A safety detection method for clinical trial data

Abstract

The present invention relates to the field of medical data processing, and in particular to a safety detection method for clinical trial data, comprising: continuously collecting clinical trial data of patients at various times over multiple days, performing periodic analysis and volatility analysis to extract periodic characteristic values ​​and volatility characteristic values ​​of the clinical trial data of each day, obtaining an initial period number by comprehensively considering the volatility characteristic values ​​and the periodic characteristic values, and adjusting the initial period number according to the noise level of the clinical trial data, thereby effectively reducing the influence of noise on prediction results, improving prediction accuracy, and realizing adaptive adjustment of the period number, constructing a moving average prediction model using the adjusted initial period number, predicting clinical trial data after the current day, performing safety detection based on the deviation between the predicted value and the true value, being able to monitor the clinical trial data in real time and accurately, effectively preventing and identifying potential safety risks, and providing a certain safety guarantee for clinical trial data management.

Description

Technical Field

[0001] The present invention relates to the field of medical data processing, and in particular to a safety detection method for clinical trial data. Background Art

[0002] With the rapid development of modern medical technology and the increasing complexity of clinical trials, ensuring the security and accuracy of medical data has become the basic guarantee for the credibility of scientific research results. The diversity and complexity of medical data have brought unprecedented challenges to medical data management.

[0003] In this context, PBRTQC (patient data real-time quality control) as an advanced medical data processing technology plays a vital role. By deeply analyzing, monitoring and evaluating medical data in real time, this technology can keenly detect and promptly correct errors and anomalies in the data, ensure the accuracy and completeness of medical data, thereby minimizing the deviation of experimental results caused by data errors and deviations, greatly improving the credibility and reliability of clinical trial data in the medical process, and ensuring the security of clinical trial data.

[0004] As an effective prediction method, the moving average method is often used in PBRTQC technology. It mainly predicts the clinical trial data of patients and judges whether there are abnormalities in the clinical trial data based on the deviation value between the predicted results and the actual results, so as to make safety warnings. For example, the patent document with publication number CN110090005A discloses a medical data processing method, which mainly performs discretization sampling processing and moving average processing on the sampled signal to obtain the electromyography sampling signal for subsequent analysis and processing. Another example is the patent document with publication number CN117095828A, which discloses a blood sugar value prediction and alarm method based on the follow-up records of diabetic patients, which mainly collects the patient's blood sugar value data, performs interpolation filling after cleaning, and finally uses the exponentially weighted moving average method to smooth and denoise the data. Finally, a model is built to predict the blood sugar trend and generate warnings when abnormal.

[0005] In the above-mentioned moving average method, "period number" is a key parameter of the moving average method. The size of the period number directly affects the accuracy of the prediction results. Although a smaller period number can respond quickly to data changes, it may cause the results to be too sensitive to noise and fail to effectively smooth the data. On the contrary, although a larger period number can better smooth the data, it may cause the prediction results to lag and fail to reflect the true status of the data in real time. When the above-mentioned technical solution applies the moving average method to clinical trial data prediction, it is difficult to take into account both the data response speed and the smoothing effect because the period number is often set manually, resulting in poor accuracy and timeliness of the prediction results, which in turn affects the safety of clinical trial data in the medical process. Summary of the invention

[0006] In order to solve the problem that when the moving average method is applied to clinical trial data prediction, the artificial setting of the period parameter is difficult to take into account both the data response speed and the smoothing effect, resulting in poor accuracy and timeliness of the prediction results, thereby affecting the security of the clinical trial data, the present invention proposes a clinical trial data security detection method, which specifically adopts the following scheme:

[0007] A safety detection method for clinical trial data, comprising:

[0008] Collect clinical trial data of patients for multiple consecutive days, and perform periodic analysis and volatility analysis on the clinical trial data to extract periodic characteristic values ​​and volatility characteristic values ​​of the clinical trial data for each day;

[0009] Calculate the initial number of periods per day used to build the moving average forecast model ,in, is the serial number of the day, For the The initial period of days, For the The volatility characteristic value of the clinical trial data of the day, For the The periodic characteristic value of clinical trial data of days, is the ceiling function, is the absolute value function;

[0010] Obtaining the fitting value of the clinical trial data of each day through data fitting, and taking the difference between the fitting value and the actual value as the noise level of the clinical trial data, and adjusting the initial period number according to the size of the noise level;

[0011] The clinical trial data are predicted based on the moving average prediction model corresponding to the adjusted initial period number, and the clinical trial data are safety tested based on the prediction results.

