A clinical trial data real-time acquisition and quality control management platform

Abstract

The present application relates to the technical field of clinical trial data collection and intelligent quality control, in particular to a clinical trial data real-time collection and quality control management platform, in the present application, an operation starting time recording unit acquires a subject disinfection operation starting time stamp, an operation sequence delay calculation unit calculates an actual operation delay based on a standard time consumption, a transmission delay processing unit separates transmission and operation delay by double time stamp, an actual collection time determination unit subtracts pure transmission delay time data and operation delay time data from the device upload time stamp by a reverse time restoration algorithm to obtain a real blood collection completion time, a time window dynamic determination unit determines eligibility by comparing real time and a preset time window through dynamic boundary offset, a physiological index correction unit corrects blood glucose and insulin values in real time by a hierarchical architecture and a correction rule library, and a dynamic correlation logic verification unit adjusts correlation rules and verifies index logic by combining a three-dimensional decision matrix to provide support for clinical trials.

Description

Technical Field

[0001] This invention relates to the field of clinical trial data acquisition and intelligent quality control technology, specifically, to a real-time clinical trial data acquisition and quality control management platform. Background Technology

[0002] Clinical trial data collection and intelligent quality control is an important technology, specifically applied to the real-time management of physiological indicators such as blood glucose and insulin during the blood collection process in clinical trials. The core is to ensure the authenticity and reliability of data by accurately handling time deviations and operational differences, so as to provide accurate data support for trial conclusions. Currently, clinical trial data collection relies heavily on basic platforms to record data, but it is difficult to adapt to the complex interference factors in the blood collection scenario.

[0003] Existing clinical trial data acquisition and quality control platforms cannot eliminate the combined interference of blood collection operation delays and Bluetooth transmission delays, leading to deviations in the calculation of actual blood collection completion time. This results in erroneous rejection or retention of data in the acceptance criteria for the collection time window. Furthermore, they do not provide real-time correction for the impact of blood collection posture on blood glucose levels and blood collection site on insulin levels. The original indicators are distorted due to operational differences. In addition, the logic verification does not combine the adjusted correlation rules between the corrected indicators and the collection time window attributes, which easily leads to the misjudgment of data with logical contradictions between blood glucose and insulin as acceptable. These problems result in low data quality control accuracy and poor reliability, directly affecting the accuracy of clinical trial conclusions. To solve this technical problem, we provide a real-time clinical trial data acquisition and quality control management platform. Summary of the Invention

[0004] The purpose of this invention is to provide a real-time acquisition and quality control management platform for clinical trial data to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, a real-time clinical trial data acquisition and quality control management platform is provided, comprising:

[0006] The operation start time recording unit is used to obtain the timestamp of the subject starting the disinfection operation sequence in real time, as the starting reference of the operation sequence;

[0007] The operation sequence delay calculation unit calculates the actual operation delay time based on a preset standard operation sequence time range;

[0008] The transmission delay processing unit is used to monitor the transmission delay time of the device transmitting data to the platform via Bluetooth, and to separate the transmission delay from the operation delay;

[0009] The actual collection time determination unit determines the actual blood collection completion time by subtracting the outputs of the operation sequence delay calculation unit and the transmission delay processing unit from the timestamp uploaded by the device, so as to eliminate the superimposed effects of operation delay and transmission delay.

[0010] The time window dynamic judgment unit compares the output of the actual collection time determination unit with the preset collection time window. When the actual blood collection completion time falls within the time window, it is judged as qualified; otherwise, it is judged as timeout, thus avoiding incorrect rejection or incorrect retention due to the operation start time not being identified.

[0011] The physiological index correction unit stores a correction rule library for different blood collection postures and sites, including the correction coefficients for blood glucose values ​​for sitting and standing postures and the correction differences for insulin values ​​for fingertips and veins, and corrects the original blood glucose and insulin values ​​in real time based on the blood collection posture and site information input by the subject.

[0012] The dynamic correlation logic verification unit performs verification based on the output of the physiological indicator correction unit, applying dynamic correlation logic rules that match the corrected values. This includes the correlation between blood glucose threshold and insulin threshold. The unit also adjusts the applicable conditions of the correlation logic according to the differences in posture and location within the time window to prevent erroneous pass / fail judgments.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0014] This invention separates operation delay and transmission delay using a dual timestamp separation mechanism, determines the actual blood collection completion time using a reverse time restoration algorithm to eliminate time superposition interference, eliminates indicator distortion caused by operation differences using a layered correction architecture, and adjusts the correlation verification rules of blood glucose and insulin by combining the corrected indicators with time window attributes using a three-dimensional decision matrix. Overall, it achieves the effect of accurate calculation of actual blood collection time, reliable physiological indicator data, and no logical contradictions. It effectively solves the problems of low data quality control accuracy and the impact on the accuracy of clinical trial conclusions caused by time delay interference, uncorrected operation differences, and inaccurate logical verification in existing platforms. Attached Figure Description

[0015] Figure 1 This is an overall block diagram of the present invention.

[0016] The meanings of the labels in the diagram are as follows:

[0017] 1. Operation start time recording unit; 2. Operation sequence delay calculation unit; 3. Transmission delay processing unit; 4. Actual acquisition time determination unit; 5. Time window dynamic judgment unit; 6. Physiological index correction unit; 7. Dynamic correlation logic verification unit. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] This invention provides a real-time clinical trial data acquisition and quality control management platform. Please refer to [link / reference]. Figure 1 As shown, it includes:

[0020] Operation start time recording unit 1 is used to obtain the timestamp of the subject starting the disinfection operation sequence in real time, as the starting reference of the operation sequence;

[0021] The operation sequence delay calculation unit 2 calculates the actual operation delay time based on the preset standard time range of the operation sequence;

[0022] The transmission delay processing unit 3 is used to monitor the transmission delay time of the device transmitting data to the platform via Bluetooth, and to separate the transmission delay from the operation delay;

[0023] The actual collection time determination unit 4 determines the actual blood collection completion time by subtracting the outputs of the operation sequence delay calculation unit 2 and the transmission delay processing unit 3 from the timestamp uploaded by the device, so as to eliminate the superimposed effects of operation delay and transmission delay.

