Clinical research scheme optimization method and system based on AI

Abstract

The invention provides an AI-based clinical research scheme optimization method and system, and relates to the technical field of clinical research, and the method comprises the steps: collecting current clinical test data, and extracting basic features and derivative features related to the recruitment efficiency and shedding rate of a patient from the current clinical test data; the extracted basic features and derivative features are input into a plurality of different pre-trained multivariable regression models at the same time, preliminary optimization suggestions are output through the multivariable regression models, and the multiple different multivariable regression models comprise a multiple linear regression base model, a Logistic regression base model and a random forest base model; inputting the preliminary optimization suggestions output by each multivariable regression model into a pre-trained meta-model, and generating a final optimization suggestion by integrating the multiple preliminary optimization suggestions by the meta-model; and based on the final optimization suggestion, executing the optimization of the clinical research scheme, and achieving the purposes of improving the recruitment efficiency of the patient and reducing the falling-off rate.

Description

Technical Field

[0001] The present invention relates to the field of clinical research technology, and in particular to an optimization method and system for a clinical research program based on AI. Background Art

[0002] In the field of clinical research, patient recruitment efficiency and dropout rate have always been key factors affecting research progress and the accuracy of results. Inefficient patient recruitment may lead to extended research cycles and increased costs, while high dropout rates may weaken the reliability and effectiveness of research results. Traditional clinical research program optimization methods often rely on the experience and intuition of researchers, lack of systematicity and scientificity, and are difficult to accurately solve the problems of patient recruitment and dropout.

[0003] With the rapid development of artificial intelligence (AI) technology, its application in the medical field is becoming more and more extensive, providing new possibilities for the optimization of clinical research programs. AI technology can dig out the laws and trends hidden in complex clinical data through big data analysis and machine learning algorithms, and provide more scientific and accurate decision support for clinical research.

[0004] Therefore, it is necessary to provide an AI-based optimization method for clinical research programs and its system to solve the above technical problems. Summary of the invention

[0005] In order to solve the above technical problems, the present invention provides an AI-based optimization method and system for clinical research plans, which utilizes the advantages of AI technology in data processing and decision support, and realizes precise optimization of clinical research plans by constructing multiple regression models and meta-models, thereby significantly improving patient recruitment efficiency and reducing dropout rates.

[0006] The present invention provides an optimization method for a clinical research plan based on AI, aiming to improve the efficiency of patient recruitment and reduce the dropout rate, and the optimization method comprises the following steps:

[0007] Collecting current clinical trial data, and extracting basic features and derived features associated with patient recruitment efficiency and dropout rate from the current clinical trial data;

[0008] The extracted basic features and derived features are simultaneously input into a plurality of pre-trained different multivariate regression models, and each multivariate regression model outputs a preliminary optimization suggestion, wherein the plurality of different multivariate regression models include a multivariate linear regression base model, a logistic regression base model, and a random forest base model;

[0009] The preliminary optimization suggestions output by each multivariate regression model are input into the pre-trained meta-model, and the meta-model generates the final optimization suggestions by integrating multiple preliminary optimization suggestions;

[0010] Based on the final optimization suggestions, the clinical research plan is optimized to improve patient recruitment efficiency and reduce dropout rate.

[0011] Preferably, the collecting of current clinical trial data and extracting basic features and derived features associated with patient recruitment efficiency and dropout rate from the current clinical trial data include:

[0012] Integrate current data from multiple clinical trial databases;

[0013] Preprocessing the integrated current data, wherein the preprocessing includes data cleaning, missing value filling and outlier processing;

[0014] Using feature selection algorithms to identify basic features associated with patient recruitment efficiency and dropout rate, where the basic features include patient age, gender, disease type, and past medical history;

[0015] Based on the basic features, derived features are constructed by statistical analysis methods, wherein the derived features include disease severity scores and treatment response prediction indexes.

[0016] Preferably, the extracted basic features and derived features are simultaneously input into a plurality of pre-trained different multivariate regression models, and each multivariate regression model outputs preliminary optimization suggestions, including:

[0017] Combining the basic features and the derived features into a feature vector;

[0018] The feature vector is simultaneously input into a pre-trained multivariate linear regression base model, a logistic regression base model, and a random forest base model;

[0019] Based on the input feature vector, each model uses its own learning algorithm and parameter settings to calculate preliminary optimization suggestions related to patient recruitment efficiency and dropout rate. Among them, the multivariate linear regression base model provides predictions and optimization suggestions based on linear relationships, the logistic regression base model provides probability-based classification and optimization suggestions, and the random forest base model provides optimization suggestions through integrated learning.