[0012] The present invention obtains the periodic characteristic value and volatility characteristic value of the clinical trial data of each day by performing periodic analysis and volatility analysis on the clinical trial data, dynamically determines the initial period number for constructing a moving average prediction model in combination with the periodic characteristic value and the volatility characteristic value, realizes adaptive adjustment of the period number, enhances the adaptability of the model to data characteristics, and accurately quantifies the noise level of the data through data fitting, and corrects the initial period number accordingly, effectively reduces the impact of noise on the prediction result, improves the prediction accuracy, constructs a moving average prediction model using the adjusted initial period number, predicts the clinical trial data after the day, and performs safety detection based on the deviation between the predicted value and the actual value, thereby realizing real-time and accurate monitoring of the clinical trial data, effectively preventing and identifying potential safety risks, and providing a certain safety guarantee for clinical trial data management.

[0013] Preferably, the initial period number also includes an adjustment process, which is: obtaining the fitting value of the clinical trial data of each day through data fitting, and taking the difference between the fitting value and the actual value as the noise level of the clinical trial data; taking the average noise level of the clinical trial data at all times of each day as the overall noise level of the clinical trial data of each day; if the overall noise level is less than a preset noise level threshold, and the initial period number is equal to the preset adjustment factor, the initial period number remains unchanged; if the overall noise level is less than a preset noise level threshold, and the initial period number is not equal to the preset adjustment factor, the initial period number is subtracted from the preset adjustment factor.

[0014] The initial number of periods of the prediction model is adaptively adjusted according to the noise intensity in the actual data to effectively cope with data fluctuations, ensure that the model robustness is enhanced when the noise intensifies, improve the prediction stability, reduce the model complexity when the noise is small, and maintain sensitivity, thereby optimizing the prediction performance under different conditions and improving the accuracy and reliability of clinical trial data analysis.

[0015] Preferably, the method of predicting the clinical trial data after each day according to the moving average prediction model corresponding to the adjusted initial period number includes:

[0016] set up For the Days adjusted initial period number;

[0017] The moving average prediction model is constructed as:

[0018]

[0019] In the formula, For the Days The predicted value of clinical trial data at a certain moment, For the Days Clinical trial data at the moment, is the sequence number of the moment, is the day number, The value of All integers in the range.

[0020] By adjusting the moving average prediction model of the initial period, the data from multiple days in the past are flexibly and adaptively integrated to provide more accurate predictions for clinical trials on each subsequent day.

[0021] Preferably, whether there is a safety issue in the clinical trial data is judged based on the difference between the predicted value and the actual value in the prediction result, including: if the deviation between the predicted value and the actual value exceeds a preset deviation threshold, the clinical trial data has a safety issue; otherwise, the clinical trial data does not have a safety issue.

[0022] Preferably, the method for extracting the periodic characteristic value of the daily clinical trial data is: select any moment of the day; take the sum of the difference values ​​of the clinical trial data at the moment and the previous moment and the next moment respectively as the change rate of the clinical trial data at the moment; take the average of the difference values ​​between the change rate of the clinical trial data at the moment and the change rate of the clinical trial data at all the same moments in multiple consecutive days as the periodic characteristic value of the clinical trial data at the moment; the difference value is the absolute value of the difference; take the average of the periodic characteristic values ​​of the clinical trial data at all moments of the day as the periodic characteristic value of the clinical trial data of the day.

