Integrated clinical trial support platform and support method

Abstract

The present invention relates to an integrated clinical trial support platform and support method, wherein the integrated clinical trial support platform comprises an integrated system comprising: a recruitment management system for performing a service of recruiting clinical trial participants according to a request of a clinical trial consumer; a data analysis system for receiving an input of clinical trial information from the clinical trial consumer or a clinical trial provider to analyze clinical trial data, and predicting a clinical trial result; and a document management system for managing a clinical trial document. Therefore, the present invention can improve the operational efficiency of a clinical trial by integrally supporting a process related to execution of the clinical trial.

Description

Clinical Trial Integrated Support Platform and Support Methods

[0001] The present invention relates to an integrated platform for supporting clinical trials, and more particularly, to an integrated platform for supporting clinical trials and a support method for supporting clinical trials that improve the operational efficiency of clinical trials by comprehensively supporting processes related to conducting clinical trials.

[0002] Numerous drugs currently in use require a process of proving their efficacy and safety, a process known as clinical trials. A clinical trial is a series of tests or studies conducted on humans prior to development and marketing. These tests aim to demonstrate the safety and therapeutic efficacy of a drug by examining its distribution, metabolism, excretion, pharmacological and clinical effects, and to investigate side effects.

[0003] In particular, clinical trials for new drug approval must be submitted to the licensing agency (e.g., the Ministry of Food and Drug Safety) prior to initiating the clinical trial, and must be conducted in accordance with strict scientific ethics regulations.

[0004] However, for clinical trials, it is necessary to recruit users who meet various conditions such as gender, age, and medical history. However, since the process of sounding out users' willingness to participate in clinical trials is conducted in extremely limited environments such as hospitals, it is not easy to recruit clinical trial participants.

[0005] Furthermore, these clinical trials are difficult to predict for patients, and they often take years to conduct and are extremely costly. Therefore, efforts are ongoing to reduce the time and cost of clinical trials, and the development of systems and methods for predicting clinical trial outcomes is required.

[0006] In addition, conventional paper-based clinical trial data management has inherent problems such as extremely weak data storage, maintenance, and security, as well as extremely limited data sharing, data reprocessing, variability or fluidity of the test or review period, and subsequent reference and utilization.

[0007] Therefore, the clinical trial market requires a technology for a clinical trial integrated support platform and support method that comprehensively supports the processes essential for conducting each clinical trial while efficiently carrying out activities such as recruitment and management of clinical trial participants, analysis and prediction of clinical trial data, and management of clinical trial documents.

[0008] The purpose of the present invention is to provide an integrated clinical trial support platform and support method that improve the operational efficiency of clinical trials by comprehensively supporting processes related to conducting clinical trials.

[0009] In addition, another purpose is to provide an integrated clinical trial support platform and support method that allows clinical trials to be conducted based on individual user consent by providing a transcription signature procedure for consent in bulk through the user terminal when recruiting participants for clinical trial participation.

[0010] In addition, another objective is to provide an integrated clinical trial support platform and support method that predicts the success rate of clinical trials for new treatments using machine learning algorithms.

[0011] In addition, another purpose is to provide a clinical trial integrated support platform and support method that provides standardized clinical trial documents and allows for integrated management of each clinical trial-related document.

[0012] In order to achieve the above-described purpose, the integrated clinical trial support platform of the present invention includes an integrated system (10) that provides clinical trial support services through an integrated network (40) to a clinical trial demander (20) who requires a clinical trial and a clinical trial supplier (30) who performs a clinical trial, and the integrated system (10) is characterized by including a recruitment management system (100) that performs a service of recruiting clinical trial participants at the request of the clinical trial demander (20) or the clinical trial supplier (30); a data analysis system (200) that receives clinical trial-related information from the clinical trial supplier (30), analyzes clinical trial-related data, and predicts clinical trial results; and a document management system (300) that manages the clinical trial-related documents.

[0013]

[0014] In addition, the recruitment management system (100) includes a center server (130) that transmits and receives signals and data with a user terminal (110) through a recruitment network (120) and performs a clinical trial information provision and health confirmation process through the user terminal; a contracting agency server (140) that performs tasks such as preparing a clinical trial plan, transmitting the clinical trial plan, and generating review results; a request request server (150) that transmits a clinical trial request to the contracting agency server (140) through the recruitment network (120) for a clinical trial subject; an authentication server (160) that receives selected clinical trial participant information from the center server (130) through the recruitment network (120) and then transmits a consent form to the user terminal (110) corresponding to the clinical trial participant information according to a clinical trial consent request recommended to the user terminal (110) included in the clinical trial participant information; And it is characterized by including a smart server (170) that receives user image information from a user terminal (110) through a recruitment network (120) and provides clinical trial participation information to the user terminal (110) using the received user image.

[0015]

[0016] In addition, the data analysis system (200) is characterized by including an analysis device (220) that receives clinical trial-related information and predicts clinical trial results using a prediction model for which machine learning has been completed; a user terminal (110) that is connected to the analysis device and an analysis network (230) to receive a prediction result of the prediction device; and an analysis server (210) that provides data for prediction processing of the analysis device (220) and performs a login procedure at the user's request.

[0017]

[0018] In addition, the document management system (300) is characterized by including a management server (310); a database (320) in which clinical trial documents managed by the management server (310) are stored; and a management device (330) for calculating the similarity of clinical trial documents managed by the management server (310).

[0019]

[0020] In addition, a clinical trial request request step (S100) in which a requesting agency server (150) requests a clinical trial request to a contracting agency server (140) through a recruitment network (120); a clinical trial plan provision step (S110) in which, after the clinical trial request request step, the contracting agency server (140) provides a clinical trial plan to a center server (130) through a recruitment network (120); a review result generation step (S120) in which, after the clinical trial plan provision step, the center server (130) generates an IRB review result for the clinical trial plan received in the clinical trial plan provision step; a clinical trial participant recruitment request step (S130) in which, after the IRB review result generation step, the contracting agency server (140) transmits a request for clinical trial participant recruitment to the center server (130) through the recruitment network (120); a clinical trial participant selection step (S140) in which, after the clinical trial participant recruitment request step, the center server (130) selects a clinical trial participant; It is characterized by including a clinical trial participant information generation step (S160) for generating information on a clinical trial participant selected in the clinical trial participant selection step; and a clinical trial participant information transmission step (S160) for transmitting the clinical trial participant information generated in the clinical trial participant generation step to a center server (300).

[0021]

[0022] In addition, it is characterized by including a prediction model generation step (S200) for generating a prediction model that predicts clinical trial results by performing machine learning using clinical trial case data; a related information reception step (S210) for receiving clinical trial related information from a user terminal (110) after the prediction model generation step; and a prediction step (S220) for predicting clinical trial results according to clinical trial related information using the prediction model after the related information reception step.

[0023]

[0024] In addition, it is characterized by including a document template selection request reception step (S300) in which the clinical trial document management server (100) receives a document template selection request from a first user; an information reception step (S310) in which, after the document template selection request reception step, the management server (310) provides the selected or newly created document template to the first user and receives input of information that needs to be written on the document template from the first user; a clinical trial document sharing step (S320) in which, after the information reception step, the management server (310) shares the first written clinical trial document with the document management system (300); and, after the clinical trial document sharing step, when the management server (310) receives a request for submission of the first final clinical trial document to the review agency from the first user on the document management system (300), a clinical trial document submission step (S330) in which the first final clinical trial document is submitted to the review agency.

[0025] According to the present invention, there is an effect of improving the operational efficiency of clinical trials by comprehensively supporting processes related to conducting clinical trials.

[0026] In addition, when recruiting participants for clinical trials, the consent signature procedure is provided in bulk through the user terminal, so that the clinical trial is conducted based on the individual consent of the user, which has the effect of making the identity authentication and verification process transparent and efficient.

[0027] Additionally, it has the effect of providing the probability of success or failure of clinical trials with greater reliability.

[0028] Additionally, it has the effect of providing standardized clinical trial documents and enabling integrated management of each clinical trial-related document.

[0029] Figure 1 is a diagram showing the configuration of the clinical trial integrated support platform of the present invention.

[0030] Figure 2 is a drawing showing the configuration of the integrated system of the present invention.

[0031] Figure 3 is a drawing showing the configuration of the recruitment management system of the present invention.

[0032] Figure 4 is a flowchart of a recruitment management method using the recruitment management system of the present invention.

[0033] Figure 5 is a flowchart showing the detailed sequence of the participation information display step in the recruitment management method of the present invention.

[0034] Figure 6 is a diagram showing the configuration of the data analysis system of the present invention.

[0035] Figure 7 is a drawing showing the configuration of the analysis device of the present invention.

[0036] Figure 8 is a flowchart showing an analysis method according to one embodiment of the present invention.

[0037] Figure 9 is a flowchart of the prediction model generation method of the present invention.

