Clinical test data intelligent analysis method and system

Abstract

The invention relates to the technical field of data processing, in particular to an intelligent analysis method and system for clinical test data, and the method comprises the steps: obtaining a data sequence of each physiological index of a patient, and calculating the deviation degree through quantifying the difference between a target data point and a corresponding physiological index normal range, calculating an abnormal index based on the deviation degree and the distribution characteristics of the corresponding data in the data sequence of the corresponding physiological indexes, and obtaining a fitting curve of each physiological index by fitting the abnormal index, so as to screen out time periods in which all the fitting curves are in a rising trend; and taking the physiological index pairs of which the correlation is greater than a preset value in the time period meeting the screening condition as target index pairs, constructing a causal graph by using a PC algorithm based on the target index pairs, and identifying the causal relationship of each physiological index. According to the method, the complexity of constructing the causal graph by the PC algorithm can be reduced while the important causal relationship among the physiological indexes is reserved.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing. More specifically, the present invention relates to a method and system for intelligent analysis of clinical trial data. Background Art

[0002] With the rapid development of modern clinical trials, data acquisition technologies have become increasingly diverse, and massive heterogeneous data has been generated from data sources such as electronic medical records, gene sequencing, wearable sensors, and medical images. Especially in the scenario of remote patient monitoring, wearable devices collect multi-dimensional physiological indicators such as heart rate, blood pressure, blood glucose level, and exercise volume in real time, forming a high-density and high-dimensional time-series data stream. The scale and complexity of such data pose severe challenges to traditional causal inference methods.

[0003] In related technologies, constraint-based causal discovery algorithms, such as the PC algorithm (Peter-Clark Algorithm), can identify causal relationships between different health indicators and are usually used to construct causal graphs between health indicators and predict the effects of interventions. The core steps of the PC algorithm include: (1) constructing a complete undirected graph; (2) gradually removing irrelevant edges through zero-order and higher-order conditional independence tests; (3) determining the direction of causal relationships based on orientation rules.

[0004] However, in high-dimensional and large-scale data scenarios, higher-order conditional independence tests need to traverse exponentially growing combinations of conditional variables, resulting in a sharp increase in computational complexity. For example, for a dataset containing dozens of physiological indicators such as heart rate, blood pressure, and blood glucose level, the algorithm needs to perform millions of independence tests, which not only takes too long but also is prone to misjudgment due to multiple hypothesis testing problems (such as false causal relationships or missed true associations), thereby affecting the effect prediction of intervention measures, leading to clinical decision-making biases, and reducing the credibility of the system. Summary of the Invention

[0005] In order to solve the problem that due to the huge amount of clinical trial data, the computational complexity of the PC algorithm in identifying causal relationships between various physiological indicators is large, thereby affecting the accuracy of the constructed causal graph, the present invention provides a method and system for intelligent analysis of clinical trial data.

[0006] According to a first aspect of the present invention, there is provided a method for intelligent analysis of clinical trial data, including: Obtaining clinical trial data of a patient and performing preprocessing to obtain data sequences of various physiological indicators aligned in time, where the physiological indicators include heart rate, blood pressure, and blood glucose level; Selecting target data points from the data sequence of the target physiological indicator, and determining the deviation degree of the target data points by quantifying the difference between the target data points and the normal range of the target physiological indicator at the corresponding moment; Based on the degree of deviation, as well as the occurrence frequency and average time interval of data points with the same value as the target data point in the data sequence of the target physiological index, calculate the anomaly index of the target data point. The anomaly index is positively correlated with both the occurrence frequency and the degree of deviation, and negatively correlated with the average time interval; By fitting the anomaly indices of the data points in the data sequences of each physiological index, obtain the fitting curves of each physiological index. Screen the time periods during which all the fitting curves show an upward trend, and take the pairs of physiological indices with a correlation greater than a preset value within the time periods that meet the screening conditions as the target index pairs, so as to construct a causal graph using the PC algorithm based on the target index pairs and identify the causal relationships of each physiological index.