[0023] By analyzing the differences between clinical data at a single moment and adjacent moments, quantifying the rate of change, and comparing it with the average rate of change over multiple days, the periodic features are effectively extracted. Overall, this strategy improves the accuracy of identifying periodic features, enabling clinical research to more accurately grasp the periodic change trends of data.

[0024] Preferably, the method for extracting the volatility characteristic value of the daily clinical trial data is: selecting any moment of the day, taking the average of the difference values ​​between the clinical trial data at that moment and other clinical trial data within its local range as the fluctuation amplitude of the clinical trial data at that moment; the difference value is the absolute value of the difference; taking the variance of the fluctuation amplitude of the clinical trial data at all moments of the day as the volatility characteristic value of the clinical trial data of that day.

[0025] By comparing the data at a single moment with the average difference within its local range, the fluctuation amplitude was quantified, and then the fluctuation characteristics of the whole day were summarized, which effectively captured the local change intensity and overall variation of the clinical data, and improved the sensitivity and accuracy of identifying daily fluctuation characteristics.

[0026] Preferably, the local range is the previous All clinical trial data within the hour is regarded as the local scope of the clinical trial data. is the default value.

[0027] The present invention has the following effects:

[0028] The present invention obtains the periodic characteristic value and volatility characteristic value of the clinical trial data for each day by performing periodic analysis and volatility analysis on the clinical trial data, and dynamically determines the initial period number for constructing the moving average prediction model in combination with the periodic characteristic value and the volatility characteristic value, thereby realizing adaptive adjustment of the period number and enhancing the adaptability of the model to data characteristics, and accurately quantifies the noise level of the data through data fitting, and corrects the initial period number accordingly, effectively reducing the impact of noise on the prediction results and improving the prediction accuracy, and judging whether the clinical trial data is abnormal based on the deviation between the accurate predicted value and the actual value, thereby realizing real-time and accurate monitoring of the clinical trial data, effectively preventing and identifying potential safety risks, and providing a certain safety guarantee for clinical trial data management. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0030] Figure 1 It is a schematic flow chart of the method of the present invention. DETAILED DESCRIPTION

[0031] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0032] The specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0033] Reference Figure 1 , a safety detection method for clinical trial data, comprising steps S1-S6, specifically as follows:

[0034] S1: Collect clinical trial data of patients for multiple consecutive days.

[0035] In the solution for clinical trial data security detection in this embodiment, the collected clinical trial data include blood pressure, blood sugar, heart rate and any other physiological parameters that require real-time monitoring and safety assurance. In this embodiment, the patient's heart rate data is used as an example for explanation, so the "clinical trial data" referred to below refers to heart rate data. The remaining parameter data can also be predicted according to this method. This method has wide applicability and good versatility. Through this dynamically optimized prediction method, the safety and reliability of clinical trial data can be comprehensively improved.

[0036] First, the clinical trial data of patients are collected in real time at a sampling frequency of once per second. These raw data are seamlessly transmitted to the detection system built based on the real-time quality control method of patient data to ensure the real-time and accuracy of the data.

[0037] Then, taking the implementation day as the starting point, the clinical trial data of the patient in the past week (7 days) are obtained, and based on these clinical trial data, the clinical trial data on the eighth day can be predicted.

[0038] S2: Perform periodic analysis and volatility analysis on the clinical trial data to extract periodic characteristic values ​​and volatility characteristic values ​​of the clinical trial data for each day.