[0038] Figure 10 is a flowchart of a method for generating a prediction model according to another embodiment of the present invention.

[0039] Figure 11 is a flowchart of a method for predicting clinical trial results according to another embodiment of the present invention.

[0040] Figure 12 is a diagram showing the configuration of the document management system of the present invention.

[0041] Figure 13 is a drawing showing the configuration of the management device of the present invention.

[0042] Figure 14 is a flowchart showing the document management method of the present invention.

[0043] Figure 15 is a flowchart showing a method for calculating similarity and suggesting documents using the management device of the present invention.

[0044] The present invention will now be described in detail. In describing specific embodiments below, various specific details have been included to further illustrate the invention and aid understanding. However, those skilled in the art will recognize that the present invention can be utilized without these specific details.

[0045]

[0046] Figure 1 is a diagram showing the configuration of the clinical trial integrated support platform of the present invention.

[0047] Referring to FIG. 1, the clinical trial integrated support platform (1) of the present invention may include a clinical trial demander (20) requiring clinical trial performance work, a clinical trial supplier (30) supplying clinical trial performance work, and an integrated system (10) that can be accessed by a clinical trial demander (20) or a clinical trial supplier (30) through an integrated network (40).

[0048] Clinical trial demanders (20) may include various companies requiring clinical trials, such as pharmaceutical companies, medical device manufacturers, functional food manufacturers, cosmetics manufacturers, therapeutic material manufacturers, and university hospitals.

[0049] Clinical trial suppliers (30) may include various institutions capable of performing clinical trial work, such as market research companies, big data companies, human resources agencies, contract agencies, and centers.

[0050] The integrated network (40) can be configured regardless of the communication mode, such as wired or wireless, and can be implemented in various forms to perform communication between servers and between servers and terminals. More specifically, the integrated network (40) is a communication network that is a high-speed backbone network of a large communication network capable of large-capacity, long-distance voice and data services, and may be a next-generation wired or wireless network for providing the Internet or high-speed multimedia services. If the network (200) is a mobile communication network, it may be a synchronous mobile communication network or an asynchronous mobile communication network.

[0051]

[0052] Figure 2 is a drawing showing the configuration of the integrated system of the present invention.

[0053] Referring to FIG. 2, the integrated system (10) of the present invention may include a recruitment management system (100) that performs a service of recruiting clinical trial participants at the request of a clinical trial demander (20), a data analysis system (200) that receives clinical trial information from a clinical trial demander (20) or a clinical trial supplier (30), analyzes clinical trial data, and predicts clinical trial results, and a document management system (300) that manages clinical trial documents.

[0054]

[0055] Figure 3 is a diagram showing the configuration of the recruitment management system of the present invention.

[0056] Referring to FIG. 3, the recruitment management system (100) of the present invention may include a user terminal (110), a recruitment network (120), a center server (130), a consignment institution server (140), a request request server (150), an authentication server (160), and a smart server (170).

[0057]

[0058] The user terminal (110) is a wired or wireless communication terminal, such as a smartphone, tablet, or PC, used by each user, including hospitals or institutions, and can provide various data held by the user to the center server (130). To this end, the user terminal (110) is a wireless terminal equipped with computing capabilities and can be operated by installing various apps.

[0059]

[0060] The recruitment network (120) may be configured with the same configuration as the integrated network (40) regardless of the communication mode, such as wired or wireless. In other words, the recruitment network (120) refers to a network configuration applied within the recruitment management system (100).

[0061]

[0062] The center server (130) can transmit and receive signals and data with the user terminal (110) via the recruitment network (120), and can provide information on clinical trials matching diseases of interest and perform health check processes through the user terminal (110). Thereafter, users of the user terminal (110) who have passed the health check can be registered as clinical trial participants in the database.

[0063]

[0064] The contracted institution server (140) can perform tasks such as drafting a clinical trial plan, transmitting the clinical trial plan, and generating review results. More specifically, the contracted institution server (140) can create a clinical trial plan for a clinical trial subject and transmit the created clinical trial plan to the center server (130) via the recruitment network (120). Furthermore, the center server (130) that receives the clinical trial plan can have its own Institutional Review Board (IRB) review the clinical trial plan and generate review results.

[0065] In addition, when the consignment institution server (140) receives a message regarding approval or disapproval of the clinical trial plan based on the review results, it can transmit a request for recruitment of clinical trial participants to the center server (130) through the recruitment network (120).

[0066]

[0067] The request server (150) can transmit a request for clinical trial request to the contract institution server (140) for clinical trial subjects through the recruitment network (120). The request server can include various companies requiring clinical trials, such as pharmaceutical companies, medical device manufacturers, functional food manufacturers, and cosmetics manufacturers.

[0068]

[0069] The authentication server (160) may receive information on selected clinical trial participants from the center server (130) through the recruitment network (120), and then transmit a consent form to a user terminal (110) corresponding to the clinical trial participant information according to a clinical trial consent request recommended to the user terminal (110) included in the clinical trial participant information.

[0070]

[0071] The smart server (170) can receive user image information from a user terminal (110) through a recruitment network (120) and provide clinical trial participation information to the user terminal (110) using the received user image.

[0072]

[0073] Figure 4 is a flowchart of a recruitment management method using the recruitment management system of the present invention.

[0074] Referring to FIG. 4, the recruitment management method of the present invention comprises a clinical trial request request step (S100) for requesting a request, a clinical trial plan provision step (S110) for providing a clinical trial plan to a center server (130) through a recruitment network (120) after the clinical trial request request step, a review result generation step (S120) for generating an IRB review result for the clinical trial plan received in the clinical trial plan provision step at the center server (130) after the clinical trial plan provision step, a clinical trial participant recruitment request step (S130) for transmitting a request for recruitment of clinical trial participants to the center server (130) through the recruitment network (120) after the IRB review result generation step, a clinical trial participant selection step (S140) for selecting clinical trial participants by the center server (130) after the clinical trial participant recruitment request step, a clinical trial participant information generation step (S150) for generating information on clinical trial participants selected in the clinical trial participant selection step, and transmitting the clinical trial participant information generated in the clinical trial participant generation step to the center. The method may include a step (S160) of transmitting clinical trial participant information to a server (130), a step (S170) of transmitting clinical trial participant information to a contracting institution server (140) via a recruitment network (120) by a requesting party server (150), a step (S170) of generating a consent form by a center server (130) of transmitting the IRB review result to a smart server via the recruitment network (120) and having the smart server generate a consent form based on the IRB review result, a step (S180) of verifying the consent form generated in the consent form generation step by the center server (130) via the recruitment network (120) after the consent form generation step; and an electronic signature processing step (S190) of performing an electronic signature process by a smart server (170) via a user terminal (110) and the recruitment network (120) if the consent form verification is passed in the consent form verification step.

[0075]

[0076] In the review result generation step (S120), the center server (130) performs a review of the clinical trial plan by its own IRB to generate a review result, and if the IRB review result is approval of the clinical trial plan, a clinical trial plan approval message can be transmitted to the contract institution server (140) through the recruitment network (120).

[0077] In the clinical trial participant information transmission step (S160), the center server (130) can transmit the IRB review results to the smart server (170) via the recruitment network (120) and also transmit the clinical trial participant information in the same manner.

[0078] In the consent form generation step (S170), a mobile electronic consent form can be generated to be transmitted to a user terminal (110) corresponding to clinical trial participant information based on the IRB review results by a smart server, or an existing mobile electronic consent form can be extracted from a database (not shown).

[0079] In the consent verification step (S180), the center server (130) can independently perform mobile electronic consent verification on the mobile electronic consent generated by the smart server through the recruitment network (120) or can perform verification by transmitting the generated mobile electronic consent to the consignment agency server (140) and requesting verification through the smart server (170).

[0080] In the electronic signature process step (S190), the smart server (170) can proceed with the mobile electronic consent form signing step through the user terminal (110) and the recruitment network (120).

[0081] In addition, according to one embodiment of the present invention, in the clinical trial participant information generation step (S150), a smart server (170) may generate clinical trial participation information and may include a participation information display step for displaying the generated information on a user terminal (110), which will be described in detail later with reference to FIG. 5.

[0082]

[0083] Figure 5 is a flowchart showing the detailed sequence of the participation information display step in the recruitment management method of the present invention.

[0084] Referring to FIG. 5, the participation information display step may include an image acquisition step (S151) of acquiring an image of a clinical trial participant, a first input signal generation step (S152) of encoding the image acquired in the image acquisition step to generate a first input signal, a neural network input step (S153) of inputting the first input signal generated in the first input signal generation step into a convolutional neural network, a network input step (S154) of inputting the output value of the convolutional neural network and the first output signal previously stored in the database into the neural network after the neural network input step, a second output signal acquisition step (S155) of acquiring a second output signal after the network input step, and a participation information display step (S156) of transmitting the acquired second output signal to a user terminal (110) so that clinical trial participation information is displayed on the user terminal (110).