[0007] The present invention comprehensively considers the anomaly index and time persistence, screens the target index pairs, and can accurately identify the physiological indices that have a significant impact on the patient's health, enabling the causal graph constructed based on the screened target indices to intuitively show the causal relationships of each physiological index of the patient. Thus, while reducing the amount of data to be processed when constructing the causal graph using the PC algorithm, the accuracy of the constructed causal graph is ensured.

[0008] Preferably, the method for obtaining the degree of deviation of the target data point includes: If the value of the target data point is greater than the upper bound value of the normal range, then take the difference between the value of the target data point and the upper bound value, and divide it by the upper bound value as the degree of deviation of the target data point; If the value of the target data point is less than the lower bound value of the normal range, then take the difference between the lower bound value and the value of the data point, and divide it by the lower bound value as the degree of deviation of the target data point; If the value of the target data point is within the normal range, then set the degree of deviation of the target data point to zero.

[0009] The present invention can standardize the degree of deviation to a relative ratio, which enables the degrees of deviation of different indices to be compared on the same scale, avoiding the difficulty of comparison caused by differences in index dimensions or numerical ranges.

[0010] Preferably, when the healthy range of the target physiological index is affected by the patient's exercise amount, the method for obtaining the normal range includes: Take the product of the difference between the patient's exercise amount at the moment corresponding to the target data point and the preset exercise amount threshold and the preset weight as the change amount, obtain the static healthy range of the target physiological index, and update the lower bound value and the upper bound value of the static healthy range through a summation operation based on the change amount; Obtain the maximum value among the lower bound values before and after the update, and the minimum value among the upper bound values before and after the update, and take the range composed of the maximum value and the minimum value as the normal range.

[0011] By dynamically adjusting the normal ranges of various physiological indicators, the present invention can avoid misjudging the normal changes of physiological indicators affected by the patient's exercise volume as abnormalities, ensuring the accuracy of the deviation degrees of each data point.

[0012] Preferably, when the healthy range of the target physiological indicator is a fixed range, the static healthy range of the target physiological indicator is used as the normal range of the target physiological indicator at each moment.

[0013] Preferably, the method for obtaining the abnormality index of the target data point includes: Taking the opposite number of the average time interval as the power of the exponential function and performing a power operation to obtain a severity index; Performing a normalization process on the occurrence frequency, calculating the sum of the normalized values obtained and the severity index, and performing a normalization process on the product of the sum and the severity to obtain the abnormality index of the target data point.

[0014] The present invention can comprehensively consider the frequency and time interval of abnormalities, and obtain a quantified abnormality index through weighted summation and normalization processing, which is comparable, enabling the abnormality indexes of different data points to be compared on the same scale, thereby providing a quantified basis for subsequent screening operations.

[0015] Preferably, by fitting the abnormality indexes of the data points in the data sequence of each physiological indicator, the fitting curve of each physiological indicator is obtained, including: Using the least squares method to perform curve fitting on the abnormality indexes of all data points in the data sequence of each physiological indicator respectively, to obtain the fitting curve of each physiological index.

[0016] Preferably, screening the time periods during which all the fitting curves show an upward trend includes: Obtaining the slope values of the data points on each fitting curve. If the slope values of the data points on all the fitting curves within any time period are positive, then retain that any time period to obtain the time periods that meet the screening conditions.

[0017] Preferably, the method for obtaining the target indicator pair includes: For any time period that meets the screening conditions, calculate the absolute value of the Spearman correlation coefficient of the data sequences of any two physiological indicators within any time period to obtain the correlation of the corresponding physiological indicator pair, and use the physiological indicator pair with a correlation greater than the preset correlation threshold as the target indicator pair.

[0018] The present invention uses correlation to screen the target indicator pair, which can ensure that there is a relatively important causal relationship in the screened target indicator pair, thereby providing a data basis for the subsequent construction of the causal diagram.

[0019] Preferably, when constructing a causal graph using the PC algorithm based on target metrics, the conditional variables for any pair of target metrics are physiological metrics with an average correlation greater than a set value with the pair of target metrics at the same time period.

[0020] According to a second aspect of the present invention, there is provided an intelligent analysis method for clinical trial data. The system includes a memory and a processor. A computer program is stored on the memory, and the processor executes the computer program to implement the steps of the first aspect of the present invention.