[0039] Because the human heart rate is precisely regulated by the autonomic nervous system, it exhibits obvious circadian rhythmic changes, that is, there is a significant difference between the resting and awake states, usually manifested as an increased heart rate when awake and a decreased heart rate when asleep. In addition, factors such as the time accuracy of heart rate measuring instruments and individual biological clocks may also cause clinical trial data to reflect certain periodic characteristics. In a clinical trial environment, a patient's heart rate will also be affected by daily life activities and emotional fluctuations in the immediate future. For example, exercise and emotional tension can cause the heart rate to increase instantly, and these short-term changes are distinctive, giving clinical trial data a strong short-term volatility. At the same time, drug interventions or treatments used in clinical trials, such as Beta-blockers and calcium channel blockers may cause a decrease in heart rate, while electrophysiological therapy or the recovery phase after surgery may cause a temporary increase in heart rate. The combination of these internal and external factors further exacerbates the periodic and volatile changes in clinical trial data during and after treatment. When predicting patients' clinical trial data, the volatility and periodic characteristics of the data must be taken into account when adapting the moving average method's interim number. Among them, the periodic changes are mainly related to the patient's inherent physiological routine, showing a relatively stable and long period; while the volatility changes are mainly due to factors such as drug use, daily life activities and emotional reactions, and are highly unstable and immediate. Therefore, when adaptively determining the number of periods in the heart rate prediction model, although the periodic law needs to be considered, it is more focused on capturing and adapting to the volatility changes caused by various immediate factors, thereby improving the sensitivity and accuracy of the prediction model to changes in clinical trial data.

[0040] The clinical trial data is subjected to periodic analysis and volatility analysis to extract periodic characteristic values ​​of the clinical trial data for each day, as follows:

[0041] (1) Analyze the periodic characteristic values ​​of clinical trial data at each moment of the day, including:

[0042] First, calculate the rate of change of clinical trial data at each moment:

[0043] The difference between the clinical trial data at this moment and the clinical trial data at the previous moment is recorded as the first difference value, and the difference between the clinical trial data at this moment and the clinical trial data at the next moment is recorded as the second difference value. The absolute values ​​of the first difference value and the second difference value are calculated, and the accumulated value is used as the rate of change of the clinical trial data at this moment, wherein the difference value refers to the absolute value of the difference.

[0044] First Days For example, the moment Days The rate of change of clinical trial data collected at all times is expressed as:

[0045]

[0046] In the formula, For the Days The value of clinical trial data collected at all times, For the Days Collect clinical trial data at all times. For the Days The value of clinical trial data collected at all times, for and The difference value of for and The difference value of .

[0047] Then, based on the change rate of the clinical trial data at each moment of each day and the change rate of the clinical trial data at the same moment before the day, the periodic characteristic value of the clinical trial data at each moment of the day is evaluated, including:

[0048] First Days are the object of evaluation;

[0049] Calculate the The rate of change of clinical trial data at each moment of the day;

[0050] Get the Day before The change rate of clinical trial data at the same time of the day is obtained The change rate of clinical trial data is recorded as the set of change rates of clinical trial data at the same time. is the serial number of the day, is the default value, and , Indicates taking and The smaller value between .

[0051] For example , indicating the 6th day, If is 2, Take 2, and take the first moment on the 6th day as an example. At this time, obtain the change rate of the clinical trial data at the first moment on the 2 days before the 6th day, that is, obtain the change rate of the clinical trial data at the first moment on the 4th day and the change rate of the clinical trial data at the first moment on the 5th day, and record these two change rates as the change rate set of the clinical trial data at the first moment;

[0052] Next, calculate the The difference between the change rate of the clinical trial data at each moment of the day and the change rate of each clinical trial data in the change rate set of the clinical trial data at the same moment is used to obtain the change rate difference value set of the clinical trial data at the same moment. For example, the difference between the change rate of the clinical trial data at the first moment of the 6th day and the change rate of the clinical trial data at the first moment of the 4th day, and the difference between the change rate of the clinical trial data at the first moment of the 6th day and the change rate of the clinical trial data at the first moment of the 5th day are calculated, and these two difference values ​​are used as the change rate difference value set of the clinical trial data at the first moment;

[0053] Finally, the mean of the set of change rate difference values ​​of the clinical trial data at the same time is taken as the first The periodic characteristic value of the clinical trial data at this time of the day.

[0054] For example, the mean of the set of change rate difference values ​​of the clinical trial data at the first moment is used as the periodic characteristic value of the clinical trial data at the first moment within the sixth day.