[0024] The time window dynamic judgment unit 5 compares the output of the actual collection time determination unit 4 with the preset collection time window. When the actual blood collection completion time falls within the time window, it is judged as qualified; otherwise, it is judged as timeout, thus avoiding incorrect rejection or incorrect retention due to the operation start time not being identified.

[0025] The physiological index correction unit 6 stores a correction rule library for different blood collection postures and sites, including the correction coefficients for blood glucose values ​​for sitting and standing postures and the correction differences for insulin values ​​for fingertips and veins, and performs real-time correction of the original blood glucose and insulin values ​​based on the blood collection posture and site information input by the subject.

[0026] The dynamic correlation logic verification unit 7 verifies the value based on the output of the physiological index correction unit 6 by applying dynamic correlation logic rules that match the corrected value, including the correlation between blood glucose threshold and insulin threshold. It also adjusts the applicable conditions of the correlation logic according to the differences in posture and location within the time window to prevent incorrect qualification judgment.

[0027] Transmission delay processing unit 3 separates transmission delay from operation delay through a dual timestamp separation mechanism:

[0028] When the device starts up, a first operation end timestamp is automatically generated and stored locally. When the platform receives a data packet, a second platform receiving timestamp is generated. The difference between the two is calculated as the pure transmission delay time. At the same time, the transmission delay processing unit 3 interacts with the operation sequence delay calculation unit 2 in real time to obtain the actual delay time data of the operation sequence. Through the collaborative comparison of timestamp difference calculation and operation sequence delay data, the transmission delay component in the mixed delay is independently separated to ensure that the device upload timestamp only retains the operation end time information.

[0029] The actual collection time determination unit 4 uses a reverse time reconstruction algorithm to accurately locate the blood collection completion time.

[0030] Using the device upload timestamp as a reference point, the pure transmission delay time data output by the transmission delay processing unit 3 and the operation delay time data output by the operation sequence delay calculation unit 2 are received synchronously. The actual blood collection completion time point is generated by subtracting the sum of the pure transmission delay time data and the operation delay time data in reverse along the time axis. The operation delay time adopts a sequence accumulation model, which includes the superposition of the measured time consumption of the three stages of disinfection, blood collection operation and device start-up, to ensure that the time restoration result is consistent with the actual operation node of the subject.

[0031] The time window dynamic judgment unit 5 performs qualification verification through a dynamic offset window boundary judgment mechanism:

[0032] After receiving the actual blood collection completion time point output by the actual collection time determination unit 4, it compares it with the start and end boundaries of the preset collection time window in real time. When the time point is after the start boundary and before the end boundary, a qualified mark is triggered and output to the quality control database. When the time point is earlier than the start boundary or later than the end boundary, a timeout alarm is triggered and data upload is frozen. This mechanism eliminates the systematic deviation between the restored time point and the original upload timestamp through dynamic offset compensation of the time window boundary.

[0033] The correction rule base of physiological indicator correction unit 6 contains a multi-dimensional operation parameter mapping table:

[0034] The system stores the numerical conversion relationship between sitting and standing postures for blood glucose values ​​at the blood collection posture dimension, and the numerical conversion relationship between fingertip blood and venous blood values ​​at the blood collection site dimension. The conversion relationship at the posture dimension includes a blood glucose concentration gradient compensation rule caused by differences in systemic circulation, and the conversion relationship at the site dimension includes a plasma separation coefficient compensation rule caused by differences in capillary and venous blood components. The mapping table is generated through training with big data from clinical trials and supports dynamic loading according to research project type.

[0035] The rule base adopts a layered correction architecture:

[0036] The first layer is a gravity effect compensation model based on posture dimension, which automatically matches the vertical displacement correction coefficient of blood glucose value according to the sitting and standing posture input by the subject. The second layer is a plasma osmosis compensation model based on location dimension, which automatically matches the plasma proportion correction coefficient of insulin value according to the input type. The two correction coefficients are superimposed in parallel through a mapping table to generate the final correction value to eliminate the physiological index deviation caused by micro-operational differences.

[0037] The real-time correction operation of the physiological indicator correction unit 6 is executed through a human-machine collaborative triggering mechanism:

[0038] After the subject inputs the blood collection posture and site information through a mobile terminal, the platform automatically collects the raw blood glucose and insulin values ​​uploaded by the device, calls the hierarchical correction architecture, corrects the raw blood glucose value according to the posture dimension, and corrects the raw insulin value according to the site dimension. The corrected index values ​​immediately overwrite the original data storage, and simultaneously mark the source of the correction factor for the dynamic correlation logic verification unit to trace.

[0039] Dynamic association logic verification unit 7 performs verification through the conditional rule engine:

[0040] After receiving the corrected blood glucose and insulin values, the engine activates the corresponding association rule set based on the current time window attributes. The rule set includes logical matching conditions between blood glucose and insulin threshold ranges, and dynamic association threshold functions for blood glucose and insulin. The engine calculates the threshold range to which blood glucose belongs in real time based on the corrected values ​​and retrieves the corresponding insulin threshold range for compliance matching. If the measured insulin value exceeds the association threshold range, it is marked as a logical anomaly.

