Clinical test CRF automatic filling system and method

Abstract

The invention discloses a clinical test CRF automatic filling system and method, and the system comprises a data preprocessing end, a feature extraction end, a filling prediction end, a model training and optimization end, and a CRF filling evaluation end. The data preprocessing end, the feature extraction end, the filling prediction end, the model training and optimizing end and the CRF filling evaluation end are in data communication. According to the method and the device, the data in the clinical research form (CRF) can be conveniently and automatically filled, so that the efficiency and the accuracy of data entry are improved, the workload of data entry can be reduced, and the consistency and the accuracy of the data are improved.

Description

Technical Field

[0001] This invention relates to the field of drug clinical trials, and more specifically to an automated CRF filling system and method for clinical trials. Background Technology

[0002] In clinical trials, the CRF (Case Report Form) is an important tool for recording clinical trial data. It's used to collect and organize detailed information about the study subjects, including the role of the CRF, its content, and important considerations.

[0003] 1. The Role of CRF: CRF is crucial for data collection and management in clinical trials, ensuring the accuracy, completeness, and consistency of research data. Through CRF, researchers can meticulously record basic information about research subjects, medical history, laboratory test results, treatment processes, and adverse events, facilitating the analysis and evaluation of research outcomes.

[0004] 2. Contents of CRF: CRF usually includes the following parts: (1) Basic information: such as the subject's ID, name, gender, age, etc.; (2) Medical history: including past medical history, family medical history, allergy history, etc.; (3) Laboratory test results: such as blood routine, urine routine, biochemical tests, etc.; (4) Treatment process: record the treatment methods, dosage, course of treatment, etc. received by the subject; (5) Adverse events: record the adverse reactions or complications that occurred in the subject during the trial.

[0005] 3. Precautions for CRF: (1) When filling out CRF, be objective, truthful and accurate, and avoid subjective assumptions and false records; (2) CRF should be filled out in accordance with the pre-designed format and requirements to ensure data consistency; (3) During the experiment, CRF should be updated in a timely manner to ensure data integrity; (4) CRF should be kept safe and confidential to prevent data leakage or tampering.

[0006] In existing technologies, CRF data is typically entered manually during clinical trials, which results in very low efficiency and accuracy. Summary of the Invention

[0007] In order to solve the above problems, the present invention provides an automated CRF filling system and method for clinical trials.

[0008] To achieve the above objectives, the technical solution of the present invention is as follows:

[0009] An automated CRF filling system for clinical trials is provided, comprising a data preprocessing end, a feature extraction end, a filling prediction end, a model training and optimization end, and a CRF filling evaluation end, wherein the data preprocessing end, the feature extraction end, the filling prediction end, the model training and optimization end, and the CRF filling evaluation end communicate with each other.

[0010] In a preferred embodiment of the present invention, the data preprocessing end includes a missing value processing module, an outlier processing module, a data type conversion module, a data normalization module, and a data cleaning and correction module.

[0011] In a preferred embodiment of the present invention, the missing value processing module includes a missing value processing unit, which is used to process missing values ​​in the CRF that need to be entered.

[0012] The outlier processing module includes an outlier processing unit, which is used to detect and process outliers in the CRF that need to be entered.

[0013] The data type conversion module includes a data type conversion unit, which is used to convert the data in the CRF that needs to be entered into an appropriate data type.

[0014] The data normalization module includes a data normalization unit, which is used to normalize the data in the CRF to eliminate the dimensional differences between different data items.

[0015] The data cleaning and correction module includes a data cleaning and correction unit, which is used to detect and correct data errors, inconsistencies, or logical errors in the CRF.

[0016] In a preferred embodiment of the present invention, the feature extraction end includes a text feature extraction module, a numerical feature extraction module, a time series feature extraction module, an image feature extraction module, and a domain knowledge-based feature extraction module.

[0017] In a preferred embodiment of the present invention, the text feature extraction module includes a text feature extraction unit, which is used to extract key information and features from text data in a CRF;

[0018] The numerical feature extraction module includes a numerical feature extraction unit, which is used to extract important numerical features from the numerical data in the CRF.