[0021] The present invention has the following effects: The present invention determines the anomaly index of each data point by integrating various indicators, can accurately evaluate the possibility of anomalies in each data point, and through the fitting method, can facilitate the determination of the change trend of each physiological indicator. Thus, based on the anomaly duration of each physiological indicator, by evaluating the correlation between any two pairs of physiological indicators, important pairs of physiological indicators, that is, target metric pairs, can be quickly and accurately screened out. When constructing a causal graph based on the screened target metric pairs, the causal relationship between each physiological indicator can be more accurately identified, so that while effectively reducing the computational complexity of the PC algorithm when processing large-scale data, the accuracy of the constructed causal graph can be ensured. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] 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 drawings, several embodiments of the present invention are shown by way of illustration and not limitation, and like or corresponding reference numerals indicate like or corresponding parts, wherein: Figure 1 is a schematic flowchart of the steps of an intelligent analysis method for clinical trial data according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0024] The following will describe the specific embodiments of the present invention in detail with reference to the drawings.

[0025] Referring to Figure 1 , an intelligent analysis method for clinical trial data includes steps S1 - S4, specifically as follows: S1: Obtain the clinical trial data of the patient and perform preprocessing to obtain data sequences of various physiological indicators that are aligned in time. The physiological indicators include heart rate, blood pressure, and blood glucose level.

[0026] Specifically, according to a fixed sampling frequency, such as once per minute, continuously collect the clinical trial data (such as heart rate, blood pressure, blood glucose level, etc.) of the patient within the most recent day from the patient's wearable device, sensor, or electronic medical record system to obtain a heart rate data sequence, a blood pressure data sequence, a blood glucose level data sequence, etc.

[0027] Optionally, the total number of samplings can be denoted as , then the heart rate data sequence can be denoted as ; the blood pressure data sequence can be denoted as ; the blood glucose level data sequence can be denoted as ; where , and are the heart rate value, blood pressure value, and blood glucose level value of the th sampling respectively; is the sampling order; is the total number of samplings.

[0028] It should be noted that in order to eliminate the influence of the dimensions of different physiological indicators and align the data sequences of various physiological indicators in time, it is necessary to perform normalization processing on the data sequences of each physiological indicator respectively, such as maximum-minimum normalization processing.

[0029] Moreover, when the sampling frequencies of different types of sensors are different, resampling techniques (such as upsampling or downsampling) may be required to align the data sequences of various physiological indicators in time.

[0030] Optionally, when there are missing values for any physiological indicator at any sampling moment, linear interpolation is used to replace the missing values to ensure the integrity of the data of the physiological indicators at each sampling moment. Among them, normalization processing, resampling techniques, and linear interpolation method are all existing technologies, and will not be elaborated in detail in this embodiment.

[0031] S2: Select target data points from the data sequence of the target physiological indicator, and determine the deviation degree of the target data points by quantifying the difference between the target data points and the normal range of the target physiological indicator at the corresponding moment.

[0032] Among them, the target physiological indicator refers to a randomly selected physiological indicator; the target data point refers to a randomly selected data point from the data sequence of the target physiological indicator; the deviation degree refers to the deviation size between the value of the target data point and the normal level of the target physiological indicator.

[0033] In an exemplary embodiment of the present invention, when the healthy range of a target physiological index is affected by the patient's exercise amount, the determination of the normal range of the target physiological index at each moment can be achieved through the following steps: Multiply the difference between the patient's exercise amount at the moment corresponding to the target data point and the preset exercise amount threshold by the preset weight as the change amount, obtain the static healthy range of the target physiological index, and update the lower bound value and the upper bound value of the static healthy range through a summation operation based on the change amount; obtain the maximum value among the lower bound values before and after the update, and the minimum value among the upper bound values before and after the update, and use the range composed of the maximum value and the minimum value as the normal range.

[0034] It should be noted that during exercise, the changes in human physiological indexes (such as heart rate and blood glucose level) are normal physiological reactions rather than health problems. If the healthy range in the static state (static healthy range) is used to evaluate the data during exercise, misjudgment may occur. Therefore, this embodiment proposes a method for dynamically adjusting the healthy range. For physiological indexes affected by the patient's exercise amount, the normal range is adjusted in real time according to the patient's exercise state, thereby avoiding misjudgment and improving the accuracy of evaluation.