[0020] Preferably, the pre-training process of the multivariate linear regression base model, the logistic regression base model and the random forest base model includes:

[0021] Collecting historical clinical trial data, and dividing the historical clinical trial data into a training data set and a validation data set;

[0022] For the multivariate linear regression base model, the historical basic features and historical derived features in the training data set were used as input variables, and the patient recruitment efficiency and dropout rate were used as target variables. Linear regression analysis was performed, and the trained multivariate linear regression base model was obtained by adjusting the model parameters to minimize the prediction error;

[0023] For the Logistic regression base model, the historical basic features and historical derived features in the training data set were used as input variables, and the patient recruitment efficiency and dropout rate were used as target variables and converted into binary forms. Logistic regression analysis was performed, and the trained Logistic regression base model was obtained by maximizing the log-likelihood function.

[0024] For the random forest base model, the historical basic features and historical derived features in the training data set are used as input variables, multiple decision trees are constructed as base learners, and the optimal number and depth of trees are selected through the cross-validation method to obtain the trained random forest base model.

[0025] Preferably, the preliminary optimization suggestions output by each multivariate regression model are input into a pre-trained meta-model, and the meta-model generates a final optimization suggestion by integrating multiple preliminary optimization suggestions, including:

[0026] Receive preliminary optimization suggestions from the multivariate linear regression base model, the logistic regression base model, and the random forest base model;

[0027] Inputting the preliminary optimization suggestion as an input variable into a pre-trained meta-model;

[0028] The meta-model uses its own learning algorithm and parameter settings to perform comprehensive processing on multiple preliminary optimization suggestions, wherein the comprehensive processing includes evaluating the consistency score, complementarity degree and reliability score of each preliminary optimization suggestion;

[0029] Based on the results of the comprehensive processing, the meta-model generates a final optimization suggestion that combines the advantages of multiple preliminary optimization suggestions, and outputs the final optimization suggestion.

[0030] Preferably, the training process of the pre-trained meta-model includes:

[0031] Collect multiple historical clinical research plans and their corresponding optimization results as training data sets, where each clinical research plan includes basic features, derived features, and patient recruitment efficiency and dropout rate data obtained after implementation;

[0032] For each clinical research plan, multiple historical preliminary optimization suggestions are generated based on the patient recruitment efficiency and dropout rate data after implementation, as well as the prediction results of the multivariate regression model;

[0033] The generated multiple historical preliminary optimization suggestions and their corresponding historical clinical research plans and optimization results are used as input variables and target variables to build and train a meta-model, wherein the meta-model is learned using a machine learning algorithm, and a trained meta-model is obtained by minimizing the prediction error.

[0034] Preferably, the optimization of the clinical research plan is performed based on the final optimization suggestion to achieve the purpose of improving patient recruitment efficiency and reducing dropout rate, including:

[0035] Analyze the final optimization recommendations output by the meta-model and identify measures to improve patient recruitment efficiency and dropout rate;

[0036] Adjust the recruitment strategy of the clinical research program based on the identified measures, where the recruitment strategy includes optimizing recruitment materials, broadening recruitment channels, adjusting patient screening criteria, and strengthening patient education and support;

[0037] At the same time, according to the final optimization recommendations, the treatment and management processes in the clinical research plan were adjusted to reduce the patient dropout rate. The adjustments to the treatment and management processes included improving the treatment plan, measures to enhance patient compliance, and providing psychological and social support;

[0038] Implement the adjusted clinical research plan.

[0039] The present invention also provides an AI-based clinical research program optimization system for executing the AI-based clinical research program optimization method, the optimization system comprising:

[0040] A feature extraction module, used to collect current clinical trial data and extract basic features and derived features associated with patient recruitment efficiency and dropout rate from the current clinical trial data;

[0041] A preliminary optimization suggestion generating module, used for simultaneously inputting the extracted basic features and derived features into a plurality of pre-trained different multivariate regression models, and outputting preliminary optimization suggestions from each multivariate regression model, wherein the plurality of different multivariate regression models include a multivariate linear regression base model, a logistic regression base model and a random forest base model;

[0042] The final optimization suggestion generation module is used to input the preliminary optimization suggestions output by each multivariate regression model into the pre-trained meta-model, and the meta-model generates the final optimization suggestion by integrating multiple preliminary optimization suggestions;

[0043] The program execution module is used to optimize the clinical research program based on the final optimization suggestion, so as to improve the patient recruitment efficiency and reduce the dropout rate.