[0019] The time series feature extraction module includes a time series feature extraction unit, which is used to extract time-related features from the time series data in the CRF.

[0020] The image feature extraction module includes an image feature extraction unit, which is used to extract image features using image processing and computer vision techniques if the CRF contains image data.

[0021] The domain knowledge-based feature extraction module includes a domain knowledge-based feature extraction unit, which is used to extract specific domain features related to clinical trials by utilizing the knowledge and experience of domain experts.

[0022] In a preferred embodiment of the present invention, the filling prediction end includes a model-based filling module, a similarity-based filling module, a rule-based filling module, and a collaborative filtering-based filling module.

[0023] In a preferred embodiment of the present invention, the model-based filling module includes a model-based filling unit, which is used to construct a predictive model based on existing CRF data and other relevant information, and fill in the missing data through the predictive ability of the model.

[0024] The similarity-based fill module includes a similarity-based fill unit, which is used to fill the values ​​between similar data samples based on the similarity of existing data samples in the CRF.

[0025] The rule-based fill module includes a rule-based fill unit, which is used to fill predictions based on predefined rules and logic, according to the relationship between existing data and rules in the CRF;

[0026] The collaborative filtering-based fill module includes a collaborative filtering-based fill unit, which is used to predict the value of missing data by analyzing existing data in the CRF and the subject's behavioral patterns using collaborative filtering technology.

[0027] In a preferred embodiment of the present invention, the model training and optimization end includes a feature engineering module, a model selection module, a model training module, a model optimization module, and a model optimization module.

[0028] In a preferred embodiment of the present invention, the feature engineering module includes a feature engineering unit, which is used to preprocess and transform CRF data to extract useful features for model training;

[0029] The model selection module includes a model selection unit, which is used to select a suitable filler prediction model for the task.

[0030] The model training module includes a model training unit, which is used to train and populate a prediction model based on existing CRF data and labels.

[0031] The model optimization module includes a model optimization unit, which is used to further optimize the performance of the infill prediction model.

[0032] The model optimization module includes a model optimization unit, which is used to evaluate the performance of the inflated prediction model.

[0033] The CRF filling evaluation module includes an accuracy evaluation module, an integrity evaluation module, a consistency evaluation module, an error analysis module, and a model evaluation module.

[0034] The accuracy evaluation module includes an accuracy evaluation unit, which is used to evaluate the consistency between the filled result and the true value.

[0035] The integrity assessment module includes an integrity assessment unit, which is used to assess the data integrity of the filling system's filling results.

[0036] The consistency assessment module includes a consistency assessment unit, which is used to assess the consistency indicators between the filling results and other existing data.

[0037] The error analysis module includes an error analysis unit, which is used to analyze and identify errors in the filling results;

[0038] The model evaluation module includes a model evaluation unit, which is used to evaluate the performance and prediction accuracy of the model if machine learning or statistical models are used for filling predictions.

[0039] An automatic CRF filling method for clinical trials, the automatic CRF filling method for clinical trials includes:

[0040] (1) Improve the quality and consistency of CRF data, reduce bias and error in the data analysis process, and ensure the reliability and availability of data;

[0041] (2) Extract key, informative features from the CRF for subsequent data filling, model building and data analysis;

[0042] (3) Use machine learning or other predictive models to automatically predict and fill missing data or blank fields in CRF;

[0043] (4) Improve the accuracy and generalization ability of the model to better predict and fill missing data in the CRF;

[0044] (5) Evaluate the filling results.

[0045] The beneficial effects of this invention are:

[0046] This invention facilitates the automatic filling of data in Clinical Research Forms (CRFs), thereby improving the efficiency and accuracy of data entry, reducing the workload of data entry, and improving the consistency and accuracy of data. Attached Figure Description

[0047] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0048] Figure 1 This is a schematic diagram of the data preprocessing end.