[0055] If the Days For example, the moment Days Periodic characteristic values ​​of clinical trial data collected at all times The formula is:

[0056]

[0057] In the formula, For the Days The rate of change of clinical trial data collected at all times, For the Days The rate of change of clinical trial data collected at all times, that is, Day before Days The rate of change of clinical trial data collected at all times, express and The difference value of is the absolute value function, For the front The total number of changes in clinical trial data collected within a day, The value of All integers in the range, The value of ,express Pick and The smaller value in ensures the rationality of the value. For example, if the preset value is 5, is 4. If not used If the function takes the smaller value of the two, it will cause an unreasonable situation such as "taking the first 5 days of the 4th day". After that, the value selection will become "take the 3 days before the 4th day", ensuring the rationality of the value.

[0058] Based on the above formula, in another specific example:

[0059] like , , then on the 4th day The periodic characteristic value of the clinical trial data collected at the time is obtained by calculating the periodic characteristic value of the clinical trial data collected at the time. The change rate of the clinical trial data collected at the moment is compared with the change rate of the first three days (i.e., the first day, the second day, and the third day) The difference value (absolute value of the difference) of the change rate of the clinical trial data collected at each time point is 3 difference values ​​in total. The average of these 3 difference values ​​is used as the first Periodic characteristic values ​​of clinical trial data collected at all times.

[0060] (2) The average of the periodic characteristic values ​​of the clinical trial data at all times of the day is taken as the periodic characteristic value of the clinical trial data of the day.

[0061] The clinical trial data is subjected to periodicity analysis and volatility analysis, and the volatility characteristic values ​​of the clinical trial data are extracted every day, as follows:

[0062] First, set the local range of clinical trial data at each moment of the day. The difference of data within the local range can reflect the fluctuation of data within the local range. Therefore, set the previous range of clinical trial data at each moment of the day. Hour is the local range of the moment. In this embodiment, , take the clinical trial data collected within 1 hour before each moment as the local range data at that moment, 1 is the experience value, which can be set by yourself, because it is a local range analysis, The value should not be too large.

[0063] Then, obtain the fluctuation range of clinical trial data at each moment, specifically:

[0064] According to the difference between the clinical trial data at that moment and other clinical trial data in the local range, a difference value set is obtained; the mean of the difference value set is used as the fluctuation range of the clinical trial data at that moment, where the difference value refers to the absolute value of the difference, such as the difference between 1 and 2 is .

[0065] In a specific example, Days Taking the moment as an example, the fluctuation range of the clinical trial data at that moment is expressed by the formula:

[0066]

[0067] In the formula, For the Days The fluctuation range of clinical trial data at each moment, For the The total number of clinical trial data contained in the local scope at the moment, is the serial number of the clinical trial data contained in the local scope, The value of All integers in the range, For the Days The value of clinical trial data collected at all times, For the The local range of the moment The value of clinical trial data, express and The difference value of is the absolute value function.

[0068] Finally, according to the fluctuation range of the clinical trial data at each moment of the day, the volatility characteristic value of the clinical trial data of the day is obtained, including:

[0069] The variance of the fluctuation range of the clinical trial data at all times within the day is calculated, and the variance is used as the volatility characteristic value of the clinical trial data of the day.

[0070] In a specific example, For example, The volatility characteristic value of the clinical trial data of 10 days is expressed as follows:

[0071]

[0072] In the formula, For the The volatility characteristic value of the clinical trial data of the day, Indicates The number of moments in the day, Indicates the serial number of the time. The value of All integers in the range, Indicates Days The fluctuation range of clinical trial data at each moment, For the The average fluctuation range of clinical trial data at all times during the day, the overall expression of the formula The variance of the fluctuation range of clinical trial data at all times of the day.

[0073] When predicting the heart rate data for the day, the initial period setting of the prediction model should be appropriately reduced for the characteristics of strong data volatility. This is because greater volatility may cause rapid changes in data, and if the period is too large, important information points may be missed. On the contrary, when the heart rate data of the day shows obvious periodicity, the impact of periodic characteristics should be considered more when adaptively adjusting the initial period to ensure that the prediction model can fit and capture the periodic change trend of the data to improve the accuracy and reliability of the prediction. In other words, the situation of large volatility tends to choose a smaller initial period, while the situation of prominent periodicity will prompt us to pay more attention to the adaptation and adjustment of the initial period to match the periodicity.