[0044] Compared with related technologies, the AI-based clinical research program optimization method and system provided by the present invention have the following beneficial effects:

[0045] The present invention collects current clinical trial data, extracts basic features and derived features associated with patient recruitment efficiency and dropout rate, and then uses a plurality of different pre-trained multivariate regression models (including a multivariate linear regression base model, a logistic regression base model, and a random forest base model) to analyze the extracted features and output preliminary optimization suggestions. These preliminary optimization suggestions are then input into the pre-trained meta-model, which generates final optimization suggestions by integrating a plurality of preliminary optimization suggestions. Finally, based on these final optimization suggestions, the optimization of the clinical research plan is performed to achieve the purpose of improving patient recruitment efficiency and reducing dropout rate. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 A flowchart of an optimization method for an AI-based clinical research program provided by the present invention;

[0047] Figure 2 A schematic diagram of the module structure of an optimization system for an AI-based clinical research program provided by the present invention. DETAILED DESCRIPTION

[0048] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only the parts related to the present invention, rather than all structures, are shown in the accompanying drawings. In addition, the embodiments of the present invention and the features in the embodiments may be combined with each other without conflict.

[0049] It should also be noted that, for ease of description, only the part relevant to the present invention but not all content is shown in the accompanying drawings. It should be mentioned before discussing exemplary embodiments in more detail that some exemplary embodiments are described as processing or methods depicted as flow charts. Although the flow chart describes each operation (or step) as sequential processing, many operations therein can be implemented in parallel, concurrently or simultaneously. In addition, the order of each operation can be rearranged. When its operation is completed, the processing can be terminated, but it can also have additional steps not included in the accompanying drawings. The processing can correspond to methods, functions, procedures, subroutines, subprograms, etc.

[0050] Embodiment 1

[0051] The present invention provides an AI-based optimization method for clinical research protocols, aiming to improve patient recruitment efficiency and reduce dropout rates. Figure 1As shown, the optimization method comprises the following steps:

[0052] S1: Collect current clinical trial data, and extract basic features and derived features associated with patient recruitment efficiency and dropout rate from the current clinical trial data.

[0053] Specifically, step S1 includes the following steps:

[0054] S11: Integrate current data from multiple clinical trial databases.

[0055] In this embodiment, it is first necessary to integrate current data from multiple clinical trial databases. Since clinical trials usually involve a large amount of patient information, treatment details, and outcome data, these data are scattered and stored in different databases, such as hospital information systems, electronic health record systems, clinical trial registration platforms, etc. Therefore, integrating these data sources can not only provide more comprehensive and accurate data support, but also help discover the correlation and potential patterns between data from different sources, thereby laying a solid foundation for subsequent feature extraction and model training.

[0056] S12: Preprocessing the integrated current data, wherein the preprocessing includes data cleaning, missing value filling and outlier processing.

[0057] In this embodiment, data preprocessing mainly includes data cleaning, missing value filling, and outlier processing. Data cleaning involves operations such as removing duplicate records and correcting erroneous data to ensure data quality; missing value filling is to take corresponding filling measures for the vacant values ​​in the data set, such as using the average value, median, or a method based on model prediction; outlier processing is to identify and process data points that significantly deviate from the normal range to avoid their negative impact on subsequent analysis. Through this series of preprocessing steps, data quality can be effectively improved to ensure the accuracy of subsequent analysis and modeling.

[0058] S13: Use feature selection algorithm to identify basic features related to patient recruitment efficiency and dropout rate, where the basic features include patient age, gender, disease type and previous medical history.

[0059] In this embodiment, the basic features here generally include but are not limited to the patient's age, gender, disease type, previous medical history, etc. The feature selection process can be implemented through a variety of algorithms, including but not limited to recursive feature elimination (RFE), model-based selection methods, etc. Through feature selection, the key factors that best reflect the patient recruitment efficiency and dropout rate can be screened out from a large number of original features, which is of great significance for subsequent model training because it can help reduce the impact of noise features and improve the generalization ability and interpretability of the model.

[0060] S14: Based on the basic features, derived features are constructed by statistical analysis methods, wherein the derived features include disease severity scores and treatment response prediction indexes.

[0061] In this embodiment, derived features refer to new features generated by mathematical operations or logical deductions of basic features, including disease severity scores, treatment response prediction indexes, etc. These derived features can often more directly reflect certain complex relationships or patterns, which helps to improve the predictive performance of the model. For example, the disease severity score can quantify the severity of the patient's condition by comprehensively considering multiple indicators, while the treatment response prediction index may predict the effectiveness of a specific treatment plan for the patient based on historical data. By constructing such derived features, not only can the depth and breadth of the model be increased, but it can also provide clinicians with more intuitive and practical information to help them make better decisions.

[0062] S2: The extracted basic features and derived features are simultaneously input into a plurality of pre-trained different multivariate regression models, and each multivariate regression model outputs a preliminary optimization suggestion, wherein the plurality of different multivariate regression models include a multivariate linear regression base model, a logistic regression base model and a random forest base model.