[0041] The correlation between blood glucose threshold and insulin threshold employs a dynamic interval coupling mechanism:

[0042] For the blood glucose threshold interval division, each sub-interval is bound to an independent insulin threshold range. The lower limit of the insulin threshold range corresponding to the hyperglycemia interval is significantly higher than that of the hypoglycemia interval, and the threshold range boundary value is dynamically adjusted according to the plasma proportion correction coefficient. When the blood glucose value is at the boundary of the adjacent interval, a threshold gradient algorithm is used to smoothly transition the insulin threshold range to avoid misjudgment caused by interval jump.

[0043] The adjustment of the applicable conditions for the association logic is implemented through a three-dimensional decision matrix:

[0044] The first dimension selects a rule library specific to venous blood or fingertip blood based on the type of blood sampling site. The second dimension adjusts the boundary of the blood glucose correlation interval based on the posture correction coefficient. The third dimension combines the judgment time window location to dynamically load the window time period-specific verification threshold. After the matrix outputs the optimal combination of correlation rules, the rule engine performs multi-layer logic verification to eliminate false qualified judgments caused by the interaction effect of posture, site and time window.

[0045] Further explanation is needed regarding the specific implementation of the dual timestamp separation mechanism in the transmission delay processing unit. After the operation start time recording unit 1 obtains the start timestamp of the subject disinfection operation and the operation sequence delay calculation unit 2 preliminarily calculates the actual operation delay, a new transmission delay will be generated when the device transmits blood collection data to the platform via Bluetooth. If this delay is not separated from the operation delay, the superposition of the two will lead to a serious deviation in the subsequent calculation of the actual blood collection completion time. Therefore, the transmission delay processing unit 3 needs to accurately separate the transmission delay from the operation delay through the dual timestamp separation mechanism to provide the actual collection time determination unit 4 with pure transmission delay data. The specific implementation method is as follows:

[0046] The core of the dual timestamp separation mechanism of the transmission delay processing unit 3 is to lock the pure transmission delay by generating and calculating the difference between two key timestamps. At the same time, combined with operation delay data verification, it ensures that the delay splitting is free from cross-interference. First, the device sets the automatic timestamp generation trigger condition. When the subject completes the blood collection operation and the device starts up, that is, when the device has completed the blood collection data and is ready to transmit to the platform, the device system will automatically generate the first operation end timestamp. This timestamp accurately marks the final end time of the operation, without including the subsequent transmission time. After generation, it is immediately stored in the device's local memory, rather than being transmitted with the data packet, to avoid being tampered with or affected by delays during transmission. For example, if the device starts up at "14:25:30.120", then the first operation end timestamp is recorded at this time, ensuring the originality and accuracy of the operation end time.

[0047] When the platform receives the blood collection data packet sent by the device, containing raw blood glucose and insulin data and device identification information, the transmission delay processing unit 3 triggers the platform system to generate a second platform reception timestamp. This timestamp marks the moment the platform successfully receives the data packet. For example, if the platform receives the data packet at "14:25:30.380", the second platform reception timestamp is recorded as this moment. At this time, by calculating the difference between the second platform reception timestamp and the first operation end timestamp, the pure transmission delay time can be obtained. For example, if the difference between the two timestamps is 0.26 seconds, this time is the pure time taken for the data to be transmitted from the device to the platform. Since the first timestamp marks the end of the operation and the second timestamp marks the completion of the transmission, the interval between the two only includes the Bluetooth transmission phase without any operational phase. Therefore, the difference can be directly used as the pure transmission delay time. After calculating the pure transmission delay using only the two timestamps, it is still necessary to eliminate the potential cross-interference between the operation delay and the transmission delay. The data packets uploaded by the device to the platform may include a timestamp containing a mixture of operation and transmission delays. Directly using this mixture of delays will lead to errors in subsequent time restoration. Therefore, the transmission delay processing unit 3 needs to establish a real-time data interaction channel with the operation sequence delay calculation unit 2 to continuously obtain the actual delay of the operation sequence. The time data, namely the total delay of the three stages of disinfection, blood collection, and equipment startup calculated by the operation sequence delay calculation unit 2 based on the preset standard time consumption, is then compared with the operation sequence delay data through timestamp difference calculation. The mixed delay time is compared with the theoretical sum of the operation delay and the pure transmission delay. If the two are consistent, it means that the pure transmission delay calculation is accurate. If there is a deviation, the generation timing of the first operation end timestamp (whether it is indeed generated when the equipment startup is completed) and the accuracy of the second platform receiving timestamp (whether there is a platform receiving buffer delay) are further verified. The deviation is controlled within 0 by fine-tuning the timestamp value. Within 0.5 seconds, the transmission delay component in the mixed delay is finally separated from the mixed delay, and only the operation end time information corresponding to the operation delay is retained. Through the above process, the transmission delay processing unit 3 not only locks the pure transmission delay through dual timestamps, but also verifies the accuracy of delay splitting through collaborative comparison with the operation sequence delay calculation unit 2. Finally, it ensures that the timestamp uploaded by the device to the platform only retains the operation end time information, without any transmission delay superposition. This provides accurate transmission delay data support for the subsequent actual collection time determination unit 4 to reverse restore the real blood collection completion time, avoiding time restoration deviation due to delay confusion.