[0085]

[0086] At this time, in the image acquisition step (S151), the smart server (170) can acquire a photo of the clinical trial subject using the camera of the user terminal (110). For example, when an image of the clinical trial subject is captured using the camera of the user terminal (110), the smart server (170) can acquire image information of the photo from the user terminal (110).

[0087]

[0088] In the first input signal generation step (S152), the smart server (170) can generate the first input signal by encoding the pixels of the photo into color information. The color information can include RGB color information, brightness information, and saturation information. The smart server (170) can convert the color information into numerical values ​​and encode the photo in the form of a data sheet including these values.

[0089]

[0090] In the neural network input stage (S153), the first input signal can be input into a pre-trained convolutional neural network of an embedded computer (not shown) connected to a smart server (100).

[0091]

[0092] In the network input step (S154), the smart server (170) may input the output value of the convolutional neural network and the first output signal previously stored in a database connected to the smart server (170) into the pre-trained neural network based on the result of the input of the convolutional neural network. According to one embodiment, the smart server (170) may use the first output signal previously stored in the database as an input to the neural network. The first output signal used as an input may be for analyzing the output value output through the operation of the convolutional neural network through comparison and utilizing the information accumulated in the first output signal.

[0093]

[0094] In the second output signal acquisition step (S155), the smart server (170) can acquire a second output signal based on the result of the input of the neural network. According to one embodiment, the second output signal can be generated by comparing and accumulating the output value, which is the result of the operation of the convolutional neural network, with the first output signal input through the neural network. The generated second output signal may include, but is not limited to, clinical trial participation information.

[0095]

[0096] In the participation information display step (S156), the smart server (170) can process the clinical trial participation information to be displayed on the user terminal (110) based on the second output signal. According to one embodiment, the clinical trial participation status may include information such as the number of clinical trial participants, participation frequency, health status of clinical trial participants, insufficient number of participants, whether participants are participating in the clinical trial overlapping, and clinical trial refresh period. The user terminal (110) can display the clinical trial participation status in numbers and graphs, and can provide participation status display in various forms such as text and tables.

[0097]

[0098] The convolutional neural network can take as input a first input signal generated by encoding an image captured by a camera of a user terminal (110), and output the type of clinical trial and whether there is duplicate participation in the clinical trial.

[0099] Encoding according to one embodiment may be performed by storing color information for each pixel of a photo in the form of a numerical data sheet, and the color information may include saturation information, brightness information, RGB color, etc. of each pixel.

[0100] According to one embodiment, the convolutional neural network is composed of a classification neural network and a feature extraction neural network, and the feature extraction neural network can perform the task of separating the face of a clinical trial subject from the background in a photograph of the face of the clinical trial subject. The classification neural network can perform the task of recognizing the face of the clinical trial subject to identify which user the user is, classifying the clinical trials in which the user has participated by type, and determining whether there is duplicate participation in the clinical trials by type of clinical trials. The method for the feature extraction neural network to distinguish the face of the clinical trial subject from the background may be, but is not limited to, a group of pixels in which changes in each value of color information from a data sheet of a first input signal encoding the photograph are detected to have occurred by 30% or more in 6 or more of 8 pixels including one pixel, as the boundary between the face of the clinical trial subject and the background.

[0101] According to one embodiment, a feature extraction neural network sequentially stacks convolutional layers and pooling layers on an input signal. The convolutional layer includes a convolution operation, a convolutional filter, and an activation function. The calculation of the convolutional filter is adjusted according to the matrix size of the target input, but a 9X9 matrix is ​​typically used. The activation function typically uses, but is not limited to, the ReLU function, the sigmoid function, and the tanh function. The pooling layer is a layer that reduces the matrix size of the input, and uses a method of extracting a representative value by grouping pixels in a specific area. The operation of the pooling layer typically uses, but is not limited to, the average or the maximum value. The operation is performed using a square matrix, typically a 9X9 matrix. The convolutional layer and the pooling layer are alternately repeated until the input is sufficiently small while maintaining the difference.

[0102] In one embodiment, a classification neural network recognizes the faces of clinical trial subjects, distinguished from the background by a feature extraction neural network, identifies the user through facial recognition, categorizes the clinical trials the user has participated in by type, and checks for duplicate clinical trial participation. Information stored in a database can be utilized to recognize the faces of clinical trial subjects. The classification neural network prioritizes identifying the type of clinical trial and can then determine whether there is duplicate clinical trial participation based on the identified clinical trial type.

[0103]

[0104] In one embodiment, a classification neural network has hidden layers and output layers. A convolutional neural network for a big data-based clinical trial information management method typically has five or more hidden layers, and each hidden layer typically has 80 nodes, though this number can be increased in some cases. The hidden layer's activation function can be, but is not limited to, the ReLU function, the sigmoid function, or the tanh function. The output layer of the convolutional neural network can have a total of 100 nodes.

[0105]

[0106] According to an embodiment, the output of the convolutional neural network may include 100 nodes in the output layer, wherein the top 50 nodes may indicate the type of the target clinical trial, and the bottom 50 nodes may indicate whether there is duplicate participation in the clinical trial corresponding to the top node. The method of corresponding the top 50 nodes and the bottom 50 nodes may be performed in a way that the top n-th node and the bottom n-th node correspond to the top n-th node, so that the n-th node in total corresponds to the 50+n-th node in total. For example, the 1st node may correspond to the 51st node, the 2nd node may correspond to the 52nd node, the 10th node may correspond to the 60th node, and the 50th node may correspond to the 100th node. The type of clinical trial may be output as code information corresponding to the clinical trial, but is not limited thereto. Among the 100 output layer nodes of the convolutional neural network (501), output layer nodes without output values ​​can output the number '0' as their output value. Nodes that include this number '0' among the top 50 nodes are considered to have no corresponding clinical trials and can be excluded from future neural network calculations. If there are 50 or more types of classified clinical trials, the remaining clinical trials can be automatically processed after all previously generated output values ​​have been processed.

[0107]

[0108] In one embodiment, a convolutional neural network can learn by receiving a first learning signal generated by a user's inputted corrected answer when the user identifies a problem in assessing clinical trial participation status based on the convolutional neural network. Problems in assessing clinical trial participation status based on the convolutional neural network may refer to problems with the type of clinical trial or the presence of duplicate clinical trial participation.

[0109] In one embodiment, the first learning signal is generated based on the error between the correct answer and the output value. Depending on the case, methods such as SGD using delta, batching, or backpropagation may be used. Based on the first learning signal, the convolutional neural network performs learning by modifying existing weights, and momentum may be used in some cases. A cost function may be used to calculate the error, and the ss entropy function of the trustee may be used as the cost function.

[0110]

[0111] A neural network according to an embodiment may have as input the type of clinical trial, whether there is duplicate participation in the clinical trial, and an existing first output signal pre-stored in a database, which is an output of a convolutional neural network.

[0112] According to one embodiment, the input of the neural network may be, but is not limited to, a matrix of 1 row and 50 columns, which is the sum of a matrix of 1 row and 50 columns, which includes the type of clinical trial and whether there is overlapping participation in the clinical trial, which is generally the output of a convolutional neural network, and a matrix of 1 row and 500 columns, which is the first output signal. In particular, the number of columns of the first output signal may increase depending on the type of clinical trial.

[0113] Therefore, the input layer nodes that serve as inputs to the neural network can consist of a total of 550 nodes, and the output layer nodes can consist of 5 nodes.

[0114] According to one embodiment, a neural network may be configured to evaluate the type of clinical trial and whether there is duplicate clinical trial participation, which are outputs of a convolutional neural network, based on information contained in a first output signal. The neural network's evaluation may include whether there is duplicate clinical trial participation, etc. If the occurrence of duplicate clinical trial participation is confirmed through the duplicate clinical trial participation, a notification signal may be generated and transmitted to the user via the user terminal (10).

[0115]

[0116] According to one embodiment, a neural network of a big data-based clinical trial information management method has three or more hidden layers, and each hidden layer has 100 nodes, but may be set to more than that in some cases. The activation function of the hidden layer may use, but is not limited to, the ReLU function, the sigmoid function, and the tanh function. The number of output layer nodes may be five, but is not limited thereto. Each output layer node may include the type of target clinical trial, whether or not there is duplicate participation in the clinical trial and the frequency of participation, abnormal signals, and clinical trial participation information.

[0117] The output value of the output layer node changes at 0.02 second intervals, and the corresponding output value can be extracted for each type of clinical trial.

[0118]

[0119] According to one embodiment, the neural network can modify the existing first output signal based on the output results collected over a period of 0.5 seconds, i.e., while the output values ​​of the output layer nodes represent the output results for a total of 25 clinical trials at 0.02 second intervals. The neural network (200) can modify the values ​​of the corresponding clinical trials included in the existing first output signal, and can preserve the values ​​of the clinical trials that were not included in the first input signal created from the captured photos. The neural network can output the second output signal as the output value.