[0063] Specifically, step S2 includes the following steps:

[0064] S21: Combining the basic features and derived features into a feature vector.

[0065] In this embodiment, in order to prepare the data for input into each multivariate regression model, the basic features and derived features extracted from the current clinical trial data need to be combined into a feature vector. This step is crucial because a correctly constructed feature vector can effectively represent the information of individual patients, so that the model can learn patterns related to patient recruitment efficiency and dropout rate.

[0066] Specifically, the construction of the feature vector involves arranging all selected basic features and derived features in a certain order and digitizing them for the input of the model. For example, if the selected basic features include the patient's age, gender, and disease type, and the derived features include the disease severity score and the treatment response prediction index, then these features will be organized into an ordered list or array, providing structured data input for subsequent model training, which helps to improve the efficiency of model training and the accuracy of prediction.

[0067] S22: Inputting the feature vector into a pre-trained multivariate linear regression base model, a logistic regression base model and a random forest base model simultaneously.

[0068] In this embodiment, the constructed feature vector is then simultaneously input into the pre-trained multivariate linear regression base model, logistic regression base model, and random forest base model. These three models each have different working principles and applicable scenarios.

[0069] The meta-linear regression model is suitable for predicting continuous target variables, such as patient recruitment efficiency; the logistic regression model is suitable for binary classification problems and can be used to predict whether patients will drop out; and the random forest model is a method based on ensemble learning that can improve the stability and accuracy of predictions by constructing multiple decision trees.

[0070] After inputting the feature vector, each model will process the input data according to its internal learning algorithm and parameter settings, and output preliminary optimization suggestions related to patient recruitment efficiency and dropout rate. By utilizing different types of models, the same problem can be analyzed from multiple angles, thereby obtaining more comprehensive and reliable preliminary optimization suggestions.

[0071] In step S22, the pre-training process of the multivariate linear regression base model, the logistic regression base model and the random forest base model includes:

[0072] Historical clinical trial data are collected, and a training data set and a validation data set are divided from the historical clinical trial data.

[0073] For the multivariate linear regression base model, the historical basic features and historical derived features in the training data set were used as input variables, and the patient recruitment efficiency and dropout rate were used as target variables. Linear regression analysis was performed, and the prediction error was minimized by adjusting the model parameters to obtain the trained multivariate linear regression base model.

[0074] In this embodiment, the pre-training of the multivariate linear regression base model begins with the collection of historical clinical trial data. These data usually include detailed records of a large number of patients' personal information, disease conditions, treatment responses, and final clinical trial results, such as patient recruitment efficiency and dropout rate. After the data collection is completed, these data need to be divided into a training data set and a validation data set. The training data set is used for model training, and the validation data set is used to evaluate the generalization ability of the model. During the training process, the model uses the historical basic features and derived features in the training data set as input variables, and the patient recruitment efficiency and dropout rate as target variables to perform linear regression analysis. By adjusting model parameters, such as weights and bias terms, to minimize the prediction error, a trained multivariate linear regression base model is finally obtained.

[0075] For the Logistic regression base model, the historical basic features and historical derived features in the training data set were used as input variables, and the patient recruitment efficiency and dropout rate were used as target variables and converted into binary form. Logistic regression analysis was performed, and the trained Logistic regression base model was obtained by maximizing the log-likelihood function.

[0076] In this embodiment, the Logistic regression base model is also trained based on historical clinical trial data, but its target variable is to convert the patient recruitment efficiency and dropout rate into a binary classification problem, that is, whether the patient will participate in the clinical trial or drop out midway. During the training process, the model uses the historical basic features and derived features in the training data set as input variables, and the target variable is processed by binarization. For example, a dropout rate of 0 indicates that the patient has not dropped out, and a dropout rate of 1 indicates that the patient has dropped out. The model estimates parameters by maximizing the log-likelihood function to find the best parameter combination so that the model can best fit the training data. Logistic regression can effectively handle binary classification problems and provide probability-based classification results, which is particularly useful for predicting whether patients will drop out of clinical trials.

[0077] For the random forest base model, the historical basic features and historical derived features in the training data set are used as input variables, multiple decision trees are constructed as base learners, and the optimal number and depth of trees are selected through the cross-validation method to obtain the trained random forest base model.

[0078] In this embodiment, the random forest base model is a method based on ensemble learning, which improves the stability and accuracy of the model by constructing multiple decision trees. In the pre-training process, a training data set is first extracted from historical clinical trial data, which contains the basic characteristics and derived characteristics of the patient. Then, the model constructs multiple decision trees by random sampling and feature selection, and each decision tree is trained based on part of the samples and part of the features. The optimal number and depth of trees are selected by cross-validation method to ensure that the model does not overfit the training data while maintaining good generalization ability. Finally, by integrating the prediction results of multiple decision trees, the random forest base model can provide more robust and accurate predictions. Random forests can not only handle complex nonlinear relationships, but also effectively reduce the deviation and variance of a single model, and improve the reliability and accuracy of predictions.