[0074] S3: Calculate the initial number of periods used to build the moving average forecast model each day.

[0075] In the actual application of the PBRTQC method, the performance of the data prediction module was particularly improved, aiming to reduce the subjective bias caused by human settings by adaptively adjusting the key parameters (number of periods) of the prediction model, thereby enhancing the objectivity and accuracy of the prediction results, which is crucial for the safety monitoring of clinical trial data.

[0076] The initial number of periods used to construct the moving average forecast model for the day is determined according to the following relationship:

[0077]

[0078] In the formula, is the serial number of the day, For the The initial period of the day, For the The volatility characteristic value of the clinical trial data of the day, For the The periodic characteristic value of clinical trial data of days, is the ceiling function, It is an absolute value function to avoid negative results.

[0079] This model integrates quantitative indicators of data volatility and periodicity of clinical trial data, allowing the model to automatically adjust prediction parameters based on actual data volatility and periodicity, reducing the subjectivity of manually preset periodic numbers, improving the adaptability and generalization ability of the model, and being suitable for clinical data changes under different conditions. It can more accurately predict clinical data trends, reduce prediction errors, and improve the objectivity and accuracy of prediction results.

[0080] S4: Perform data fitting on the clinical trial data, calculate the noise level based on the data fitting results, and use the noise level to adjust the initial number of periods.

[0081] The adjustment process of the initial period is:

[0082] Data fitting is performed on the clinical trial data at all times of each day to obtain the fitting value of the clinical trial data at each time. The difference between the fitting value and the actual value of the clinical trial data at each time is used as the noise level of the clinical trial data at that time. Furthermore, the mean of the noise level of the clinical trial data at all times of each day is calculated, and the value after normalization of the mean is used as the overall noise level of the clinical trial data of each day. The least squares method can be used for data fitting.

[0083] The noise level threshold and the initial period adjustment factor are preset. In this embodiment, the noise level threshold is set to 0.62, and the initial period adjustment factor is set to 1.

[0084] When the overall noise level of the clinical trial data on a certain day is greater than or equal to 0.62, the clinical trial data on that day is considered to have a large noise performance, and the initial period number used to construct the moving average prediction model on that day is added with the period adjustment factor to expand the initial period number;

[0085] If the overall noise level of the day is less than 0.62, and the initial period number is equal to the preset adjustment factor, the initial period number remains unchanged, avoiding the situation where the period number is 0 after adjustment;

[0086] If the overall noise level of the day is less than 0.62, and the initial period number is not equal to the preset adjustment factor, the preset adjustment factor is subtracted from the initial period number to reduce the initial period number, thereby completing the adjustment process of the initial period number.

[0087] S5: Predict the clinical trial data based on the moving average prediction model corresponding to the adjusted initial period number, and perform safety testing based on the prediction results.

[0088] First, build a moving average forecasting model, including:

[0089] set up For the Days adjusted initial period number;

[0090] The moving average prediction model is constructed as:

[0091]

[0092] In the formula, For the Days Clinical trial data at the moment, For the Days Clinical trial data at the moment, is the day number, The value of All integers in the range, is the sequence number of the moment. Through this prediction model, we can get the The predicted value of clinical trial data at every moment of the day.

[0093] For example, the initial period after adjustment on the 7th day has been calculated to be 3, that is, , ,but The range is ;

[0094] Then the first Clinical trial data at the moment, within 6 days Clinical trial data at the moment, within 7 days The clinical trial data at the moment, the average of these three data, is used as the first The value of clinical trial data at a certain moment.