[0120]

[0121] According to one embodiment, a neural network can learn based on user input. If the user determines that the neural network's analysis is incorrect, the user can generate a second learning signal for the neural network (200) to learn.

[0122]

[0123] In one embodiment, the second learning signal is generated based on the error between the correct answer and the output value. Depending on the case, methods such as SGD using delta, batching, or backpropagation may be used. Based on this second learning signal, the neural network performs learning by modifying existing weights, and momentum may be used in some cases. A cost function can be used to calculate the error, which may be a trustee function.

[0124]

[0125] Figure 6 is a diagram showing the configuration of the data analysis system of the present invention.

[0126] Referring to FIG. 6, the data analysis system may include a user terminal (110), an analysis server (210), and an analysis device (230) connected through an analysis network (240).

[0127]

[0128] The user terminal (110) may have the same configuration as the terminal (110) of the recruitment management system, and may use the web service provided by the analysis device (230) via a web browser. The user terminal (110) may transmit clinical trial-related information input by the user via an input means to the analysis device (220), and may receive the prediction results transmitted from the analysis device (220) and output them via an output means.

[0129]

[0130] The analysis server (210) can perform a login procedure at the user's request. That is, when a user inputs his / her identification information, such as an ID and password, through the user terminal (110), the analysis server (210) receives the ID and password input from the user terminal (110), verifies whether the user is registered in the database (DB), and approves or denies the user's permission to use the web service.

[0131] When the analysis device (220) receives clinical trial-related information from a user terminal (110) or an analysis server (210), it can predict the success rate of the clinical trial using a prediction model for which machine learning has been completed, and transmit the prediction result to the user terminal (110) via a network.

[0132] The clinical trial-related information input into the analysis device (220) may include clinical trial performance-related data and clinical trial case data.

[0133]

[0134] Figure 7 is a configuration diagram of a data analysis device according to one embodiment of the present invention.

[0135] Referring to FIG. 7, the analysis device (220) may include a communication unit (221), a memory unit (222), a prediction unit (223), a control unit (224), an input unit (225), and an output unit (226).

[0136]

[0137] The communication unit (221) can perform data communication with the user terminal (110), and can transmit and receive data with the user terminal (110) through an analysis network (230) such as LAN, WAN, Ethernet, WiFi, Wibro, Wimax, and HSDPA.

[0138] The communication unit (221) can receive user information (such as ID and password) and / or clinical trial-related information (clinical trial conditions or clinical trial characteristics) transmitted from the user terminal (110). The communication unit (221) can transmit the clinical trial success rate prediction results to the user terminal (110) via the control unit (224) and can receive clinical trial case data (trial instances).

[0139] Additionally, the communication unit (221) can directly transmit the received data to the prediction unit (223) or transmit it to the prediction unit (223) through the input unit (225).

[0140]

[0141] The memory unit (222) can store a program for the operation of the prediction unit (223) and can also temporarily store input / output data of the prediction unit (223). In addition, the memory unit (222) can store a user DB containing user information.

[0142] The memory unit (222) can store machine learning algorithms, prediction models, learning data, and clinical trial-related information (clinical trial features).

[0143] In addition, the memory unit (222) can store data generated during a learning process using a machine learning algorithm and result values ​​predicted by a prediction model.

[0144] The memory unit (222) may be implemented as at least one storage medium (recording medium) among storage media such as flash memory, a hard disk, an SD card (Secure Digital Card), a random access memory (RAM), a read only memory (ROM), a programmable ROM (PROM), an erasable and programmable ROM (EPROM), an electrically erasable and programmable ROM (EEPROM), a register, a removable disk, and web storage.

[0145] In addition, the memory unit (222) may be a device that stores various databases or tables, etc. For example, the memory unit (222) may store chemical characteristic information of a drug, target-based characteristic information, and / or information on whether a clinical trial for a drug has succeeded or failed. For example, the chemical characteristic information of a drug may include one or more of molecular weight, XLogP, polar surface area, number of hydrogen bond donors, number of hydrogen bond acceptors, formal charge, number of rings, number of rotatable bonds, refractive index, and AlogP solubility, and the target-based characteristic information may include information obtained from one or more of tissues of adipose tissue, adrenal gland, bladder, blood, blood vessel, brain, breast, cervix, colon, esophagus, fallopian tube, heart, kidney, liver, lung, muscle, nerve, ovary, pituitary gland, prostate, salivary gland, skin, small intestine, spleen, stomach, testis, thyroid, uterus, and vagina.

[0146]

[0147] The prediction unit (223) can perform machine learning using clinical trial case data.

[0148] To this end, the prediction unit (223) may include a learning module (223a) and a prediction module (223b) that predicts the clinical trial success rate using a machine-learned prediction model. The prediction model may predict the clinical trial success rate using multiple machine learning algorithms.

[0149]

[0150] The learning module (223a) can perform a three-stage learning process consisting of a first learning stage (training, level 1), a second learning stage (meta-training, level 2), and a testing and optimization stage (testing and optimizing).

[0151] The learning module (223a) can receive learning data (dataset) through the input unit (150) and classify it into a dataset for each learning step.

[0152] The learning module (223a) can enable a plurality of first learning algorithms to learn the relationship between clinical trial conditions (features, Xs) and clinical trial results (Y) through a dataset for the first learning stage in the first learning stage.

[0153] At this time, the first learning algorithm may include a K-Nearest Neighbor (KNN) algorithm, a Gradient Boosting Machine (GBM) algorithm, a Neural Network algorithm, a Random Forest algorithm, extra trees, and a logistic regression algorithm, and these algorithms may have different predictive powers for the same clinical trial.

[0154] The learning module (223a) can perform machine learning to determine the most predictive algorithm among multiple first learning algorithms in the second learning stage. The learning module (223a) can train the second learning algorithm using the dataset for the second learning stage.

[0155] At this time, the second learning algorithm can be implemented as any one of Bayesian Optimization, Particle Swarm Optimization, Genetic Algorithm, and Differential Evolution Algorithm.

[0156]

[0157] Bayesian optimization is an optimization technique for finding the optimal value of a function. It utilizes the principles of Bayesian statistics and probability models to efficiently search for optimal values ​​within the input variables of a given function. Bayesian optimization can identify feature sets with high demand forecasting performance through key concepts such as probability model construction, prior distribution setting, data collection and update, point selection and exploration, function evaluation and update, optimization, and termination. Probabilistic model construction involves constructing a probability model to estimate the function to be optimized (the objective function). This probability model can be expressed as a probability distribution without knowing the function's shape or values. Prior distribution setting involves assigning a prior distribution to the parameters of the probability model. This prior distribution represents the probability distribution over possible parameter values, allowing the function's estimated value to be adjusted. Data collection and update involves sampling a few initial data points and updating the probability model. Point selection and exploration involves selecting the next point to explore based on the probability model. This point selection can be performed by considering unevaluated regions or regions with high function uncertainty. Function evaluation and update can calculate function values ​​at selected points and update the probability model based on these values. Optimization and termination can iterate the optimization process by selecting the next point to explore along with the optimal function value estimated through the probability model, and terminate when no further improvements are possible.

[0158] Particle swarm optimization (PSO) is one of the metaheuristic algorithms for solving optimization problems. It mimics the concept of membership and allows individuals to explore the solution space and find the optimal solution. PSO focuses on finding new solutions and can be used in particular for optimization problems in continuous spaces.

[0159] In particle swarm optimization, each individual is called a "particle." Each particle represents a potential solution and can play a role in finding the optimal solution by exploring and moving through the solution space. Each particle has a position and velocity. The position represents the potential solution, and the velocity determines the direction in which the particle moves in the next step. Furthermore, each particle records the best solution it has found, which is called the "individual optimal solution" and is denoted by pBest. Furthermore, the best solution among all particles so far is found and recorded, which is called the "swarm optimal solution" and is denoted by gBest. Furthermore, each particle can move and update in the next step based on its current velocity and position, and the information in pBest and gBest can be used to update the velocity and position. Furthermore, particles move individually while exploring the solution space, and as individual particles gradually converge to a good solution, the optimal solution for the swarm can also be found and converge. The particle swarm optimization algorithm can be repeatedly executed a certain number of times or until a given termination condition is met, and when the termination condition is satisfied, the algorithm can be terminated and the optimal solution of the cluster finally found can be returned.

[0160] A genetic algorithm is an optimization technique used to solve optimization problems inspired by the theory of biological evolution. It can be operated by simulating the process of finding a solution by imitating the process of mutation, selection, and crossover of genes.