[0048] The specific implementation of the reverse time restoration algorithm for the actual collection time determination unit is as follows: After the transmission delay processing unit 3 accurately separates the pure transmission delay through the dual timestamp separation mechanism and the operation sequence delay calculation unit 2 initially calculates the operation delay, the timestamp uploaded by the device to the platform still contains the superimposed information of the pure transmission delay time data and the operation delay time data. If it is directly used as the blood collection completion time, it will cause deviation in the subsequent time window determination. Therefore, the actual collection time determination unit 4 needs to trace back the real blood collection completion time from the superimposed delay timestamp based on the reverse time restoration algorithm to achieve accurate positioning. The specific implementation is as follows:

[0049] When the actual collection time determination unit 4 executes the reverse time restoration algorithm, it first uses the device upload timestamp as the core reference point. The device upload timestamp is a time stamp automatically generated and sent to the platform along with the data packet when the device completes blood collection and starts Bluetooth transmission. This timestamp by default includes the entire process delay from the completion of the blood collection operation to the platform receiving the data, that is, the sum of the operation delay and the transmission delay, and is not the actual blood collection completion time. Therefore, it needs to be corrected through reverse restoration. For example, if the device generates and uploads the timestamp at "14:25:31.500", it already implicitly contains the sum of the operation delay and the transmission delay. It needs to be deduced backward from this starting point. Subsequently, the unit synchronously receives the pure transmission delay time data output by the transmission delay processing unit 3 and the operation delay time data output by the operation sequence delay calculation unit 2 through the real-time data interaction interface. The synchronous reception mechanism can ensure that the pure transmission delay time data and the operation delay time data are consistent with the time dimension of the device upload timestamp, avoiding To avoid mismatches in latency values ​​caused by data transmission time differences, for example, if the pure transmission latency is 0.26 seconds and the operation latency is 0.8 seconds, the algorithm ensures that both latency values ​​correspond to the blood collection operation associated with the upload timestamp "14:25:31.500". After obtaining the pure transmission latency data and the operation latency data, the algorithm first calculates the sum of the pure transmission latency data and the operation latency data, and then subtracts this sum backwards along the timeline to generate the actual blood collection completion time. The core logic of the backward subtraction on the timeline is to work backwards from the upload timestamp, successively subtracting the time consumed in the transmission and operation stages, returning to the moment when the blood collection operation actually ends. For example, if the upload timestamp is "14:25:31.500", the sum of the pure transmission latency data and the operation latency data is 0.26 + 0.8 = 1.06 seconds. Subtracting 1.06 seconds backwards from "14:25:31.500" gives the actual blood collection completion time as "14:25:30"."440" marks the moment when the blood collection needle is removed and the blood collection operation is truly completed. The operation delay time is calculated using a sequence accumulation model to ensure a perfect match between the delay value and the actual operation process. This model includes the sum of the measured timestamps for three stages: disinfection, blood collection, and device startup. The first stage, disinfection, is calculated from the disinfection start timestamp obtained by operation start time recording unit 1 to the timestamp corresponding to the disinfection end signal detected by the device sensor when the disinfectant cotton pad leaves the skin. The difference between these two timestamps is the measured disinfection time. The second stage, blood collection, is calculated from the disinfection end signal timestamp to the timestamp corresponding to the blood collection completion sensor signal triggered by the blood collection needle. The difference represents the actual blood collection time. The third stage, device startup time, is calculated from the time stamp of the sensor signal after blood collection completion to the time stamp of the device generating the first operation completion. The difference is the actual device startup time. Adding the actual times of the three stages together yields the operation delay time output by operation sequence delay calculation unit 2. Through this process, combined with the sequence accumulation model for the measured decomposition of operation delay, the reverse time restoration algorithm completely eliminates the superposition interference of operation delay and transmission delay, ensuring that the generated actual blood collection completion time is completely consistent with the subject's actual operation node. This avoids time deviations caused by delay superposition and provides accurate time basis for the subsequent qualification verification of time window dynamic judgment unit 5.

[0050] The specific implementation of the dynamic offset window boundary determination mechanism of the time window dynamic determination unit: After the actual collection time determination unit 4 obtains the actual blood collection completion time point after removing operation delay and transmission delay through the reverse time restoration algorithm, if a fixed collection time window boundary is directly used for qualification verification, it is easy to misjudge due to the systematic deviation between the restored time point and the original upload timestamp. It may misjudge a time point that should be qualified as a timeout, or retain abnormal data that actually timed out. Therefore, the time window dynamic determination unit 5 needs to use a dynamic offset window boundary determination mechanism to compensate for the deviation while accurately comparing the time points, ensuring that the qualification verification results are completely matched with the requirements of the clinical trial protocol. The specific implementation is as follows:

[0051] When the time window dynamic determination unit 5 performs the eligibility verification, it first receives the actual blood collection completion time point output by the actual collection time determination unit 4. This time point is the true operation node time after all delay disturbances have been eliminated. For example, "14:25:30.440" represents the instant when the subject's blood collection needle is pulled out and the blood collection operation is truly completed. The unit will synchronously record the subject identification, blood collection device number, and clinical trial project number corresponding to this time point to ensure that the subsequent verification results can be accurately associated with specific operation links and avoid data confusion. Subsequently, the unit retrieves the preset collection time window for this clinical trial project. The preset collection time window is a time range set according to the regulations on blood collection time in the clinical trial protocol, including the start boundary and the end boundary. For example, if the administration time is "14:25:00.000", the start boundary of the initial preset collection time window is "14:25:28.000" and the end boundary is "14:25:32.000". In theory, all actual blood collection completion time points need to fall within this interval to be determined as qualified. However, due to possible systematic deviations in the restored time points, if fixed boundaries are directly used, time points such as "14:25:27.950" that should be qualified but are too early due to deviations may be misjudged as overtime. Therefore, it is necessary to start the dynamic offset window boundary compensation. The core of the dynamic offset compensation is that the unit calculates and adjusts the start boundary and the end boundary of the time window based on historical verification data and real-time deviation monitoring. The unit will automatically retrieve the deviation records corresponding to this blood collection device within the last 24 hours, calculate the average trend of the deviations, and use this average deviation value as the dynamic offset amount. If the deviation is positive, the boundary will be shifted backward; if it is negative, the boundary will be shifted forward. For example, in the scenario where the average is 0.1 seconds earlier as above, the start boundary will be shifted forward 0.1 seconds from "14:25:28.000" to "14:25:27.900", and the end boundary will be shifted forward 0.1 seconds from "14:25:32.000" to "14:25:31.900". By shifting the boundary, the systematic deviation is fully compensated to ensure that the verification standard adapts to the actual deviation situation of the restored time point. After completing the dynamic offset of the boundary, the unit compares the actual blood collection completion time point with the adjusted start boundary and end boundary in real time. If this time point (such as "14:25:30.440") is after the adjusted start boundary ("14:25:27.900") and before the adjusted end boundary ("14:25:31.900"), it immediately triggers the generation of a qualified mark. The qualified mark includes the "time qualified" identification, the actual blood collection completion time point, the adjusted time window boundary, and the dynamic offset amount, which are synchronously output to the quality control database for archiving, and the blood collection data is allowed to enter the subsequent physiological index correction link. If this time point is earlier than the adjusted start boundary (such as "....If the timeout value is "000", a timeout alarm is immediately triggered. The alarm information is pushed to the clinical trial monitor's terminal device in real time, indicating "Subject XX completed blood collection at 14:25:32.000, exceeding the adjusted time window (14:25:27.900-14:25:31.900), data abnormal." Simultaneously, the upload process for this blood collection data is automatically frozen, preventing it from proceeding to subsequent stages until the monitor verifies the cause of the deviation and manually confirms whether to retain the data. This dynamic offset window boundary determination mechanism strictly adheres to the time requirements of the clinical trial protocol and eliminates the systematic deviation between the restored time point and the original upload timestamp through deviation compensation. This completely avoids erroneous rejection or retention due to inconsistent time bases, providing precise assurance for the time dimension quality control of clinical trial data.

[0052] The specific implementation of the physiological indicator correction unit's rule base and multidimensional operation parameter mapping table: After the time window dynamic judgment unit 5 completes the qualification verification of the actual blood collection time, although the verified blood collection data meets the time dimension requirements, the original blood glucose and insulin values ​​are still affected by the micro-operational differences in blood collection posture and blood collection site. For example, when standing, the blood return speed of the lower limbs is slowed down due to gravity, which may lead to a deviation in blood glucose concentration compared to when sitting. The different proportions of plasma components in fingertip capillary blood and venous blood will directly affect the insulin detection results. If the original data is used directly, it will lead to the distortion of physiological indicators. Therefore, the physiological indicator correction unit 6 needs to rely on the correction rule base containing the multidimensional operation parameter mapping table to achieve accurate indicator correction. The specific implementation is as follows:

[0053] The core carrier of the correction rule base of physiological indicator correction unit 6 is a multi-dimensional operational parameter mapping table. This mapping table uses operational dimensions, indicator types, and numerical conversion relationships as its core framework. It constructs data mapping logic for key operational variables affecting physiological indicators, ensuring that raw values ​​under different operational scenarios can be converted into standardized correction values. The first dimension is the blood sampling posture dimension, specifically storing the numerical conversion relationship between sitting and standing postures on blood glucose values. Because there are significant differences in systemic circulation between sitting and standing postures (i.e., "systemic circulation difference"), gravity increases the resistance to venous blood return in the lower limbs when standing, prolonging the time blood remains in peripheral vessels. In contrast, blood return is more balanced when sitting. This difference can lead to blood collection difficulties for the same subject under different postures. Since there are gradient differences in blood glucose concentration, this dimension incorporates a glucose concentration gradient compensation rule caused by differences in systemic circulation. By pre-recording the difference in blood glucose levels at the same time point and location under different postures, this difference is converted into a compensation coefficient. The coefficient is then optimized according to the age and weight of the subjects to ensure the compensation rule is applicable to people with different physiological characteristics, avoiding a "one-size-fits-all" compensation bias. The second dimension is the blood sampling site dimension, focusing on storing the numerical conversion relationship between finger-prick blood and venous blood for insulin values. Finger-prick blood is capillary blood, while venous blood is venous circulating blood. Due to differences in blood source and component filtration pathway, there are significant component differences between the two. Capillary blood contains more tissue fluid return components, while venous blood has passed through the liver. The metabolic regulation of organs such as capillaries and venous blood can lead to deviations in insulin test values. Therefore, it is necessary to introduce a compensation rule for the plasma separation coefficient caused by the difference in composition between capillary and venous blood. The plasma separation coefficient is a parameter that quantifies the ratio of the actual effective concentration of insulin in capillary blood to that in venous blood. By analyzing the plasma separation process of the two in the laboratory, the plasma separation coefficient under different conditions is determined, and this coefficient is stored in association with the blood collection site and blood collection time (fasting / postprandial). This ensures that the corresponding conversion relationship is invoked in different scenarios. For example, the original value of insulin from fingertip blood in fasting condition needs to be corrected according to the coefficient, while in postprandial condition it is corrected according to the adjusted coefficient, accurately eliminating the index deviation caused by site differences. The generation of this multidimensional operational parameter mapping table does not rely on theoretical derivation, but rather on large-scale... The system employs large-scale clinical trial big data training to ensure the authenticity and universality of the transformation relationships. Data sources encompass multi-center clinical trials, multi-sample populations, and multiple operational scenarios. By collecting blood glucose and insulin test data from the same subject at the same time point under different operational conditions, a corresponding dataset is established. Statistical regression analysis is then used to fit the numerical transformation relationships under different operational conditions, ultimately forming a structured, multi-dimensional operational parameter mapping table. New clinical trial data is incorporated quarterly to iteratively optimize the mapping table, ensuring its adaptation to new testing equipment or population characteristics. Furthermore, the mapping table supports dynamic loading according to research project type to accommodate the specific needs of different clinical trials, such as research projects on "blood glucose control trials in patients with type 2 diabetes."Because many participants have insulin resistance, the difference in blood glucose concentration gradient between sitting and standing postures may be more significant than in healthy individuals. Therefore, the compensation coefficient for the posture dimension in the loaded mapping table is optimized for diabetic patients. For the "Childhood Metabolic Syndrome Screening Trial," because children's capillaries are thinner, the insulin composition of fingertip blood and venous blood differs from that of adults. Therefore, the plasma separation coefficient for the site dimension in the loaded mapping table is adjusted to parameters specific to children. The platform automatically matches and loads the corresponding mapping table version by identifying the trial protocol number of the research project, eliminating the need for manual adjustments. This improves correction efficiency and avoids correction errors caused by inappropriate project adaptation, providing a standardized physiological indicator data foundation for the subsequent dynamic correlation logic verification unit 7.