[0079] S23: Each model calculates preliminary optimization suggestions related to patient recruitment efficiency and dropout rate based on the input feature vector, using its own learning algorithm and parameter settings; wherein the multivariate linear regression base model provides predictions and optimization suggestions based on linear relationships, the logistic regression base model provides probability-based classification and optimization suggestions, and the random forest base model provides optimization suggestions through ensemble learning.

[0080] In this embodiment, the preliminary optimization suggestions calculated by the multivariate linear regression base model, the logistic regression base model and the random forest base model are further processed. The core of this step is to evaluate and integrate these preliminary suggestions to generate the final optimization suggestions.

[0081] Specifically, each model’s initial recommendations are evaluated based on the accuracy and confidence of their predictions and then combined through a meta-model, which may use weighted averaging, voting mechanisms, or other ensemble learning strategies to consider the recommendations of individual models.

[0082] In addition, the meta-model will also evaluate the consistency score, complementarity and reliability score of each preliminary optimization suggestion to ensure that the final optimization suggestion is both consistent and complementary while maintaining high reliability. The technical effect of this is that by integrating the outputs of multiple models, the limitations of a single model can be overcome and the quality of the final optimization suggestion can be improved, thereby better guiding the optimization of clinical research plans and achieving the goal of improving patient recruitment efficiency and reducing dropout rates.

[0083] S3: The preliminary optimization suggestions output by each multivariate regression model are input into the pre-trained meta-model, and the meta-model generates the final optimization suggestions by integrating multiple preliminary optimization suggestions.

[0084] Specifically, step S3 includes the following steps:

[0085] S31: Receive preliminary optimization suggestions from the multivariate linear regression base model, the logistic regression base model, and the random forest base model.

[0086] In this embodiment, preliminary optimization suggestions output by the multivariate linear regression base model, the logistic regression base model, and the random forest base model are received. These preliminary optimization suggestions are calculated based on the basic features and derived features extracted in the previous stage through their respective model algorithms. The multivariate linear regression base model provides predictions based on linear relationships, the logistic regression base model provides classification suggestions based on probability, and the random forest base model provides optimization suggestions through ensemble learning, ensuring that the preliminary optimization suggestions from different models can be effectively integrated together for subsequent comprehensive processing.

[0087] S32: Input the preliminary optimization suggestion as an input variable into a pre-trained meta-model.

[0088] In this embodiment, these preliminary optimization suggestions are used as new input variables and are input into the pre-trained meta-model. The training process of the meta-model is based on historical clinical research plans and their optimization results. These historical data include basic features, derived features, and patient recruitment efficiency and dropout rate data after implementation.

[0089] In order to train the meta-model, it is first necessary to generate multiple preliminary optimization suggestions from historical data. This step simulates the output of multiple models in actual operations. Then, these preliminary optimization suggestions and their corresponding historical clinical research plans and optimization results are used as input variables and target variables for learning using machine learning algorithms. The learning goal of the meta-model is to find the best parameter configuration by minimizing the prediction error or maximizing the log-likelihood function, so that when new preliminary optimization suggestions are input, more accurate and effective final optimization suggestions can be output.

[0090] In step S32, the training process of the pre-trained meta-model includes:

[0091] Multiple historical clinical research plans and their corresponding optimization results are collected as training data sets, where each clinical research plan includes basic features, derived features, and patient recruitment efficiency and dropout rate data obtained after implementation.

[0092] In this embodiment, multiple historical clinical research schemes and their corresponding optimization results are collected as training data sets. These data sets cover a large number of actual cases of clinical studies, and each case contains basic features, derived features, and patient recruitment efficiency and dropout rate data after implementation. Basic features include basic information such as the patient's age, gender, disease type, and past medical history, while derived features include disease severity scores and treatment response prediction indexes. The collection of these data not only needs to be carried out across multiple data sources, but also the accuracy and completeness of the data must be guaranteed, because high-quality data is a prerequisite for training efficient models.

[0093] For each clinical research plan, multiple historical preliminary optimization recommendations are generated based on the patient recruitment efficiency and dropout rate data after implementation, as well as the prediction results of the multivariate regression model.