[0161] Genetic algorithms have the advantage of being able to effectively solve complex optimization problems by performing a search process through stages such as individual representation and initialization, fitness evaluation, selection, crossover, mutation, replacement, and iteration.

[0162] Individual representation and initialization represent individuals representing solutions to the problem being optimized and generate an initial population. Each individual can be considered a solution candidate. Fitness evaluation evaluates the fitness of each individual to gauge how well it is a candidate for a solution. The fitness function varies depending on the problem and can be defined according to the optimization goal. Selection selects individuals based on their fitness to become parents of the next generation, and these individuals can be passed on to the next generation. Crossover is the process of exchanging genetic information between selected parents. This creates new offspring individuals, maintaining diversity and helping to explore the solution space. Mutation increases diversity by randomly applying mutations to some individuals. It can introduce new traits or properties to help explore new areas. Replacement selects individuals from the current generation to replace descendants. This process creates a new generation, which can find better solutions than the previous generation. Iteration is the process of repeating the above steps to perform optimization over multiple generations, so that the fitness of individuals can evolve to improve with each generation.

[0163]

[0164] The differential evolution algorithm (DEA) is a metaheuristic technique used to solve function optimization problems. It initially generates random candidate solutions and then uses differential vectors to update these candidate solutions to find the optimal solution. The DEA is applicable to a wide range of problems, exhibits effective performance even with a simple structure, and possesses the advantage of achieving good optimization performance with limited information and low complexity.

[0165] The core idea of ​​the differential evolution algorithm is to calculate the rate of change of the fitness function value among the current solution candidates and select a solution with a higher rate of change. Through this, the solution candidates are evolved and the search process is carried out to find a solution with a high rate of change.

[0166] A differential evolution algorithm may include a solution initialization step, a fitness function evaluation step, a rate of change calculation step, a selection and evolution step, and a convergence verification step.

[0167] The initialization step randomly generates solution candidates. The fitness function evaluation step calculates the fitness function value for each solution candidate. The rate of change calculation step calculates the rate of change in the fitness function value of each solution candidate. The selection and evolution step selects and evolves solution candidates with high rates of change. The selected solution candidates are selected based on the rate of change calculated in the previous step. The convergence verification step verifies whether convergence has occurred after sufficient generations or evolution iterations, and terminates the algorithm if the termination condition is met.

[0168]

[0169] Meanwhile, the learning module (223a) can perform a testing and optimization step after the first and second learning steps are completed. At this time, the learning module (223a) can test the first and second learning algorithms learned using the dataset for the testing and optimization step, and optimize the parameters of each algorithm based on the test results.

[0170] The learning module (223a) can generate a prediction model after testing and optimization of the learned first and second learning algorithms are completed. The learning module (223a) can store the generated prediction model in the memory unit (222). The learning module (223a) can periodically update the prediction model through machine learning.

[0171]

[0172] The prediction module (223b) can receive clinical trial-related information transmitted from the user terminal (110) via the communication unit (221). At this time, the input unit (225) can process the clinical trial-related information received via the communication unit (221) and provide it to the prediction module (223b).

[0173] The prediction module (223b) can predict the success rate of a clinical trial based on the clinical trial-related information received using the prediction model stored in the memory unit (222). The prediction module (223b) can transmit the predicted result (prediction result) using the prediction model to the user terminal (110) that requested the prediction of the success rate of the clinical trial. The user terminal (110) can display the predicted result of the clinical trial success rate provided by the prediction module (223b) on a display.

[0174]

[0175] The control unit (224) is a component that controls the analysis device (220), and may include, for example, a processing unit such as a CPU or GPU. The control unit (224) may train models to be described later using information stored in the memory unit (222), and may also perform prediction value calculation for new inputs using the trained model. Specifically, the control unit (224) may control the prediction unit (223) that predicts the success and failure of a clinical trial. To this end, the control unit (224) may include a control program such as an operating system (OS), a program that defines various processing orders, and an internal memory for storing data. In addition, the control unit (224) may perform information processing to execute various processes by these programs, etc.

[0176]

[0177] The input unit (225) can process data received through the communication unit (221) and transmit it to the prediction unit (223). That is, the input unit (225) can preprocess user information and / or clinical trial-related information into a data format that can be processed by the prediction unit (223) and transmit the preprocessed data to the prediction unit (223).

[0178] The input unit (225) processes clinical trial case data into a machine learning-enabled format and outputs it. For example, the input unit (225) transmits clinical trial case data to the prediction unit (223) in the form of a table consisting of multiple independent variables and one dependent variable (state variable).

[0179] In addition, the input unit (225) generates input data according to the user's operation. The input unit (225) may be composed of a keyboard, a keypad, a touch pad, a touch screen, a mouse, a bar code reader, a QR (Quick Response) code scanner, and a joystick.

[0180] Additionally, the input unit (225) can process clinical trial case data and input it as learning data into the prediction unit (223). The input unit (225) can preprocess a dataset (clinical trial case data) extracted from the FDA's database.

[0181]

[0182] Clinical trial case data may include clinical trial conditions and clinical trial results of actual clinical trial cases, wherein the clinical trial conditions may include at least one of a number of characteristics. The characteristics of the clinical trial conditions may include phase, treatment area, gender of participants, healthiness of participants, number of participants, indication, sponsor, study type, duration, geographical location, molecule type, mechanism of action, target of action, route of administration, and whether the drug is designated by the Ministry of Food and Drug Safety.

[0183]

[0184] The output unit (226) is for outputting information such as visual information, auditory information, and / or tactile information, and may include a display, an audio output module, and a haptic module.

[0185]

[0186] Figure 8 is a flowchart showing a data analysis method according to one embodiment of the present invention.

[0187] Referring to FIG. 8, the data analysis method of the present invention may include a prediction model generation step (S200) of generating a prediction model that predicts clinical trial results by performing machine learning using clinical trial case data; a related information reception step (S210) of receiving clinical trial-related information from a user terminal (110) after the prediction model generation step; and a prediction step (S220) of predicting clinical trial results according to clinical trial-related information using the prediction model after the related information reception step.

[0188] At this time, the prediction model used in the prediction step (S220) can be created through a prediction model creation method, which will be described in detail later with reference to FIG. 9.

[0189]

[0190] Figure 9 is a flowchart of a method for generating a prediction model of the present invention.

[0191] Referring to FIG. 9, the method for generating a prediction model may include a machine learning performing step (S221) of performing machine learning using a plurality of first learning algorithms and one second learning algorithm; a performance index calculating step (S222) of calculating a performance index of a prediction model based on clinical trial results predicted by the plurality of learned first learning algorithms and the second learning algorithm through a test dataset extracted from the clinical trial case data and actual clinical trial results, after the performance index calculating step, an optimization step (S223) of optimizing parameters of the plurality of learned first learning algorithms and the second learning algorithm according to the calculated performance index of the prediction model; and, after the optimization step, a determination step (S224) of performing learning by the second learning algorithm to determine an algorithm having the best predictive ability among the plurality of first learning algorithms.

[0192]

[0193] Figure 10 is a flowchart of a method for generating a prediction model according to another embodiment of the present invention.

[0194] Referring to FIG. 10, a method for generating a prediction model according to another embodiment of the present invention may include a first learning step (S226) in which each of a plurality of first learning algorithms learns a relationship between clinical trial conditions and clinical trial results in a first learning step dataset extracted from clinical trial case data, and a second learning step (S227) in which the plurality of first learning algorithms learned in the first step learn a second learning algorithm by considering the results predicted through clinical trial conditions in a second learning step dataset extracted from clinical trial case data and the clinical trial results in the second learning step dataset.

[0195]

[0196] Figure 11 is a flowchart of a method for predicting clinical trial results according to another embodiment of the present invention.

[0197] Referring to FIG. 11, the method may include a user login step (S2000) in which a login procedure is performed at the analysis server (210) according to a user request; a clinical trial related information input step (S2100) in which clinical trial related information is input through the analysis device (220) user terminal (110) or the analysis server (210) after the user login step; a clinical trial result prediction step (S2200) in which the analysis device (220) predicts clinical trial results using a prediction model in which machine learning has been completed after the clinical trial result prediction step; and a clinical trial prediction result output step (S2300) in which the analysis device (220) outputs the clinical trial results predicted in the clinical trial result prediction step.

[0198]

[0199] More specifically, in the user login step (S2000), the control unit (224) of the analysis device (220) can transmit a web page (login page) for inputting user information for login to the user terminal (110) at the request of the user terminal (110). The user terminal (110) can display the login page on the display screen through the analysis server (210). The user inputs an ID and password by operating the input means of the user terminal (110) and presses the 'sign in' button. The user terminal (110) transmits the ID and password entered by the user to the analysis server (210). The analysis server (210) can receive user information including the ID and password through the analysis network (234) and can confirm whether the user is a registered user based on the received user information and approve or reject the user.