[0035] Specifically, the normal range of the heart rate at any moment satisfies the following relational expression: ; In the formula, is the normal range of the heart rate at the th moment; 60 and 100 are respectively the lower bound value and the upper bound value of the static healthy range of the heart rate. Among them, the static healthy range of the heart rate refers to the healthy range of the heart rate monitored when the patient's exercise amount is lower than the set exercise amount threshold, usually 60 - 100 beats per minute; , are respectively functions that return the maximum value and the minimum value; is the patient's exercise amount at the th moment; is the preset exercise amount threshold, is a value evaluated by a doctor for the patient's actual exercise ability and can effectively distinguish the patient's physiological state; 0.01 is the preset weight used to control the adjustment range of the heart rate boundary value.

[0036] The normal range of the blood glucose level at any moment , satisfies the following relational expression: ; In the formula, is the normal range of the blood glucose level at the The normal range at a moment; 3.9 and 6.1 are respectively the lower and upper bound values of the static healthy range of blood glucose level. Among them, the static healthy range of blood glucose level refers to the healthy range of the monitored blood glucose level when the patient's exercise amount is lower than the set exercise amount threshold, usually 3.9 - 6.1 mmol / L; 0.02 is the preset weight of the blood glucose level, used to control the adjustment range of the blood glucose level boundary value.

[0037] In an exemplary embodiment of the present invention, when the healthy range of the target physiological index is a fixed range, the static healthy range of the target physiological index is taken as the normal range of the target physiological index at each moment.

[0038] For example, for physiological indexes whose healthy range is not affected by the patient's exercise amount, such as blood pressure, the static healthy range of blood pressure can be directly taken as the normal range of blood pressure at each moment. Among them, the static healthy range of blood pressure refers to the healthy range of the monitored blood pressure when the patient's exercise amount is lower than the set exercise amount threshold, usually 70 - 100 mmHg.

[0039] Further, after determining the normal range of each physiological index at each moment, the deviation degree of each data point in the data sequence of each physiological index can be calculated.

[0040] In an exemplary embodiment of the present invention, the determination of the deviation degree of each data point in the data sequence of each physiological index can be achieved through the following steps: If the value of the target data point is greater than the upper bound value of the normal range, the difference between the value of the target data point and the upper bound value, divided by the upper bound value, is taken as the deviation degree of the target data point; If the value of the target data point is less than the lower bound value of the normal range, the difference between the lower bound value and the value of the data point, divided by the lower bound value, is taken as the deviation degree of the target data point; If the value of the target data point is within the normal range, the deviation degree of the target data point is set to zero.

[0041] Optionally, if the target data point is the th data point in the heart rate data sequence, then the lower limit value of the normal range of the heart rate at the th moment can be denoted as , and the upper limit value of the normal range of the heart rate at the th moment can be denoted as , then the deviation degree of the target data point satisfies the relational expression: ; In the formula, is the deviation degree of the th data point in the heart rate data sequence; is the th data point in the heart rate data sequence; , are respectively the lower limit value and the upper limit value of the normal range of the heart rate at the th moment; represents that the value of is within the normal range of the heart rate at the th moment.

[0042] It should be noted that the method for determining the deviation degree of each data point in the data sequence of other physiological indicators is the same as that for determining the deviation degree of each data point in the heart rate data sequence, and will not be elaborated in this embodiment.

[0043] S3: Based on this deviation degree, as well as the occurrence frequency and average time interval of the data points with the same value as the target data point in the data sequence of the target physiological indicator, calculate the abnormality index of the target data point. The abnormality index is positively correlated with both the occurrence frequency and the deviation degree, and negatively correlated with the average time interval.

[0044] In an exemplary embodiment of the present invention, the abnormality index of each data point in the data sequence of each physiological indicator can be determined through the following steps: Take the opposite number of the average time interval as the power of the exponential function, and perform power operation to obtain the severity index; normalize the occurrence frequency, calculate the sum of the obtained normalized value and the severity index, and normalize the product of the sum and the severity to obtain the abnormality index of the target data point.