[0094] In this embodiment, at this stage, the basic features and derived features of each historical case are analyzed using the trained multivariate linear regression base model, logistic regression base model, and random forest base model to generate a series of preliminary optimization suggestions. These suggestions can be specific measures on how to improve the efficiency of patient recruitment, or strategies on how to reduce the patient dropout rate. Each model may produce differentiated suggestions due to its different algorithmic characteristics, and these differentiated suggestions are exactly what the meta-model needs to comprehensively consider.

[0095] The generated multiple historical preliminary optimization suggestions and their corresponding historical clinical research plans and optimization results are used as input variables and target variables to build and train a meta-model, wherein the meta-model is learned using a machine learning algorithm, and a trained meta-model is obtained by minimizing the prediction error.

[0096] In this embodiment, the metamodel is learned using a machine learning algorithm, and a trained metamodel is finally obtained by minimizing the prediction error. The learning process of the metamodel is a process of continuous iteration and optimization, which attempts to find an optimal way to combine multiple preliminary optimization suggestions to form a final optimization suggestion. To achieve this, the metamodel uses a weighted average ensemble learning method to evaluate the effectiveness and applicability of each preliminary optimization suggestion. In this way, the metamodel can not only consider the problem from multiple angles, but also improve the reliability and practicality of the final suggestion.

[0097] S33: The meta-model uses its own learning algorithm and parameter settings to perform comprehensive processing on multiple preliminary optimization suggestions, wherein the comprehensive processing includes evaluating the consistency score, degree of complementarity and reliability score of each preliminary optimization suggestion.

[0098] In this embodiment, the meta-model begins to perform a comprehensive process on the input multiple preliminary optimization suggestions, which involves evaluating the consistency score, complementarity level and reliability score of each preliminary optimization suggestion.

[0099] The consistency score is used to measure the level of consistency between different model recommendations, the complementarity score examines the complementary characteristics of each recommendation, and the reliability score reflects the credibility of each recommendation. Through this comprehensive evaluation, the meta-model can identify which preliminary optimization recommendations deserve more attention and which may have greater uncertainty. This processing step is critical to ensuring the quality of the final optimization recommendations because it helps to exclude inconsistent or unreliable recommendations and retain the most valuable information.

[0100] S34: Based on the result of the comprehensive processing, the meta-model generates a final optimization suggestion that integrates the advantages of multiple preliminary optimization suggestions, and outputs the final optimization suggestion.

[0101] In this embodiment, the metamodel generates and outputs the final optimization suggestions based on the results of the above comprehensive processing. The final optimization suggestions are intended to combine the advantages of multiple preliminary optimization suggestions to provide specific guidance for the optimization of clinical research programs. The output final optimization suggestions not only include specific measures to improve the efficiency of patient recruitment, such as optimizing recruitment materials, broadening recruitment channels, etc., but also cover strategies to reduce patient dropout rates, such as improving treatment plans, measures to enhance patient compliance, and providing psychological and social support. The implementation of these optimization suggestions can significantly improve the efficiency and success rate of clinical research projects, while improving patient experience and satisfaction.

[0102] S4: Based on the final optimization recommendations, optimize the clinical research plan to improve patient recruitment efficiency and reduce dropout rates.

[0103] Specifically, step S4 includes the following steps:

[0104] S41: Analyze the final optimization recommendations output by the meta-model and identify measures to improve patient recruitment efficiency and dropout rate.

[0105] In this example, the final optimization recommendations output by the meta-model need to be parsed to identify specific measures to improve patient recruitment efficiency and reduce dropout rates. This step requires the team to have deep expertise and data analysis capabilities to accurately interpret the information provided by the model.

[0106] Specifically, team members need to carefully review the results of the model output and identify which measures are considered to be most effective, such as specific types of patient education activities, specific recruitment channels, or specific patient screening criteria. In addition, the feasibility of these measures needs to be evaluated, taking into account existing resources and conditions, to determine which measures can be implemented directly and which require further research or adjustment.

[0107] S42: Adjust the recruitment strategy of the clinical research program based on the identified measures, where the recruitment strategy includes optimizing recruitment materials, broadening recruitment channels, adjusting patient screening criteria, and strengthening patient education and support.

[0108] In this embodiment, step S42 focuses on adjusting the recruitment strategy of the clinical research program according to these measures. This involves adjustments in many aspects, such as optimizing recruitment materials to make them more attractive to the target patient group, broadening recruitment channels to cover a wider range of potential participants, adjusting patient screening criteria to better match the purpose of the study, and strengthening patient education and support to ensure that potential participants fully understand the importance of the study and the participation process. These adjustments need to be based not only on the specific content of the final optimization recommendations, but also in combination with the actual needs of the clinical research and the characteristics of the target population. By carefully designing and implementing these strategies, the efficiency of patient recruitment can be effectively improved to ensure that the study can complete the predetermined sample size on time.