[0200]

[0201] In the clinical trial related information input step (S2100), when the user logs in, the analysis server (210) can provide the user terminal (110) with a web page where the user can input information (clinical trial related information) related to the target clinical trial for which the clinical trial result prediction is to be performed.

[0202] The user terminal (110) can display the corresponding web page on the display screen, and when the user inputs clinical trial-related information into a form within the corresponding web page, the input clinical trial-related information (stage, target disease, subject information, etc.) can be transmitted to the prediction device (100). The input unit (225) of the analysis device (220) can preprocess the clinical trial-related information received through the communication unit (221) and transmit it to the processing unit.

[0203]

[0204] In the clinical trial result prediction step (S2200), the analysis device (220) transmits clinical trial related information received through the analysis network (230) to the prediction module (223b) through the input unit (225), and the prediction module (223b) can predict the clinical trial results, including the success rate of the clinical trial, based on the clinical trial related information using a prediction model.

[0205]

[0206] In the clinical trial prediction result output step (S2300), the analysis device (220) can transmit a web page displaying the predicted clinical trial results to the user terminal (110) or the analysis server (210). The user terminal (110) can display the predicted clinical trial results provided from the analysis device (220). In one embodiment, the clinical trial result display can be displayed in four statuses: achieved, inconclusive, not achieved, and partially achieved, and can be displayed as a probability percentage for each status.

[0207]

[0208] Figure 12 is a diagram showing the configuration of the document management system of the present invention.

[0209] Referring to FIG. 12, the document management system (300) of the present invention may include a management server (310), a database (320), and a management device (330), and each component may be connected to a management network (340).

[0210]

[0211] The document management system (300) is linked to an external clinical trial data management system, so that clinical trial data can be automatically reflected in clinical trial documents. Furthermore, it is linked to a review agency (not shown), so that completed clinical trial documents can be automatically submitted to the review agency without having to separately access the review agency. As the document management system (300) is linked to the clinical trial data management system and the review agency, data collection and document submission can be performed by the document management system (300).

[0212]

[0213] The management server (300) can perform overall tasks for managing clinical trial documents, and can perform information retrieval on clinical trial documents, linkage with review agencies, etc.

[0214]

[0215] The database (320) stores clinical trial documents managed by the management server (300), stores all data for initial writing, additional writing, and revised writing for clinical trial documents for each writing stage, and when information for each stage is requested, data for the corresponding writing stage can be provided.

[0216]

[0217] The management device (330) can perform tasks such as calculating similarity and recommending clinical trial documents by analyzing the similarity of clinical trial documents managed by the management server (310).

[0218]

[0219] Figure 13 is a drawing showing the configuration of the management device of the present invention.

[0220] Referring to FIG. 13, the management device (330) of the present invention may include a processor (331) and a memory (332), and the memory (332) may include a first model (333a), a second model (333b), a third model (333c), a clinical trial document DB (334), and a clinical trial entity DB (335), and at least one of the first model (333a), the second model (333b), and the third model (333c) may be a model based on artificial intelligence acquired through learning by inputting clinical data of clinical trial documents into a neural network model.

[0221]

[0222] The memory (332) can store various programs and data required for the operation of the management device (330) between clinical trial documents.

[0223] Memory (332) can be interconnected with servers operated by government agencies, such as the National Health Insurance Service and the Korea Centers for Disease Control and Prevention's Rare Disease Helpline, to collect and store clinical data defining government support measures for each rare disease in a clinical trial document DB (334). In this case, the clinical data can include information on at least one of the following: the name of the rare disease, symptoms, support amount, support organization, support country, and support procedure.

[0224] Memory (332) extracts and stores, in particular, information on rare diseases from big data managed by a server operated by a government agency. The information on rare diseases may include information on the disease, diagnosis date, medical burden details, residence, treatment hospital, gender, age, and whether medical expenses are covered.

[0225] Memory (332) collects, documents, and stores multiple clinical data provided by each clinical researcher in a clinical trial document database (334). In particular, clinical data specific to rare diseases can be collected. The clinical data may include information on the name of the rare disease, treatment, and clinical trial participation conditions, while the clinical trial participation conditions may in turn include information on the rare disease, age, gender, height, weight, past medical history, and residence.

[0226]

[0227] The first model (333a) may be one that applies context-sensitive embedding to clinical data of clinical trial documents via the processor (331). Context-sensitive word / sentence / entity / document embedding is a technique for expressing words / sentences / entities / documents in a low-dimensional space. Even for words / sentences / entities / documents with the same notation, word / sentence / entity / document embedding is performed differently depending on the context. Therefore, different vector values ​​will be extracted from words / sentences / entities / documents with the same notation depending on the context.

[0228] The first model (333a) can be acquired by the processor (331) through neural network learning to obtain context-based embedding values ​​for the learning clinical trial documents using clinical data of the learning clinical trial documents. Specifically, the first model (333a) can be acquired by the processor (331) through learning to obtain context-based embedding values ​​for each word / sentence / entity / document of the learning clinical trial documents.

[0229] According to an embodiment, a BERT (Bidirectional Encoder Representations from Transformers) model can be used as the first model (333a). The BERT model is an NLP model that learns sentences bidirectionally, and is built by performing pre-training using words from a pre-registered dictionary and fine-tuning the learned model. The BERT model exhibits high accuracy even with a small amount of data by going through a fine-tuning process, and has the advantage of maintaining accuracy even in long sentences because it is an attention-based model that improves performance by drawing attention to specific vectors, so that performance does not decrease even when the sentence becomes longer. However, BERT is only one example, and any model that can extract context-based vector values ​​can be applied to the present invention.

[0230] The first model (333a) may be implemented to apply context-sensitive embedding to entities in clinical trial documents. That is, the first model (333a) may be implemented to output embedding values ​​that reflect the context of each word, sentence, document, and / or entity in the clinical trial documents.

[0231]

[0232] The second model (333b) may be an embedding that reflects keyword frequency in clinical data of clinical trial documents through a processor (331).

[0233] The second model (333b) can be obtained through neural network learning to obtain a second embedding value based on keyword frequency for the learning clinical trial documents using clinical data of the learning clinical trial documents by the processor (331).

[0234] According to an embodiment, the second model (333b) may be a model built through unsupervised learning that discovers criteria for determining how to classify and / or output the embedding values ​​of clinical trial documents based on clinical data of the clinical trial documents.

[0235]

[0236] The third model (333c) may be a model implemented to output entity names included in each clinical trial document from clinical data of the clinical trial documents through the processor (331).

[0237] The third model (333c) can be obtained through neural network learning to obtain entity names included in the learning clinical trial documents by using clinical data of the learning clinical trial documents by the processor (331).

[0238] In an embodiment, a Named Entity Recognition (NER) model may be used as the third model (333c). Named entity recognition recognizes entities with names, and can represent an algorithm that recognizes which type a word belonging to a given name belongs to.

[0239]

[0240] The processor (331) can control the overall operation of the management device (330) between clinical trial documents.

[0241] Specifically, the processor (331) obtains a first embedding value of each of the clinical trial documents based on a first model (333a) from clinical data of the clinical trial documents, calculates a first similarity between the clinical trial documents using each of the obtained first embedding values, obtains a second embedding value of each of the clinical trial documents based on a second model (333b) from clinical data of the clinical trial documents, calculates a second similarity between the clinical trial documents using each of the obtained second embedding values, and calculates a final similarity between the clinical trial documents based on the first similarity and the second similarity.

[0242] The processor (331) can output a context-based first embedding value for clinical trial documents / a keyword frequency-based second embedding value for clinical trial documents / entity names included in clinical trial documents using the first model (333a) / the second model (333b) / the third model (333c).

[0243] In an embodiment, the processor (331) and the first model (333a) / second model (333b) / third model (333c) may be implemented as separate ICs, and according to another embodiment, the processor (331) and the first model (333a) / second model (333b) / third model (333c) may be provided in one IC chip.

[0244] According to an embodiment, the processor (331) and / or the first model (333a) / second model (333b) / third model (333c) may be implemented as a software module or manufactured in the form of at least one hardware chip and mounted on the analysis device (300) between clinical trial documents. For example, it may be manufactured in the form of a dedicated hardware chip for artificial intelligence (AI), such as an NPU (Neural Processing Unit), or manufactured as a part of an existing general-purpose processor (310) (e.g., a CPU or an Application Processor) or a graphics-only processor (310) (e.g., a GPU (Graphics Processing Unit) or a VPU (Visual Processing Unit)) and mounted on the management device (330) between clinical trial documents.

[0245]

[0246] Figure 14 is a flowchart showing a document management method of the present invention.

[0247] First, in this specification, the first user is a clinical trial investigator, and the second user may be at least one of a sponsor, a clinical trial investigator, and a clinical trial research director.

[0248] Additionally, ‘review body’ refers to the Institutional Review Board (IRB).