[0049] Figure 2 This is a schematic diagram of the feature extraction end.

[0050] Figure 3 A schematic diagram of the structure for filling the prediction end;

[0051] Figure 4 This is a schematic diagram of the structure of the model training and optimization ends;

[0052] Figure 5 A schematic diagram of the structure for filling the evaluation end of the CRF. Detailed Implementation

[0053] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention is further described below.

[0054] See Figure 1-5 The clinical trial CRF automatic filling system provided by this invention includes a data preprocessing end 1, a feature extraction end 2, a filling prediction end 3, a model training and optimization end 4, and a CRF filling evaluation end 5. The data preprocessing end 1, feature extraction end 2, filling prediction end 3, model training and optimization end 4, and CRF filling evaluation end 5 cooperate through data communication to process the CRF data to be entered into the clinical trial, which facilitates subsequent automatic filling, thereby improving the efficiency and accuracy of data entry, reducing the workload of data entry, and improving the consistency and accuracy of data.

[0055] The data preprocessing end 1, feature extraction end 2, imputation prediction end 3, model training and optimization end 4, and CRF imputation evaluation end 5 can communicate with each other via a dedicated network, which improves the security and efficiency of data communication.

[0056] See Figure 1 Data preprocessing unit 1 is used to improve the quality and consistency of the CRF data to be entered, reduce bias and errors in the data analysis process, and ensure the reliability and usability of the data. By adopting the data preprocessing step, it is possible to better address issues such as missing values, outliers, and data inconsistencies, providing a more reliable data foundation for subsequent imputation predictions and data analysis.

[0057] The data preprocessing module 1 includes a missing value handling module 101, an outlier handling module 102, a data type conversion module 103, a data normalization module 104, and a data cleaning and correction module 105.

[0058] The missing value processing module 101 includes a missing value processing unit 101a and a missing value processing database unit 101b.

[0059] The missing value processing unit 101a is used to process missing values ​​in the CRF data to be entered into the clinical trial system, i.e., those data items that have not been filled or recorded. Common missing value processing methods include deleting missing data, filling missing data with specific values ​​(such as mean, median, or mode), and using interpolation methods (such as linear interpolation, polynomial interpolation) for estimation.

[0060] Missing value processing database unit 101b is used to store missing value processing information.

[0061] The outlier handling module 102 includes an outlier handling unit 102a and an outlier handling database unit 102b.

[0062] The outlier handling unit 102a is used to detect and handle outliers in the CRF, namely those data that are significantly different from other data items or exceed the normal range. Outlier handling methods may include deleting outliers, replacing them with specific values, or correcting them using statistical methods.

[0063] The outlier handling database unit 102b is used to store outlier handling information.

[0064] The data type conversion module 103 includes a data type conversion unit 103a and a data type conversion database unit 103b.

[0065] The data type conversion unit 103a is used to convert data in the CRF into appropriate data types. For example, it converts string data into numeric, date / time, or boolean types for subsequent analysis and calculation.

[0066] Data type conversion database unit 103b is used to store data type conversion information.

[0067] The data normalization module 104 includes a data normalization unit 104a and a data normalization database unit 104b.

[0068] The data normalization unit 104a is used to normalize the data in the CRF to eliminate dimensional differences between different data items. Common normalization methods include min-max scaling and z-score normalization to ensure that the data are compared and analyzed on the same scale.

[0069] The data normalization database unit 104b is used to store data normalization information.

[0070] The data cleaning and correction module 105 includes a data cleaning and correction unit 105a and a data cleaning and correction database unit 105b.

[0071] The data cleaning and correction unit 105a is used to detect and correct data errors, inconsistencies, or logical errors in the CRF. For example, it checks the correlation, logical consistency, and reasonableness between data items and makes necessary corrections.

[0072] The data cleaning and correction database unit 105b is used to store data cleaning and correction information.

[0073] See Figure 2 Feature extraction end 2 is used to extract key, informative features from the CRF for subsequent data imputation, model building, and data analysis. Through feature extraction, the raw CRF data can be transformed into a feature representation that can be used for machine learning and data analysis, thereby improving the data's representational ability and predictive performance.