[0109] S43: At the same time, according to the final optimization recommendations, the treatment and management processes in the clinical research plan should be adjusted to reduce the patient dropout rate. The adjustments to the treatment and management processes include improving the treatment plan, measures to enhance patient compliance, and providing psychological and social support.

[0110] In this embodiment, step S43 focuses on adjusting the treatment and management processes in the clinical research plan according to the final optimization suggestions to reduce the patient dropout rate.

[0111] This step is crucial to ensure the integrity and validity of the study. Adjustments include, but are not limited to, improving the treatment plan to make it more suitable for the needs of the target patient group; implementing measures to enhance patient compliance, such as regular follow-up and providing convenient follow-up methods; and providing psychological and social support to help patients overcome the difficulties and challenges encountered during treatment. Through these measures, patients' compliance with the treatment plan can be improved, and dropouts due to various reasons can be reduced, thereby maintaining the sample size and data quality of the study.

[0112] S44: Implement adjusted clinical research plans.

[0113] In this embodiment, all the adjusted clinical research schemes described above are put into practice. The implementation phase requires close monitoring of research progress, timely resolution of problems that arise, and flexible adjustment of strategies based on actual conditions. During implementation, the team should maintain good communication with participants to ensure that they understand the goals and importance of the research, while also feeling the support and care of the research team. Through this series of optimization measures, the present invention aims to fundamentally improve the efficiency and quality of clinical research, promote the development of medical science, and bring better treatment options to patients.

[0114] Embodiment 2

[0115] The present invention also provides an AI-based clinical research program optimization system for executing the AI-based clinical research program optimization method, referring to Figure 2 As shown, the optimization system comprises:

[0116] The feature extraction module 100 is used to collect current clinical trial data and extract basic features and derived features associated with patient recruitment efficiency and dropout rate from the current clinical trial data.

[0117] The preliminary optimization suggestion generation module 200 is used to simultaneously input the extracted basic features and derived features into a plurality of pre-trained different multivariate regression models, and each multivariate regression model outputs preliminary optimization suggestions, wherein the plurality of different multivariate regression models include a multivariate linear regression base model, a logistic regression base model and a random forest base model.

[0118] The final optimization suggestion generating module 300 is used to input the preliminary optimization suggestions output by each multivariate regression model into the pre-trained meta-model, and the meta-model generates the final optimization suggestion by integrating multiple preliminary optimization suggestions.

[0119] The protocol execution module 400 is used to optimize the clinical research protocol based on the final optimization suggestion, so as to improve the patient recruitment efficiency and reduce the dropout rate.

[0120] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0121] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable programmable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0122] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. An AI-based optimization method for clinical research protocols, aiming to improve patient recruitment efficiency and reduce dropout rates, characterized in that: The optimization method comprises the following steps: Collecting current clinical trial data, and extracting basic features and derived features associated with patient recruitment efficiency and dropout rate from the current clinical trial data; The extracted basic features and derived features are simultaneously input into a plurality of pre-trained different multivariate regression models, and each multivariate regression model outputs a preliminary optimization suggestion, wherein the plurality of different multivariate regression models include a multivariate linear regression base model, a logistic regression base model, and a random forest base model; The preliminary optimization suggestions output by each multivariate regression model are input into the pre-trained meta-model, and the meta-model generates the final optimization suggestions by integrating multiple preliminary optimization suggestions; Based on the final optimization suggestions, the clinical research plan is optimized to improve patient recruitment efficiency and reduce dropout rate.

2. The method for optimizing a clinical research plan based on AI according to claim 1, characterized in that: The collecting of current clinical trial data and extracting basic features and derived features associated with patient recruitment efficiency and dropout rate from the current clinical trial data include: Integrate current data from multiple clinical trial databases; Preprocessing the integrated current data, wherein the preprocessing includes data cleaning, missing value filling and outlier processing; Using feature selection algorithms to identify basic features associated with patient recruitment efficiency and dropout rate, where the basic features include patient age, gender, disease type, and past medical history; Based on the basic features, derived features are constructed by statistical analysis methods, wherein the derived features include disease severity scores and treatment response prediction indexes.

3. The method for optimizing a clinical research plan based on AI according to claim 2, characterized in that: The extracted basic features and derived features are simultaneously input into a plurality of different pre-trained multivariate regression models, and each multivariate regression model outputs preliminary optimization suggestions, including: Combining the basic features and the derived features into a feature vector; The feature vector is simultaneously input into a pre-trained multivariate linear regression base model, a logistic regression base model, and a random forest base model; Based on the input feature vector, each model uses its own learning algorithm and parameter settings to calculate preliminary optimization suggestions related to patient recruitment efficiency and dropout rate. Among them, the multivariate linear regression base model provides predictions and optimization suggestions based on linear relationships, the logistic regression base model provides probability-based classification and optimization suggestions, and the random forest base model provides optimization suggestions through integrated learning.