[0249] In addition, the 'written clinical trial document' refers to a document written before the principal investigator's confirmation is completed, and includes documents written when modifications are made to the document and the version is updated, and refers to a clinical trial document written by the researcher. The written clinical trial document, the reviewed clinical trial document and the final clinical trial document described below include the clinical trial plan (Protocol), consent form, recruitment notice, and clinical trial result report (CSR, Case Study Report). In the case of a clinical trial document initially written by the researcher, it corresponds to a 'first written clinical trial document', and in the case of a document whose version is updated by the researcher through modification, it corresponds to a 'second written clinical trial document'. The version of the document may continue to be updated.

[0250] Additionally, a "Review Clinical Trial Document" is a revision made to a shared written clinical trial document by an investigator, sponsor, or principal investigator. A review clinical trial document can be created by one or more investigators, sponsors, or principal investigators, each of whom can make revisions to the written clinical trial document. Therefore, one or more review clinical trial documents can be created for a single written clinical trial document.

[0251] Additionally, the 'final clinical trial document' is a document that the researcher who wrote the clinical trial document has finally written based on one or more reviewed clinical trial documents, and that has completed final confirmation by the principal investigator.

[0252]

[0253] Referring again to FIG. 14, the clinical trial document management method of the present invention may include a document template selection request receiving step (S300); an information receiving step (S310); a clinical trial document sharing step (S320); and a clinical trial document submission step (S330).

[0254]

[0255] In the document template selection request reception step (S300), the management server (310) may receive a document template selection request from the first user. At this time, the document template is selected from one or more standard document templates, and the standard document template and the new document template may be a clinical trial document creation template.

[0256]

[0257] In the information receiving step (S310), the management server (310) may provide a selected or newly created document template to the first user, and receive input of information that needs to be written on the document template from the first user. At this time, the information that needs to be written on the document template may be information about the relevant clinical trial, and may be information that the first user, who is the writer, determines to be information that needs to be written.

[0258] The first user, as a researcher, can complete the writing of the above-described information on a document template and transmit it to the management server (310).

[0259]

[0260] In the clinical trial document sharing step (S320), in detail, the step may include a first writing clinical trial document sharing step in which the management server (310) shares the first writing clinical trial document, a first review clinical trial document sharing step in which the management server (310) shares the first review clinical trial document, and a first final clinical trial document sharing step in which the management server (310) shares the first final clinical trial document on the document management system (300).

[0261]

[0262] The first clinical trial document sharing step is when the management server (310) shares the first clinical trial document created including information entered by the first user on the document management system (300).

[0263] The management server (310) shares the first clinical trial document, which has been completed, on the document management system (300) at the request of the first user.

[0264] The management server (310) shares the first clinical trial document written on the document management system (300), so that second users such as the sponsor, clinical trial investigator, and clinical trial research manager can review and provide revision opinions, and collate these opinions to provide the first user with the opportunity to revise the document.

[0265] Additionally, it serves to store the version and change history of subsequent documents in the system's database, as well as for exchanging review and revision opinions.

[0266]

[0267] The first review clinical trial document sharing step is a step in which the management server (310) receives one or more first review clinical trial documents including suggested modifications to the first written clinical trial document from one or more second users, and shares the received one or more first review clinical trial documents on the document management system (300).

[0268] At the stage of sharing the first review clinical trial document on the document management system (300), the second user checks the first written clinical trial document written by the first user shared on the document management system (300) and reviews it in real time.

[0269] In the past, the process of writing clinical trial documents involved the first user writing the first clinical trial document, printing it out, and sending it, or sending the first clinical trial document file itself, and having the second user add notes to the document for review.

[0270] The second user sends the document, which has been reviewed including the request for modification, back to the first user, either as a printed document or as a document file itself, and the first user applies modifications to the requested portion.

[0271] Afterwards, the final version of the clinical trial document is submitted to the review agency after going through the review process.

[0272] Therefore, the process of writing a conventional clinical trial document required time for document transmission, a second user to add a note requesting modifications to the document, and the first user to write the document by reflecting the modifications written in the note, so writing the final document also took a lot of time.

[0273] Additionally, if files or output documents are not properly managed in order, it is difficult to manage document versions and change history later.

[0274] However, the present invention allows a second user to check a first clinical trial document shared on a clinical trial document system, and a first review clinical trial document including suggested modifications to the first clinical trial document is written and shared by the second user, thereby enabling a review process to be performed quickly and without hassle.

[0275] One method for creating a first review clinical trial document that includes suggested revisions to the first draft clinical trial document is to create a first review clinical trial document by writing comments in a separate field for the second user's comments on each section of the first draft clinical trial document. In this case, when the first review clinical trial document is shared, the first user can respond to the relevant section to coordinate their comments.

[0276] In another embodiment, a second user can create a first review clinical trial document by directly modifying the content desired in the first draft clinical trial document. In this case, the first user can determine whether to make changes based on the first review clinical trial document containing the modified content.

[0277] Additionally, the first review clinical trial document including the proposed revision may also include the reason for the revision.

[0278] In addition, the present invention has the advantage of making it easy to identify change history later by storing and managing information in the order of document creation, modification, deletion, etc. on a clinical trial document system.

[0279] Additionally, the present invention can insert a watermark into a document when it is modified or updated within the clinical trial documentation system. For example, a watermark indicating the initial modification may be inserted into a document that has been modified for the first time, a watermark indicating the second modification may be inserted into a document that has been modified for the second time, and so on. Watermarks may also be inserted according to updated versions, allowing users to check the number of modifications and updated versions through the watermarks displayed on the document.

[0280]

[0281] The first final clinical trial document sharing step is when the management server (310) shares the first final clinical trial document, which is finally created based on one or more first review clinical trial documents from the first user, on the document management system (300).

[0282] As described above, if one or more first review clinical trial documents in which a second user has written a proposed revision for a part of a first written clinical trial document that requires revision are shared by the management server (310), the first user can write a first final clinical trial document, which is a final document for submission, by referring to the contents of one or more first review clinical trial documents.

[0283] In addition, the first final clinical trial document is not a document written solely by the first user, but is a document that can be submitted to the review agency after completing the final confirmation by the research director.

[0284]

[0285] In the clinical trial document submission step (S330), when the management server (310) receives a request for submission of the first final clinical trial document to the review agency from the first user on the document management system (300), the first final clinical trial document can be submitted to the review agency.

[0286] In the clinical trial document submission step (S330), when the management server (310) receives a request from the first user to submit the first final clinical trial document that has completed the review process by the second user, the first final clinical trial document can be submitted to the review agency.

[0287]

[0288] Figure 15 is a flowchart showing a method for calculating similarity and suggesting documents using the analysis device of the present invention.

[0289] Referring to FIG. 15, the clinical trial document similarity calculation method of the present invention may include a first embedding value acquisition step (S3000), a first similarity calculation step (S3100), a second embedding value acquisition step (S3200); a second similarity calculation step (S3300); and a final similarity calculation step (S3400), a template recommendation step (S3500); and a content recommendation step (S3600).

[0290]

[0291] In the first embedding value acquisition step (S3000), the processor (331) can output word-by-word / sentence-by-sentence embedding values ​​reflecting word-by-word / sentence-by-sentence context from clinical data using the first model (333a).

[0292] Specifically, the processor (331) may perform preprocessing on clinical data of clinical trial documents. The processor (331) may perform different preprocessing depending on whether the clinical data of the clinical trial documents is structured or unstructured. For example, if the clinical data is structured, preprocessing may be performed by extracting metadata of the clinical data. On the other hand, if the clinical data is unstructured, preprocessing may be performed to delete preset clinically unusable words and / or parts of speech from the clinical data. For example, clinically unusable parts of speech may include articles, prepositions, conjunctions, and exclamations.

[0293] The processor (331) can perform morphological analysis on each preprocessed word to generate tokens that pair the corresponding morpheme with each word. Furthermore, the processor (331) can input tokenized words into the first model (333a) to output word-by-word embedding values. Furthermore, the processor (331) can input sentences composed of tokenized words into the first model (333a) to output sentence-by-sentence embedding values. The processor (331) can output the word-by-word / sentence embedding values ​​for each clinical trial document.

[0294]

[0295] In the first similarity calculation step (S3100), the first similarity between clinical trial documents can be calculated using the obtained first embedding values. At this time, rather than comparing the word-by-word / sentence-by-sentence / entity-by-entity embedding values ​​of the clinical trial documents with each other, the word-by-word / sentence-by-sentence / entity-by-entity embedding values ​​are used to separately obtain the first embedding values ​​of the clinical trial documents themselves, which can then be used to determine similarity.

[0296]

[0297] In the second embedding value acquisition step (S3200), the processor (331) can acquire the second embedding value of each of the clinical trial documents based on the second model (333b) from the clinical data of the clinical trial documents.