[0045] Exemplarily, when the target data point is the th data point in the heart rate data sequence, the abnormality index of the target data point satisfies the following relational expression: ; In the formula, is the abnormality index of the th data point in the heart rate data sequence; is the deviation degree of the th data point in the heart rate data sequence; is the number of data points with the same value as the th data point in the heart rate data sequence; is the total number of data points in the heart rate data sequence; is the average time interval of the data points with the same value as the th data point in the heart rate data sequence; is the exponential function with the natural constant as the base; is the normalization function.

[0046] Among them, reflects the occurrence frequency of data points in the heart rate data sequence that have the same value as the th data point; reflects the severity index of the th data point in the heart rate data sequence. The larger this value is, the denser the distribution of data points in the heart rate data sequence that have the same value as the th data point. At this time, if is relatively large, it indicates that the probability of abnormal heart rate at this moment is relatively large, and the corresponding abnormality index is relatively large.

[0047] Optionally, other normalization methods can also be used, such as normalizing using the sigmoid function. This embodiment does not make special limitations on the selected normalization method.

[0048] It should be noted that the method for obtaining the abnormality index of each data point in the data sequence of other physiological indicators is the same as the method for determining the abnormality index of each data point in the heart rate data sequence, and this embodiment will not elaborate here.

[0049] In another embodiment, the abnormality index of each data point in the heart rate data sequence can also be calculated through the relational expression: .

[0050] S4: By fitting the abnormality indices of the data points in the data sequences of each physiological indicator, obtain the fitting curves of the abnormality indices of each physiological indicator, screen the time periods during which all the fitting curves show an upward trend, and use the pairs of physiological indicators with a correlation greater than a preset value within the time periods that meet the screening conditions as the target indicator pairs, so as to construct a causal graph based on the target indicator pairs and identify the causal relationships of each physiological indicator.

[0051] It should be noted that when the present invention constructs a causal graph using the PC algorithm, by screening and retaining relatively important variables, the number of physiological indicators to be processed can be reduced, thereby effectively reducing the computational complexity while ensuring the accuracy of the causal relationships in the constructed causal graph.

[0052] Among them, the present invention only adds a data screening process when constructing a causal graph, and does not improve other contents of the PC algorithm, such as the construction method of the undirected graph, the test process of conditional independence, and the determination process of directed edges, etc.

[0053] In an exemplary embodiment of the present invention, the fitting curves of each physiological indicator can be determined through the following steps: Using the least squares method, perform curve fitting on the abnormality indices of all data points in the data sequences of each physiological indicator respectively to obtain the fitting curves of each physiological index.

[0054] Exemplarily, for the heart rate data sequence, the least squares method can be used to perform curve fitting on the abnormal indices of all data points in the heart rate data sequence, so as to obtain the fitting curve of the heart rate. Similarly, the fitting curve of the blood glucose level and the fitting curve of the blood pressure can be determined according to the determination method of the fitting curve of the heart rate. It should be noted that the process of performing curve fitting using the least squares method is a prior art, and this embodiment will not elaborate on it here.

[0055] Optionally, other fitting methods, such as spline interpolation, can also be used for curve fitting, and this embodiment does not make special limitations on the selected fitting method.

[0056] Furthermore, after obtaining the fitting curves of each physiological index, time intervals in which the changing trends of all fitting curves are synchronously rising can be screened out, so as to determine one or more eligible time periods. It should be noted that by screening out the time intervals in which the changing trends of all fitting curves are synchronously rising, the present invention aims to measure the duration during which the abnormal indices of all physiological indices increase. These screened time periods are exactly the time periods when all physiological indices are most likely to be abnormal, thereby providing an important basis for subsequent determination of target index pairs.

[0057] In an exemplary embodiment of the present invention, the determination of the time period satisfying the screening conditions can be achieved through the following steps: Obtain the slope values of the data points on each fitting curve. If the slope values of the data points on all fitting curves within any time period are positive, then retain that any time period to obtain the time period satisfying the screening conditions.

[0058] It should be noted that in the mathematical representation of the fitting curve, the first derivative (i.e., the slope value) of each data point essentially reflects the change rate of the physiological index near that time point, and its sign and magnitude can directly quantify the trend direction and intensity change. Based on this feature, by calculating the slope values of the data points on each fitting curve, the present invention can accurately evaluate the changing trends of each physiological index, thereby screening out one or more eligible time periods.