4. A method for optimizing a clinical research program based on AI according to any one of claims 1 to 3, characterized in that: The pre-training process of the multivariate linear regression base model, the logistic regression base model and the random forest base model includes: Collecting historical clinical trial data, and dividing the historical clinical trial data into a training data set and a validation data set; For the multivariate linear regression base model, the historical basic features and historical derived features in the training data set were used as input variables, and the patient recruitment efficiency and dropout rate were used as target variables. Linear regression analysis was performed, and the trained multivariate linear regression base model was obtained by adjusting the model parameters to minimize the prediction error; For the Logistic regression base model, the historical basic features and historical derived features in the training data set were used as input variables, and the patient recruitment efficiency and dropout rate were used as target variables and converted into binary forms. Logistic regression analysis was performed, and the trained Logistic regression base model was obtained by maximizing the log-likelihood function. For the random forest base model, the historical basic features and historical derived features in the training data set are used as input variables, multiple decision trees are constructed as base learners, and the optimal number and depth of trees are selected through the cross-validation method to obtain the trained random forest base model.

5. The method for optimizing a clinical research program based on AI according to claim 4, characterized in that: The preliminary optimization suggestions output by each multivariate regression model are input into the pre-trained meta-model, and the meta-model generates a final optimization suggestion by integrating multiple preliminary optimization suggestions, including: Receive preliminary optimization suggestions from the multivariate linear regression base model, the logistic regression base model, and the random forest base model; Inputting the preliminary optimization suggestion as an input variable into a pre-trained meta-model; The meta-model uses its own learning algorithm and parameter settings to perform comprehensive processing on multiple preliminary optimization suggestions, wherein the comprehensive processing includes evaluating the consistency score, complementarity degree and reliability score of each preliminary optimization suggestion; Based on the results of the comprehensive processing, the meta-model generates a final optimization suggestion that combines the advantages of multiple preliminary optimization suggestions, and outputs the final optimization suggestion.

6. The method for optimizing a clinical research plan based on AI according to claim 5, characterized in that: The training process of the pre-trained meta-model includes: Collect multiple historical clinical research plans and their corresponding optimization results as training data sets, where each clinical research plan includes basic features, derived features, and patient recruitment efficiency and dropout rate data obtained after implementation; For each clinical research plan, multiple historical preliminary optimization suggestions are generated based on the patient recruitment efficiency and dropout rate data after implementation, as well as the prediction results of the multivariate regression model; The generated multiple historical preliminary optimization suggestions and their corresponding historical clinical research plans and optimization results are used as input variables and target variables to build and train a meta-model, wherein the meta-model is learned using a machine learning algorithm, and a trained meta-model is obtained by minimizing the prediction error.

7. The method for optimizing a clinical research plan based on AI according to claim 6, characterized in that: Based on the final optimization suggestions, the clinical research plan is optimized to improve the efficiency of patient recruitment and reduce the dropout rate, including: Analyze the final optimization recommendations output by the meta-model and identify measures to improve patient recruitment efficiency and dropout rate; Adjust the recruitment strategy of the clinical research program based on the identified measures, where the recruitment strategy includes optimizing recruitment materials, broadening recruitment channels, adjusting patient screening criteria, and strengthening patient education and support; At the same time, according to the final optimization recommendations, the treatment and management processes in the clinical research plan were adjusted to reduce the patient dropout rate. The adjustments to the treatment and management processes included improving the treatment plan, measures to enhance patient compliance, and providing psychological and social support. Implement the adjusted clinical research plan.

8. An AI-based clinical research program optimization system, used to execute an AI-based clinical research program optimization method according to any one of claims 1 to 7, characterized in that: The optimization system comprises: A feature extraction module, used to collect current clinical trial data and extract basic features and derived features associated with patient recruitment efficiency and dropout rate from the current clinical trial data; A preliminary optimization suggestion generating module, used for simultaneously inputting the extracted basic features and derived features into a plurality of pre-trained different multivariate regression models, and outputting preliminary optimization suggestions from each multivariate regression model, wherein the plurality of different multivariate regression models include a multivariate linear regression base model, a logistic regression base model and a random forest base model; The final optimization suggestion generation module is used to input the preliminary optimization suggestions output by each multivariate regression model into the pre-trained meta-model, and the meta-model generates the final optimization suggestion by integrating multiple preliminary optimization suggestions; The program execution module is used to optimize the clinical research program based on the final optimization suggestion, so as to improve the patient recruitment efficiency and reduce the dropout rate.

Images