first quantitative scoring unit is used to obtain the scoring reference range under each feature dimension, compare the information representation under the corresponding feature dimension with the scoring reference range, and output the first quantitative score corresponding to each feature dimension. The input vector determination sub-unit is used to integrate the first quantization score corresponding to each feature dimension to obtain a multi-dimensional feature vector, and use the multi-dimensional feature vector as the input vector of the learnable model.

4. The drug clinical trial subject management system according to claim 1, characterized in that, The subject screening module includes: The input vector determination unit is used to collect and process the basic personal information data of the first few subjects currently recruited, and obtain the multi-dimensional feature input vector of each subject; The model analysis unit is used to input multi-dimensional feature input vectors into a learnable model for analysis and output a target score for each subject based on clinical trial performance. Filtering unit, used for: The target scores of each subject are sorted from highest to lowest to obtain the first ranking sequence of a certain number of subjects; Obtain a preset number of participants, and select a second group of participants with high scores in the first sorting sequence based on the preset number of participants. The second group of participants are the target participants.

5. A drug clinical trial subject management system according to claim 1, characterized in that, The subject grouping module includes: The sorting unit is used to read the target score S corresponding to the target subject based on the output of the learnable model, and sort the target scores S of the target subject in descending order to obtain the second sorting sequence. Interval partitioning units, used for: Divide the target score S into m target intervals, and simultaneously read the minimum target score S. max With the maximum target score S min ; The width ΔS of each target interval is calculated based on the minimum and maximum target scores. Based on the width ΔS of each target interval, the i-th target interval is [S min +(i-1)ΔS,S min +iΔS], where i = 1, 2, ..., m; Grouping units, used for: Obtain the number of subjects n in each target interval i Within each target interval, a preset number of k samples are extracted according to a preset extraction method. i Target subjects, i.e. Among them, k i Pre-set; Based on the sampling results, target subjects drawn from different target intervals will be grouped into m combinations.

6. A drug clinical trial subject management system according to claim 1, characterized in that, The learnable model relearning module includes: Retraining units are used for: After the completion of the current clinical trial, the target subjects will be manually scored based on their clinical trial performance. The system reads the input vector corresponding to the multi-dimensional features of the target subject, and at the same time, the result of the manual scoring of the target subject is used as the output result. The learnable model is retrained based on the input vector and output results. At the same time, the weights and biases of the learnable model are updated and saved based on the retraining results.

7. A drug clinical trial subject management system according to claim 1, characterized in that, The informed consent management module includes: The informed consent form generation module is used to retrieve the informed consent form template and improve the subject's content, risks and rights, and subject's rights in the informed consent form template to obtain the target informed consent form. The consent form issuance and feedback unit is used for: The obtained informed consent forms are sent to the subjects' terminals, and consultation data from the subjects is received in real time. Online interactive consultations and answers are provided to the subjects based on the consultation data. Based on the results of online interactive consultations, the signing status of the subject's informed consent form is obtained in real time, and the signed informed consent form is locked after successful signing; Informed consent signing management unit, used for: Make a backup of the signed target informed consent form, obtain the backup consent form, and record and archive the original signed target informed consent form on the management terminal; At the same time, a backup consent form will be sent to the subject's terminal, and the subject will be reminded on the terminal to record and archive it.

8. A drug clinical trial subject management system according to claim 1, characterized in that, The test process management module includes: The prediction unit is used to analyze and predict based on the post-trial observation data of the target subjects, determine whether the current trial process is qualified based on the analysis and prediction results, and generate an alarm report when it is unqualified and transmit the alarm report to the management terminal.

9. A method for managing subjects in a drug clinical trial, characterized in that, The following steps are implemented based on a drug clinical trial subject management system as described in any one of claims 1-8: Step 1: Build a learnable model; Step 2: Input the multi-dimensional characteristics of the subjects into the learnable model for analysis and screening to obtain the target subjects; Step 3: Manage the informed consent signing information of the target subjects; Step 4: Group the target subject data; Step 5: Monitor the post-trial observation data of the target subjects after grouping in real time, and manage the observation data; Step 6: After completing the current clinical trial, retrain the learnable model and update it.

Images