[0298] Specifically, the processor (331) may perform preprocessing on clinical data from clinical trial documents. Preprocessing may be performed to remove predetermined clinically unusable words and / or parts of speech from each clinical trial document. For example, clinically unusable parts of speech may include articles, prepositions, conjunctions, and exclamations.

[0299] In addition, the processor (331) can generate a matrix representing the relationship between preprocessed clinical trial documents and predefined keywords. In particular, the processor (331) can pre-generate a keyword dictionary specialized for clinical data of clinical trial documents and generate a matrix using the generated keyword dictionary.

[0300]

[0301] In the second similarity calculation step (S3300), the processor (331) can calculate the second similarity between clinical trial documents using each of the acquired second embedding values.

[0302] According to an embodiment, the second similarity between clinical trial documents can be calculated by applying a known similarity calculation algorithm while taking the inner product of the second embedding values ​​of each clinical trial document. In other words, according to the present invention, the second embedding values ​​of the clinical trial documents themselves can be obtained and used to determine similarity.

[0303]

[0304] In the final similarity calculation step (S3400), the processor (331) can calculate the final similarity between clinical trial documents based on the first similarity and the second similarity.

[0305] According to an embodiment, the processor (331) can calculate the final similarity between clinical trial documents by adding the result value in which the first weight is applied to the first similarity and the result value in which the second weight is applied to the second similarity.

[0306] According to an embodiment, the processor (331) separately records the final similarity between clinical trial documents whose sum is greater than or equal to a preset threshold value, and separately records and sorts clinical trial documents whose final similarity is greater than or equal to the preset threshold value, thereby enabling quick access to the document with the highest similarity.

[0307]

[0308] In the template recommendation step (S3500), the processor (331) may recommend a clinical trial document template having a high degree of similarity to the corresponding clinical trial. According to an embodiment, the processor (331) may calculate a final degree of similarity and input the calculated final degree of similarity into the memory (332), thereby recommending a clinical trial document template having the highest degree of similarity to the corresponding clinical trial document among the clinical trial documents included in the clinical trial document DB (334) of the memory (332).

[0309] In this way, the difficulty associated with creating a new clinical trial document template can be reduced by allowing recommendations for clinical trial document templates that are most similar to the clinical trial in question.

[0310]

[0311] In the content recommendation step (S3600), the processor (331) can learn the recommended document template and recommend the composition of the recommended document template, text and / or images in the document, etc.

[0312] By recommending content in this way, users can easily apply the content they need to clinical trial document templates, etc.

[0313]

[0314] In addition, although not shown, according to one embodiment of the present invention, after the template content recommendation step, a clinical trial result output step for performing basic analysis and outputting clinical trial result data may be further included.

[0315] In the clinical trial result output stage, the processor (331) can output clinical trial result data in a table format and perform experimental methodology, basic statistical analysis derived from basic analysis of clinical trial result data, etc.

[0316]

[0317] The embodiments described above may be implemented using hardware components, software components, and / or a combination of hardware components and software components. For example, the devices, methods, and components described in the embodiments may be implemented using one or more general-purpose computers or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor (310), a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing instructions and responding to them. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. The processing device may also access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing device is sometimes described as being used alone; however, one of ordinary skill in the art will recognize that the processing device may include multiple processing elements and / or multiple types of processing elements. For example, the processing device may include multiple processors (310) or one processor (310) and one controller. Other processing configurations, such as a parallel processor (310), are also possible.

[0318]

[0319] The method according to the embodiment may be implemented in the form of program commands that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program commands, data files, data structures, etc., alone or in combination. The program commands recorded on the medium may be those specially designed and configured for the embodiment or may be those known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program commands, such as ROMs, RAMs, and flash memory (320). Examples of the program commands include not only machine language codes generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter, etc. The hardware devices may be configured to operate as one or more software modules to perform the operations of the embodiment, and vice versa.

[0320]

[0321] Software may include a computer program, code, instructions, or a combination of one or more of these, and may configure a processing device to perform a desired operation or, independently or collectively, command the processing device. The software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave, for interpretation by the processing device or for providing instructions or data to the processing device. The software may also be distributed over networked computer systems and stored or executed in a distributed manner. The software and data may be stored on one or more computer-readable recording media.

[0322]

[0323] Although the embodiments described above have been described with limited drawings, those skilled in the art will appreciate that various technical modifications and variations can be applied based on the described techniques. For example, appropriate results can still be achieved even if the described techniques are performed in a different order than described, and / or components of the described systems, structures, devices, circuits, etc. are combined or combined in a different manner than described, or are replaced or substituted with other components or equivalents.

Claims

1. Includes an integrated system (10) that provides clinical trial support services through an integrated network (40) to clinical trial demanders (20) who require clinical trials and clinical trial suppliers (30) who conduct clinical trials. The above integrated system (10), A recruitment management system (100) that performs a service of recruiting clinical trial participants at the request of the clinical trial demander (20) or clinical trial supplier (30); A data analysis system (200) that receives clinical trial-related information from the clinical trial supplier (30), analyzes clinical trial-related data, and predicts clinical trial results; and A clinical trial integrated support platform including a document management system (300) for managing the above clinical trial related documents; 2. In claim 1, The above recruitment management system (100) A center server (130) that transmits and receives signals and data to and from a user terminal (110) through a recruitment network (120) and provides clinical trial information and performs a health check process through the user terminal; A contracted institution server (140) that performs tasks such as writing a clinical trial plan, transmitting the clinical trial plan, and generating review results; A request server (150) that transmits a request for clinical trial to a contract institution server (140) through a recruitment network (120) for clinical trial subjects; An authentication server (160) that receives information on selected clinical trial participants from a center server (130) through a recruitment network (120), and then transmits a consent form to a user terminal (110) corresponding to the clinical trial participant information according to a clinical trial consent request recommended to the user terminal (110) included in the clinical trial participant information; and A clinical trial integrated support platform including a smart server (170) that receives user image information from a user terminal (110) through a recruitment network (120) and provides clinical trial participation information to the user terminal (110) using the received user image.

3. In claim 1, The above data analysis system (200) An analysis device (220) that receives clinical trial-related information and predicts clinical trial results using a prediction model for which machine learning has been completed; A user terminal (110) connected to the above analysis device and analysis network (230) to receive the prediction result of the prediction device; A clinical trial integrated support platform including an analysis server (210) that provides data for predictive processing of the above analysis device (220) and performs a login procedure according to a user's request; 4. In claim 1, The above document management system (300) Management server (310); A database (320) in which clinical trial documents managed by the above management server (310) are stored; A clinical trial integrated support platform including a management device (330) that calculates the similarity of clinical trial documents managed by the above management server (310); 5. In a recruitment management method using the recruitment management system (100) of claim 1, A clinical trial request request step (S100) in which a request request server (150) requests a clinical trial request to a contract recipient server (140) through a recruitment network (120); After the clinical trial request request step, the clinical trial plan provision step (S110) in which the contract institution server (140) provides the clinical trial plan to the center server (130) through the recruitment network (120); After the clinical trial plan provision step, a review result generation step (S120) is performed in which the center server (130) generates the IRB review result for the clinical trial plan received in the clinical trial plan provision step; After the IRB review result generation step, the clinical trial participant recruitment request step (S130) in which the contracted institution server (140) transmits a request for clinical trial participant recruitment to the center server (130) through the recruitment network (120); After the clinical trial participant recruitment request step, the clinical trial participant selection step (S140) in which the center server (130) selects clinical trial participants; A clinical trial participant information creation step (S150) that creates information on clinical trial participants selected in the clinical trial participant selection step; A clinical trial integration support method, including a clinical trial participant information transmission step (S160) for transmitting clinical trial participant information created in the clinical trial participant creation step to a center server (300); 6. In a data analysis method using the data analysis system (200) of claim 1, A prediction model generation step (S200) that generates a prediction model that predicts clinical trial results by performing machine learning using clinical trial case data; After the above prediction model generation step, a related information receiving step (S210) for receiving clinical trial related information from a user terminal (110); A clinical trial integrated support method, including a prediction step (S220) for predicting clinical trial results according to clinical trial related information using a prediction model after the above-mentioned related information receiving step; 7. In a document management method using the management system (300) of claim 1, A document template selection request receiving step (S300) in which the clinical trial document management server (100) receives a document template selection request from a first user; After the document template selection request reception step, the management server (310) provides the selected or newly created document template to the first user and receives the input of information required to be written on the document template from the first user (S310); After the above information receiving step, the clinical trial document sharing step (S320) in which the management server (310) shares the first written clinical trial document with the document management system (300); A clinical trial integrated support method, including a clinical trial document submission step (S330) for submitting the first final clinical trial document to the review agency when the management server (310) receives a request for submission of the first final clinical trial document to the review agency from the first user on the document management system (300) after the above clinical trial document sharing step;

Images