[0059] In an exemplary embodiment of the present invention, the determination of the target index pair can be achieved through the following steps: For any time period satisfying the screening conditions, calculate the absolute value of the Spearman correlation coefficient of the data sequences of any two physiological indices within any time period to obtain the correlation of the corresponding physiological index pair, and use the physiological index pair with a correlation greater than the preset correlation threshold as the target index pair.

[0060] Exemplarily, the correlation threshold can be set to 0.5, and any time period satisfying the screening conditions is denoted as , the Spearman correlation coefficient between any two of the heart rate data sequence, blood pressure data sequence, and blood glucose level data sequence during this period can be calculated, denoted as , where and are any two different physiological indicators. And when , then and these two physiological indicators are retained, so as to obtain a pair of target indicators within this time period, and then pairs of target indicators within all time periods that meet the screening conditions can be obtained. In this embodiment, the size of the correlation threshold is not particularly limited.

[0061] In another embodiment, the absolute value of the Pearson correlation coefficient of any two data sequences can also be used as the correlation between the corresponding two data sequences. It should be noted that the determination methods of the Spearman correlation coefficient and the Pearson correlation coefficient are both prior arts, and this embodiment will not elaborate on them here.

[0062] Furthermore, after determining all pairs of target indicators, the PC algorithm can be used to construct the causal graph corresponding to all pairs of target indicators through the steps of initializing a complete undirected graph, gradually performing conditional independence tests, and determining the edge directions. It should be noted that, given the input data, the process of constructing a causal graph using the PC algorithm is a prior art, and this embodiment will not elaborate on it here.

[0063] In an exemplary embodiment of the present invention, when constructing a causal graph using the PC algorithm based on pairs of target indicators, the conditional variable of any pair of target indicators is a physiological indicator whose average correlation with this pair of target indicators during the same period is greater than a set value.

[0064] Exemplarily, when it is determined that within the time period of , and are a pair of target indicators, at this time, the data sequences of physiological indicators other than and during this time period can be calculated, and the correlations with the data sequences of and during this time period are calculated, and the sum is averaged. If the obtained average value is greater than the set value, such as 0.6, then the corresponding physiological indicator can be used as the conditional variable of this pair of target indicators, so that based on the determined conditional variable, the step of gradually performing conditional independence tests on this pair of target indicators can be executed to improve the test efficiency when performing the step of gradually performing conditional independence tests on each pair of target indicators. In this embodiment, the determination method of the correlation is the same as the calculation method of the correlation when screening pairs of target indicators.

[0065] Furthermore, after determining the causal graph of a patient, the causal relationships of the patient's various physiological indicators can be identified by comparing the patient's causal graph with that of the general population. For example, if the patient's causal graph shows that the blood glucose level has a greater impact on the heart rate, then special attention needs to be paid to the control of the patient's blood glucose level, thereby providing more targeted reference for doctors' diagnosis and treatment.

[0066] The present invention also provides an intelligent analysis system for clinical trial data. The system includes a memory and a processor, and a computer program is stored on the memory. The computer program integrates the functions of an intelligent analysis method for clinical trial data. When the computer program is executed, an intelligent analysis method for clinical trial data can reduce the complexity of constructing a causal graph by the PC algorithm while retaining the important causal relationships between various physiological indicators.

[0067] In the description of this specification, the meanings of "a plurality of" and "several" are at least two, such as two, three or more, etc., unless otherwise specifically and clearly defined.

[0068] Although this specification has shown and described multiple embodiments of the present invention, it is obvious to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will think of many changes, alterations and alternative ways without departing from the spirit and idea of the present invention. It should be understood that various alternative solutions to the embodiments of the present invention described herein can be adopted in the process of practicing the present invention.