[0095] Then, based on the difference between the predicted value and the actual value in the prediction result, it is judged whether there are safety issues in the clinical trial data, including:

[0096] According to The difference between the actual value and the predicted value of the clinical trial data at each moment of the day can be analyzed by multivariate statistical analysis methods, such as variance analysis, regression analysis, hypothesis testing, etc. (in actual application, the appropriate analysis method can be flexibly selected according to the data characteristics and research needs), and a threshold standard is set. In this embodiment, the threshold is set to 3, which is an empirical reference value and can be dynamically adjusted according to specific practical conditions and data analysis results. When it is found that the deviation between the actual value and the predicted value exceeds the set threshold, the system will automatically trigger the early warning mechanism to prompt possible safety risks or data quality problems. For example, it is predicted that the first The value of the clinical trial data at the moment is 100, and the actual value of the clinical trial data at the 8th day is The value of the clinical trial data at a certain moment is 95, which indicates that there may be a safety risk. This mechanism aims to identify potential abnormal conditions in a timely manner and provide real-time and effective decision-making basis for the safety monitoring of clinical trial data.

[0097] In the description of this specification, "plurality" or "several" means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0098] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A safety detection method for clinical trial data, characterized in that: include: Collect clinical trial data of patients for multiple consecutive days, and perform periodic analysis and volatility analysis on the clinical trial data to extract periodic characteristic values ​​and volatility characteristic values ​​of the clinical trial data for each day; Calculate the initial number of periods per day used to build the moving average forecast model ,in, is the serial number of the day, For the The initial period of days, For the The volatility characteristic value of the clinical trial data of the day, For the The periodic characteristic value of clinical trial data of days, is the ceiling function, is the absolute value function; Predicting the clinical trial data according to the moving average prediction model corresponding to the initial period number, and judging whether there is a safety problem in the clinical trial data according to the difference between the predicted value and the actual value in the prediction result; Also included is a process for adjusting the initial period number: Performing data fitting on the daily clinical trial data by the least squares method to obtain a fitting value, and taking the difference between the fitting value and the actual value as the noise level of the clinical trial data; The average noise level of the clinical trial data at all times of the day is normalized to be the overall noise level of the clinical trial data of the day; If the overall noise level is greater than or equal to a preset noise level threshold, the initial period number is added with a preset adjustment factor, otherwise, the initial period number is subtracted with a preset adjustment factor; If the initial period number is equal to the preset adjustment factor, and the noise level of the clinical trial data on that day is less than the noise level threshold, the initial period number remains unchanged; The method for extracting the periodic characteristic value of the daily clinical trial data is: Select any time of the day; The sum of the difference values ​​of the clinical trial data at the moment, the previous moment, and the next moment is taken as the change rate of the clinical trial data at the moment, and the average of the difference between the change rate of the clinical trial data at the moment and the change rate of all clinical trial data at the same moment in multiple consecutive days is taken as the periodic characteristic value of the clinical trial data at the moment; the difference value is the absolute value of the difference; The mean of the periodic characteristic values ​​of the clinical trial data at all times of the day is taken as the periodic characteristic value of the clinical trial data of the day; The method for extracting the volatility characteristic value of the daily clinical trial data is: Select any time of the day and take the average of the difference between the clinical trial data at that time and other clinical trial data within its local range as the fluctuation range of the clinical trial data at that time; The variance of the fluctuation range of the clinical trial data at all times within the day is taken as the volatility characteristic value of the clinical trial data of the day; Predicting the clinical trial data according to the moving average prediction model corresponding to the initial period number includes: The moving average prediction model is constructed as: In the formula, For the Days The predicted value of clinical trial data at a certain moment, For the Days Clinical trial data at the moment, is the sequence number of the moment, The value of All integers in the range; The method of judging whether there is a safety problem in the clinical trial data based on the difference between the predicted value and the actual value in the prediction result includes: If the deviation between the predicted value and the actual value exceeds a preset deviation threshold, the clinical trial data has a safety issue; otherwise, the clinical trial data does not have a safety issue.

2. The method for safety detection of clinical trial data according to claim 1, characterized in that: The local scope is to take all clinical trial data within the first M hours of the clinical trial data as the local scope of the clinical trial data, where M is a preset value.

Images