[0054] The specific implementation of the hierarchical correction architecture of the physiological indicator correction unit's correction rule base: After the multi-dimensional operation parameter mapping table provides the data foundation for indicator correction under different operation scenarios, the correction rule base of the physiological indicator correction unit 6 needs to use a hierarchical correction architecture to decompose the compensation logic in the mapping table into targeted and collaborative step-by-step correction processes. This architecture accurately corrects blood glucose and insulin values ​​according to their corresponding relationships, avoiding mutual interference between the correction logics of different operation variables, and ensuring that the final correction value can completely eliminate micro-operational differences. The specific implementation is as follows:

[0055] The first layer of the hierarchical correction architecture is a gravity effect compensation model based on posture. Its core function is to eliminate the interference of blood sampling posture (sitting / standing) on ​​blood glucose values. The gravity effect refers to the influence of gravity on the distribution of blood in the systemic circulation when the human body is in different postures. In the standing posture, gravity increases the resistance to venous blood return in the lower limbs, resulting in a longer residence time of blood in the peripheral blood vessels, and subtle changes in the metabolism and distribution of glucose in the blood. In the sitting posture, the blood return throughout the body is more balanced, and the blood glucose concentration is closer to the actual circulating level of the human body. This difference will cause deviations in the original blood glucose values ​​collected from the same subject in different postures. Therefore, this model is needed to compensate for this. When the model is working, it first receives the blood sampling data input by the subject through a mobile terminal. Posture information, which is uploaded to the unit along with blood sampling data, is used by the model to automatically retrieve a preset vertical displacement correction coefficient from the multidimensional operation parameter mapping table that matches the posture. This vertical displacement correction coefficient is a parameter derived from extensive clinical trial data, quantifying the differences in blood glucose concentration gradients under different postures. The coefficient is also fine-tuned based on the subject's basic physiological characteristics to ensure compensation is more tailored to individual circumstances. Subsequently, the model applies this vertical displacement correction coefficient to the original blood glucose value to complete the posture dimension correction. For example, if a subject's original blood glucose value is a certain value obtained from standing blood sampling, the corresponding correction coefficient is added to obtain an intermediate corrected blood glucose value that eliminates posture interference. After completing the posture dimension blood glucose value correction, the correction rule base enters the second layer, based on... This site-based plasma osmotic compensation model specifically addresses the impact of blood collection site (finger-tip / venous) on insulin levels. The core of plasma osmotic compensation is to resolve the difference in insulin composition between capillary blood (finger-tip blood) and venous blood. Finger-tip blood originates from capillaries and contains more components returning from tissue fluid, resulting in a relatively lower free insulin concentration and plasma percentage. Venous blood, on the other hand, undergoes metabolic regulation by the liver and kidneys, leading to a more stable plasma insulin percentage and a better reflection of the body's overall insulin secretion and regulation. This difference in plasma composition can cause significant deviations in the raw insulin values ​​of the same subject collected from different sites. Therefore, this model is used to compensate for this discrepancy. After the model is activated, it first reads the blood collection site type input by the subject. Next, based on this type, the corresponding plasma percentage correction coefficient is matched from the multidimensional operational parameter mapping table. The plasma percentage correction coefficient is obtained through laboratory centrifugation and concentration detection, quantifying the proportion of the actual effective insulin concentration in different blood types. For example, the fasting finger-prick insulin value needs to be multiplied by a certain coefficient to convert it to a standard consistent with the venous insulin value. In the postprandial state, the coefficient will be adjusted according to metabolic changes. The model superimposes this plasma percentage correction coefficient with the original insulin value to obtain an intermediate corrected value of insulin that eliminates site interference. The two correction models do not act sequentially on the same indicator, but rather use a parallel superposition method to specifically correct blood glucose and insulin values ​​separately, and then integrate them to generate the final corrected value.The posture-based gravity effect compensation model only applies to the original blood glucose value, outputting an intermediate blood glucose correction value. The site-based plasma osmosis compensation model only applies to the original insulin value, outputting an intermediate insulin correction value. Since the deviation in blood glucose values ​​mainly stems from posture, and the deviation in insulin values ​​mainly stems from site, this parallel logic, each responsible for its own area, avoids cross-interference in correction logic. Ultimately, the intermediate blood glucose correction value and the intermediate insulin correction value together constitute the final correction value, synchronously covering the original data storage. The source of the coefficients used in each layer of correction is marked, allowing the subsequent dynamic correlation logic verification unit 7 to trace the correction process and ensure the traceability and accuracy of the correction. Through this hierarchical architecture, the physiological indicator correction unit 6 can accurately eliminate the microscopic operational differences caused by blood collection posture and site, making the corrected physiological indicators more closely reflect the actual physiological state of the human body, providing a standardized and high-quality data foundation for subsequent dynamic correlation logic verification.