[0074] The feature extraction module 2 includes a text feature extraction module 201, a numerical feature extraction module 202, a time series feature extraction module 203, an image feature extraction module 204, and a domain knowledge-based feature extraction module 205.

[0075] The text feature extraction module 201 includes a text feature extraction unit 201a and a text feature extraction database unit 201b.

[0076] The text feature extraction unit 201a is used to extract key information and features from the text data in a CRF. Common text feature extraction methods can be employed, including the Bag-of-Words model, TF-IDF (Term Frequency-Inverse Document Frequency) weighting, and N-gram models. These methods can convert text data into numerical features for subsequent machine learning and data analysis.

[0077] Text feature extraction database unit 201b is used to store text feature extraction information.

[0078] The numerical feature extraction module 202 includes a numerical feature extraction unit 202a and a numerical feature extraction database unit 202b.

[0079] The numerical feature extraction unit 202a is used to extract important numerical features from the numerical data in the CRF. For example, it calculates the statistical characteristics of the numerical data (such as mean, standard deviation, maximum value, minimum value, etc.), and calculates the differences, ratios, or percentages between data.

[0080] The numerical feature extraction database unit 202b is used to store numerical feature extraction information.

[0081] The time series feature extraction module 203 includes a time series feature extraction unit 203a and a time series feature extraction database unit 203b.

[0082] The time series feature extraction unit 203a is used to extract time-related features from time series data in a CRF. For example, it calculates time intervals, extracts dates, days of the week, seasons, and other time-related features.

[0083] The time series feature extraction database unit 203b is used to store time series feature extraction information.

[0084] The image feature extraction module 204 includes an image feature extraction unit 204a and an image feature extraction database unit 204b.

[0085] The image feature extraction unit 204a is used to extract image features using image processing and computer vision techniques if the CRF contains image data (such as pathological slides, imaging data, etc.). Common image feature extraction methods include color histograms, texture features, and shape features.

[0086] Image feature extraction database unit 204b is used to store image feature extraction information.

[0087] The domain knowledge-based feature extraction module 205 includes a domain knowledge-based feature extraction unit 205a and a domain knowledge-based feature extraction database unit 205b.

[0088] Domain-knowledge-based feature extraction unit 205a is used to extract specific domain features relevant to clinical trials by leveraging the knowledge and experience of domain experts. Examples include features extracted from clinical diagnostic criteria, medical guidelines, or literature.

[0089] Domain-based knowledge-based feature extraction database unit 205b is used to store domain-based knowledge-based feature extraction information.

[0090] See Figure 3 The imputation prediction end 3 is used to automatically predict and impute missing data or blank fields in a CRF using machine learning or other predictive models. These algorithms are existing technologies that, based on existing data and features, learn patterns and rules between data to infer and predict imputed values. The goal of these imputation prediction algorithms is to infer the values ​​of missing data based on existing CRF data and related features, making the CRF data more complete and accurate. Through automatic imputation prediction algorithms, clinical trial CRF automatic imputation systems can improve data completeness and quality and reduce the workload of data entry.

[0091] The prediction module 3 includes a model-based filling module 301, a similarity-based filling module 302, a rule-based filling module 303, and a collaborative filtering-based filling module 304.

[0092] The model-based filling module 301 includes a model-based filling unit 301a and a model-based filling database unit 301b.

[0093] The model-based imputation unit 301a is used to build a predictive model based on existing CRF data and other relevant information, and then use the model's predictive power to fill in missing data. Common models include linear regression models, decision tree models, random forest models, and support vector machine models.

[0094] The model-based fill database unit 301b is used to store model-based fill information.

[0095] The similarity-based fill module 302 includes a similarity-based fill unit 302a and a similarity-based fill database unit 302b.

[0096] The similarity-based fill unit 302a is used to fill in the values ​​between similar data samples based on the similarity of existing data samples in the CRF. The similarity can be calculated based on features such as Euclidean distance, cosine similarity, and correlation coefficient.