Claims

1. An intelligent analysis method for clinical trial data, characterized in that, Including: Obtain the clinical trial data of the patient and perform preprocessing to obtain the data sequences of various physiological indicators aligned in time, where the physiological indicators include heart rate, blood pressure, and blood glucose level; Select target data points from the data sequence of the target physiological indicator, and determine the deviation degree of the target data point by quantifying the difference between the target data point and the normal range of the target physiological indicator at the corresponding moment; Based on the deviation degree, as well as the occurrence frequency and average time interval of the data points with the same value as the target data point in the data sequence of the target physiological indicator, calculate the abnormality index of the target data point. The abnormality index is positively correlated with both the occurrence frequency and the deviation degree, and negatively correlated with the average time interval; By fitting the abnormality indices of the data points in the data sequences of various physiological indicators, obtain the fitting curves of various physiological indicators, screen the time periods when all the fitting curves show an upward trend, and take the pairs of physiological indicators with a correlation greater than a preset value within the time periods that meet the screening conditions as the target indicator pairs, so as to construct a causal graph using the PC algorithm based on the target indicator pairs to identify the causal relationships of various physiological indicators.

2. The intelligent analysis method for clinical trial data according to claim 1, characterized in that The method for obtaining the deviation degree of the target data point includes: If the value of the target data point is greater than the upper bound value of the normal range, then take the difference between the value of the target data point and the upper bound value, and divide it by the upper bound value as the deviation degree of the target data point; If the value of the target data point is less than the lower bound value of the normal range, then take the difference between the lower bound value and the value of the data point, and divide it by the lower bound value as the deviation degree of the target data point; If the value of the target data point is within the normal range, then set the deviation degree of the target data point to zero.

3. The intelligent analysis method for clinical trial data according to claim 2, wherein When the healthy range of the target physiological indicator is affected by the patient's exercise amount, the method for obtaining the normal range includes: Take the product of the difference between the patient's exercise amount at the moment corresponding to the target data point and the preset exercise amount threshold and the preset weight as the change amount, obtain the static healthy range of the target physiological indicator, and update the lower bound value and the upper bound value of the static healthy range through a summation operation based on the change amount; Obtain the maximum value among the lower bound values before and after the update, and the minimum value among the upper bound values before and after the update, and take the range composed of the maximum value and the minimum value as the normal range.

4. The intelligent analysis method for clinical trial data according to claim 2, wherein, When the healthy range of the target physiological indicator is a fixed range, then take the static healthy range of the target physiological indicator as the normal range of the target physiological indicator at each moment.

5. The intelligent analysis method for clinical trial data according to claim 1, wherein The method for obtaining the abnormality index of the target data point includes: Take the opposite number of the average time interval as the power of the exponential function, and perform a power operation to obtain the severity index; Perform a normalization process on the occurrence frequency, calculate the accumulated sum of the obtained normalized value and the severity index, and perform a normalization process on the product of the accumulated sum and the severity to obtain the abnormality index of the target data point.

6. The intelligent analysis method for clinical trial data according to claim 1, wherein The method of obtaining the fitting curves of various physiological indicators by fitting the abnormality indices of the data points in the data sequences of various physiological indicators includes: Use the least squares method to perform curve fitting on the abnormality indices of all the data points in the data sequences of various physiological indicators respectively to obtain the fitting curves of various physiological indices.

7. An intelligent analysis method for clinical trial data according to claim 6, wherein The period of time during which all the fitted curves show an upward trend includes: Obtain the slope values of the data points in each of the fitted curves. If the slope values of the data points on all the fitted curves are positive within any period of time, then retain that period of time to obtain the period of time that meets the screening criteria.

8. The intelligent analysis method for clinical trial data according to claim 1, wherein The method for obtaining the target index pairs includes: For any period of time that meets the screening criteria, calculate the absolute value of the Spearman correlation coefficient of the data sequences of any two physiological indexes within any period of time to obtain the correlation of the corresponding physiological index pair, and use the physiological index pair with a correlation greater than the preset correlation threshold as the target index pair.

9. The intelligent analysis method for clinical trial data according to claim 1, wherein When constructing a causal graph using the PC algorithm based on the target index pairs, the conditional variable of any target index pair is the physiological index whose average correlation with that target index pair in the same period is greater than the set value.

10. An intelligent analysis system for clinical trial data, characterized in that, The intelligent analysis system for clinical trial data includes a memory and a processor. A computer program is stored on the memory, and the processor executes the computer program to implement the steps of the intelligent analysis method for clinical trial data according to any one of claims 1-9.

Images