[0056] The specific implementation of the real-time correction operation human-machine collaborative triggering mechanism of the physiological indicator correction unit: After clarifying the correction logic in the hierarchical correction architecture of the correction rule base, the physiological indicator correction unit 6 needs to seamlessly connect the manually input operation information of the subject with the automatic execution of the correction process by the platform through the human-machine collaborative triggering mechanism. This mechanism ensures the authenticity of key operation information such as blood collection posture and site, which is directly input by the subject to avoid the bias of third-party recording, and improves the correction efficiency through platform automation processing to avoid errors caused by manual calculation. The specific implementation is as follows:

[0057] The human-machine collaboration trigger mechanism is initiated when the subject completes the blood collection procedure and inputs key information via a mobile terminal. The mobile terminal is a dedicated device linked to the clinical trial data platform, pre-installed with a dedicated clinical trial app. The app includes a blood collection information input interface with a simple design and operation prompts to prevent subject misoperation. After completing the blood collection, the subject must log in to the app within 5 minutes, select the blood collection posture and site in the designated input module, and click the "Submit" button after confirming the information is correct. The input posture and site information, along with the subject's unique identifier, blood collection device number, and operation timestamp, is simultaneously uploaded to the physiological indicator correction unit 6, ensuring that each piece of information is accurately linked to the corresponding blood collection operation and avoiding misoperation by different subjects. In case of confusion regarding information from different blood collection attempts or different participants, the platform immediately initiates an automated data collection and correction process after the participant completes and submits their information. First, the physiological indicator correction unit (6) automatically collects the raw blood glucose and insulin values ​​uploaded by the blood collection device via a Bluetooth data interface. This collection process is synchronized in real-time with the participant's information upload. The platform uses a dual matching mechanism to bind the newly input posture / site information with the corresponding raw indicator values, preventing information and data mismatch. Subsequently, the unit automatically invokes the constructed hierarchical correction architecture and performs corrections according to the corresponding relationships. For the raw blood glucose value, the architecture initiates the first-layer posture dimension gravity effect compensation model, matching the corresponding values ​​from the multi-dimensional operation parameter mapping table based on the bound blood collection posture. A vertical displacement correction coefficient is used, which is then superimposed on the original blood glucose value to calculate the corrected blood glucose value to eliminate posture interference. For the original insulin value, a second-layer plasma osmosis compensation model at the site dimension is initiated. Based on the bound blood collection site, the corresponding plasma percentage correction coefficient is matched from the mapping table, and this coefficient is superimposed on the original insulin value to calculate the corrected insulin value to eliminate site interference. The entire correction process is executed automatically by the platform without human intervention. The time from data acquisition to correction completion is controlled within 10 seconds, ensuring that the real-time quality control requirements of clinical trial data are met. After correction, the platform synchronously processes data storage and correction traceability information. On the one hand, the corrected blood glucose and insulin values ​​immediately overwrite the original data in the platform database. The storage records are managed using versioning, meaning the original data is not deleted but archived as historical versions. The corrected data is stored as the current valid version at the top. Subsequent dynamic correlation logic verification unit 7 and the quality control database both use the corrected data, preventing the original data from interfering with subsequent quality control processes. Furthermore, the unit automatically generates correction factor source markers, including posture correction coefficients and their sources, site correction coefficients and their sources, correction execution timestamps, and associated subject identifiers and device numbers. These markers are bound to the corrected data and stored under the same data entry. When the dynamic correlation logic verification unit 7 performs logical verification on the corrected indicator values, it can directly retrieve these markers to trace the correction process.To ensure the traceability and verifiability of the correction operations and comply with the requirements of clinical trial data management, the physiological indicator correction unit 6, through a human-machine collaborative process, guarantees both the accuracy of operational information and the efficiency of the correction process. Furthermore, by using traceability markers, it provides data credibility support for subsequent quality control steps, completely eliminating the interference of microscopic operational differences such as blood collection posture and site on physiological indicators. This lays a standardized data foundation for the dynamic correlation logic verification unit 7 to conduct precise logic verification.

[0058] The specific implementation method of the dynamic correlation logic verification unit's conditional rule engine for verification is as follows: After the physiological indicator correction unit 6 completes the real-time correction of blood glucose and insulin values ​​through the human-machine collaborative triggering mechanism, although the corrected indicators eliminate the interference of operational differences caused by blood collection posture and site, there may still be logical contradictions between indicators. For example, under normal physiological conditions, when blood glucose rises within a specific time window after drug administration, insulin secretion should increase accordingly. If there is a situation where blood glucose is significantly high but insulin is low, it may be due to data collection errors or sudden physiological abnormalities in the subject. If it is directly included in the analysis, it will affect the accuracy of the clinical trial conclusions. Therefore, the dynamic correlation logic verification unit 7 needs to perform precise verification based on the physiological correlation law between blood glucose and insulin through the conditional rule engine to ensure data logic consistency. The specific implementation method is as follows:

[0059] The conditional rule engine of the dynamic association logic verification unit 7 takes "data reception - rule activation - interval judgment - compliance matching - anomaly marking" as its core process. First, it receives the corrected blood glucose and insulin values ​​output by the physiological indicator correction unit 6 through the real-time data interface. At the same time, it retrieves the key attribute information associated with this set of data, including the current time window attribute output by the time window dynamic judgment unit 5, that is, the preset time stage of the clinical trial to which the actual blood collection completion time point belongs. The correlation between blood glucose and insulin metabolism differs in different time stages. The basic information of the subjects ensures that the verification rules can be adapted to the specific trial scenario and avoids misjudgment due to general rules not being suitable for special cases.