[0097] The similarity-based fill database unit 302b is used to store similarity-based fill information.

[0098] The rule-based fill module 303 includes a rule-based fill unit 303a and a rule-based fill database unit 303b.

[0099] The rule-based imputation unit 303a is used to perform imputation predictions based on predefined rules and logic, and according to the relationships between existing data and rules in the CRF. For example, imputation rules are developed to predict the values ​​of missing data based on clinical guidelines or expert recommendations.

[0100] Rule-based fill database unit 303b is used to store rule-based fill information.

[0101] The collaborative filtering-based filling module 304 includes a collaborative filtering-based filling unit 304a and a collaborative filtering-based filling database unit 304b.

[0102] The collaborative filtering-based imputation unit 304a is used to predict the value of missing data by analyzing existing data in a CRF and the behavioral patterns of subjects using collaborative filtering techniques. Common collaborative filtering methods include user-based collaborative filtering and item-based collaborative filtering.

[0103] Collaborative filtering-based fill database unit 304b is used to store collaborative filtering-based fill information.

[0104] See Figure 4 The model training and optimization module 4 is used to train and optimize the imputation prediction model, improving its accuracy and generalization ability to better predict and impute missing data in a CRF. Through model training and optimization algorithms, an accurate and reliable imputation prediction model can be established, enabling automatic imputation of missing data in a CRF. This will improve data integrity and quality, and reduce the workload of data entry.

[0105] The model training and optimization module 4 includes a feature engineering module 401, a model selection module 402, a model training module 403, a model optimization module 404, and a model optimization module 405.

[0106] The feature engineering module 401 includes a feature engineering unit 401a and a feature engineering database unit 401b.

[0107] Feature engineering unit 401a is used to preprocess and transform CRF data to extract useful features for model training. These algorithms include feature selection, feature scaling, feature combination, etc., designed to extract features meaningful for imputation prediction from the raw data.

[0108] Feature engineering database unit 401b is used to store feature engineering information.

[0109] The model selection module 402 includes a model selection unit 402a and a model selection database unit 402b.

[0110] The model selection unit 402 is used to select a suitable filler prediction model for the task. Common algorithms can be used, such as cross-validation, grid search, and automated hyperparameter tuning, to select the best model by comparing the performance of different models.

[0111] Model selection database unit 402b is used to store model selection information.

[0112] The model training module 403 includes a model training unit 403a and a model training database unit 403b.

[0113] The model training unit 403 is used to train the populated prediction model based on existing CRF data and labels (populated data). Common algorithms can be used, such as linear regression, decision trees, random forests, support vector machines, and neural networks. These algorithms utilize the training data to learn the model's parameters and weights to minimize prediction error.

[0114] The model training database unit 403b is used to store model training information.

[0115] The model optimization module 404 includes a model optimization unit 404a and a model optimization database unit 404b.

[0116] The model optimization unit 404a is used to further optimize the performance of the imputed prediction model. Common algorithms can be employed, such as regularization, feature selection, cross-validation, and ensemble learning. These algorithms aim to reduce overfitting and improve the model's generalization ability.

[0117] Model optimization database unit 404b is used to store model optimization information.

[0118] The model optimization module 405 includes a model optimization unit 405a and a model optimization database unit 405b.

[0119] The model optimization unit 405a is used to evaluate the performance of the imputed prediction model. Specific evaluation metrics include common metrics such as accuracy, recall, F1 score, ROC curve, and AUC. These metrics help assess the model's predictive accuracy and stability.

[0120] The model optimization database unit 405b is used to store model optimization information.

[0121] See Figure 5The CRF infill evaluation end 5 is used to evaluate infill results and help determine performance and accuracy, as well as the quality of the infill results. Through these evaluations, the performance of the infill system can be understood and improvements can be made to ensure the reliability of the infill results and data integrity.

[0122] The CRF filling evaluation module 5 includes an accuracy evaluation module 501, an integrity evaluation module 502, a consistency evaluation module 503, an error analysis module 504, and a model evaluation module 505.

[0123] The accuracy evaluation module 501 includes an accuracy evaluation unit 501a and an accuracy evaluation database unit 501b.

[0124] The accuracy evaluation unit 501a is used to evaluate the consistency between the filled result and the true value. Specifically, it calculates the accuracy of the filled result by comparing the consistency with existing data or manual filling results. A high accuracy rate indicates that the filled result is relatively consistent with the true value.

[0125] Accuracy evaluation database unit 501b is used to store accuracy evaluation information.

[0126] The integrity assessment module 502 includes an integrity assessment unit 502a and an integrity assessment database unit 502b.

[0127] The integrity assessment unit 502a is used to evaluate the data integrity of the imputation results of this system. The integrity of the imputation results is calculated by comparing the number of missing values ​​in the imputation results with the number of missing data in the original system. A high integrity indicates that the imputation system can fill in more missing values.

[0128] Integrity assessment database unit 502b is used to store integrity assessment information.

[0129] The consistency assessment module 503 includes a consistency assessment unit 503a and a consistency assessment database unit 503b.

[0130] The consistency assessment unit 503a is used to evaluate the consistency between the populated results and other existing data. The consistency of the populated results is calculated by comparing them with other relevant data. A high consistency indicates that the populated results are relatively consistent with other relevant data.

[0131] The consistency assessment database unit 503b is used to store consistency assessment information.

[0132] The error analysis module 504 includes an error analysis unit 504a and an error analysis database unit 504b.

[0133] The error analysis unit 504a is used for the process of analyzing and identifying errors in the filling results. By detecting errors or inconsistencies in the filling results, it provides information about the type and location of the errors. This helps identify potential problems in the system and allows for improvements.

[0134] Error analysis database unit 504b is used to store error analysis information.

[0135] The model evaluation module 505 includes a model evaluation unit 505a and a model evaluation database unit 505b.

[0136] The model evaluation unit 505a is used to evaluate the model's performance and prediction accuracy if the system uses machine learning or statistical models for imputation prediction. Common evaluation metrics include accuracy, recall, F1 score, ROC curve, and AUC.

[0137] Model evaluation database unit 505b is used to store model evaluation information.

[0138] Based on the above-mentioned clinical trial CRF autofill system, the present invention also provides a clinical trial CRF autofill method, which includes:

[0139] (1) Improve the quality and consistency of CRF data, reduce bias and error in the data analysis process, and ensure the reliability and availability of data;

[0140] (2) Extract key, informative features from the CRF for subsequent data filling, model building and data analysis;

[0141] (3) Use machine learning or other predictive models to automatically predict and fill missing data or blank fields in CRF;

[0142] (4) Improve the accuracy and generalization ability of the model to better predict and fill missing data in the CRF;

[0143] (5) Evaluate the filling results.

[0144] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. An automated clinical trial CRF filling system, characterized in that, The clinical trial CRF automatic filling system includes a data preprocessing end, a feature extraction end, a filling prediction end, a model training and optimization end, and a CRF filling evaluation end, and the data preprocessing end, feature extraction end, filling prediction end, model training and optimization end, and CRF filling evaluation end communicate with each other.

2. The automated clinical trial CRF filling system according to claim 1, characterized in that, The data preprocessing module includes a missing value handling module, an outlier handling module, a data type conversion module, a data normalization module, and a data cleaning and correction module.

3. The automated clinical trial CRF filling system according to claim 2, characterized in that, The missing value processing module includes a missing value processing unit, which is used to process missing values ​​in the CRF that need to be entered. The outlier processing module includes an outlier processing unit, which is used to detect and process outliers in the CRF that need to be entered. The data type conversion module includes a data type conversion unit, which is used to convert the data in the CRF that needs to be entered into an appropriate data type. The data normalization module includes a data normalization unit, which is used to normalize the data in the CRF to eliminate the dimensional differences between different data items. The data cleaning and correction module includes a data cleaning and correction unit, which is used to detect and correct data errors, inconsistencies, or logical errors in the CRF.

4. The automated clinical trial CRF filling system according to claim 1, characterized in that, The feature extraction module includes a text feature extraction module, a numerical feature extraction module, a time series feature extraction module, an image feature extraction module, and a domain knowledge-based feature extraction module.

5. The automated clinical trial CRF filling system according to claim 4, characterized in that, The text feature extraction module includes a text feature extraction unit, which is used to extract key information and features from the text data in the CRF. The numerical feature extraction module includes a numerical feature extraction unit, which is used to extract important numerical features from the numerical data in the CRF. The time series feature extraction module includes a time series feature extraction unit, which is used to extract time-related features from the time series data in the CRF. The image feature extraction module includes an image feature extraction unit, which is used to extract image features using image processing and computer vision techniques if the CRF contains image data. The domain knowledge-based feature extraction module includes a domain knowledge-based feature extraction unit, which is used to extract specific domain features related to clinical trials by utilizing the knowledge and experience of domain experts.

6. The automated clinical trial CRF filling system according to claim 1, characterized in that, The filling prediction module includes a model-based filling module, a similarity-based filling module, a rule-based filling module, and a collaborative filtering-based filling module.

7. The automated clinical trial CRF filling system according to claim 6, characterized in that, The model-based fill module includes a model-based fill unit, which is used to construct a predictive model based on existing CRF data and other relevant information, and fill in the missing data through the predictive ability of the model. The similarity-based fill module includes a similarity-based fill unit, which is used to fill the values ​​between similar data samples based on the similarity of existing data samples in the CRF. The rule-based fill module includes a rule-based fill unit, which is used to fill predictions based on predefined rules and logic, according to the relationship between existing data and rules in the CRF; The collaborative filtering-based fill module includes a collaborative filtering-based fill unit, which is used to predict the value of missing data by analyzing existing data in the CRF and the subject's behavioral patterns using collaborative filtering technology.

8. The automated clinical trial CRF filling system according to claim 1, characterized in that, The model training and optimization module includes a feature engineering module, a model selection module, a model training module, a model optimization module, and a model optimization module.

9. The automated clinical trial CRF filling system according to claim 8, characterized in that, The feature engineering module includes a feature engineering unit, which is used to preprocess and transform CRF data to extract useful features for model training. The model selection module includes a model selection unit, which is used to select a suitable filler prediction model for the task. The model training module includes a model training unit, which is used to train and populate a prediction model based on existing CRF data and labels. The model optimization module includes a model optimization unit, which is used to further optimize the performance of the infill prediction model. The model optimization module includes a model optimization unit, which is used to evaluate the performance of the inflated prediction model. The CRF filling evaluation module includes an accuracy evaluation module, an integrity evaluation module, a consistency evaluation module, an error analysis module, and a model evaluation module. The accuracy evaluation module includes an accuracy evaluation unit, which is used to evaluate the consistency between the filled result and the true value. The integrity assessment module includes an integrity assessment unit, which is used to assess the data integrity of the filling system's filling results. The consistency assessment module includes a consistency assessment unit, which is used to assess the consistency indicators between the filling results and other existing data. The error analysis module includes an error analysis unit, which is used to analyze and identify errors in the filling results; The model evaluation module includes a model evaluation unit, which is used to evaluate the performance and prediction accuracy of the model if machine learning or statistical models are used for filling predictions.

10. A method for automatically filling in clinical trial CRFs, characterized in that, The automatic CRF filling method for the clinical trial includes: (1) Improve the quality and consistency of CRF data, reduce bias and error in the data analysis process, and ensure the reliability and availability of data; (2) Extract key, informative features from the CRF for subsequent data filling, model building and data analysis; (3) Use machine learning or other predictive models to automatically predict and fill missing data or blank fields in CRF; (4) Improve the accuracy and generalization ability of the model to better predict and fill missing data in CRF; (5) Evaluate the filling results.

Images