[0060] After acquiring key attribute information, the conditional rule engine activates the corresponding association rule set based on the current time window attribute. The association rule set is a structured rule set built based on a large amount of clinical trial metabolic data. Each time window attribute corresponds to a unique rule set. For example, the blood glucose threshold range and insulin threshold range of the rule set 30 minutes after drug administration are completely different from those of the rule set 60 minutes after drug administration. This is because blood glucose is mostly in the rising phase and insulin secretion is in the initiation phase 30 minutes after drug administration. Blood glucose may reach its peak at 60 minutes, and insulin secretion will also reach its peak accordingly. The rule set contains two types of core rules. One is the logical matching condition between the blood glucose threshold range and the insulin threshold range, such as "when the blood glucose value is in range A, the insulin value must be in range a.""When blood glucose levels are in range B, insulin levels must also be within range b." This range division is based on two principles: first, the distribution range of indicators in 95% of normal samples within that time window; and second, a dynamic correlation threshold function between blood glucose and insulin. This function is not a fixed formula but rather a logic that dynamically adjusts the insulin threshold range based on time window attributes and the subject's baseline information. For example, for diabetic patients, within the "30-minute post-drug administration" time window, if the blood glucose level is higher than the average for healthy individuals, the corresponding lower limit of the insulin threshold range will be lower than for healthy individuals, because insulin secretion in diabetic patients may be delayed. These two rules work together to ensure the scenario adaptability and physiological rationality of the correlation verification. After the rule set is activated, the conditional rule engine first calculates the threshold range to which the corrected blood glucose value belongs in real time. The engine traverses the blood glucose threshold range divisions in the current rule set. For example, the "30 minutes after drug administration" rule set divides blood glucose into "low range - normal range - high range". By comparing the corrected blood glucose value with the boundary values ​​of each range, the engine determines the specific range to which it belongs. Subsequently, based on the blood glucose threshold range, the engine retrieves the corresponding insulin threshold range from the rule set and initiates compliance matching, that is, it compares the corrected insulin value with the retrieved insulin threshold range in real time. If the corrected insulin value falls within the range, the group is judged. If the data logic is compliant, the engine automatically generates a logic compliance marker and synchronously outputs the data and marker to the quality control database, allowing it to proceed to the subsequent clinical trial data analysis stage. If the corrected insulin value exceeds the associated threshold range, it is judged as a logic anomaly. The engine immediately generates a logic anomaly marker, the marker content of which includes "anomaly type (blood glucose-insulin correlation contradiction)," "current blood glucose range," "corresponding insulin threshold range," "measured insulin value," and "associated time window attribute." Simultaneously, the anomaly information is pushed to the clinical trial monitor's terminal device in real time, and the subsequent flow of this group of data is frozen until the monitor verifies the cause of the anomaly. Data freezing is lifted only after verification results are manually entered on the platform, ensuring that every logically abnormal data point can be traced and processed. Through a conditional verification logic of "scenario-based rule activation – precise interval determination – dynamic compliance matching," the rule engine not only follows the physiological metabolic patterns of blood glucose and insulin under different time windows but also adapts to individual differences among subjects. This effectively avoids misjudgments caused by rigid, one-size-fits-all rules. Simultaneously, through anomaly marking and flow freezing mechanisms, strict quality control of the logical dimensions of clinical trial data is provided, ensuring that the data ultimately entering the analysis stage eliminates operational differences and possesses logical consistency.

[0061] After the conditional rule engine activates the association rule set and clarifies that the insulin threshold range needs to be matched through the blood glucose threshold range, if a fixed interval binding relationship is used, two types of problems are likely to occur. First, it cannot adapt to the difference in insulin value caused by the plasma proportion correction coefficient in the physiological indicator correction unit 6. Second, when the blood glucose value is at the boundary of an adjacent interval, the sudden jump in the insulin threshold can easily lead to misjudgment of the critical value. Therefore, the dynamic association logic verification unit 7 achieves flexible adaptation and smooth connection between blood glucose and insulin thresholds through a dynamic interval coupling mechanism. The specific implementation method is as follows:

[0062] The dynamic interval coupling mechanism first refines the blood glucose threshold interval based on the current time window attributes and the subject's basic information. The blood glucose threshold interval is not a universally fixed range, but rather a set of sub-intervals derived from a large amount of normal sample data, combining clinical trial protocols and physiological metabolic patterns within the time window. It is typically divided into three core sub-intervals: "hypoglycemic interval," "normal blood glucose interval," and "hyperglycemic interval." In some refined scenarios, this is further subdivided. The boundary values ​​of each sub-interval are clearly labeled with physiological significance to ensure that the interval division conforms to real physiological logic. After completing the blood glucose threshold interval division, the mechanism binds an independent insulin threshold range to each sub-interval. An independent insulin threshold range refers to a separately set normal insulin secretion range for different blood glucose sub-intervals. Its binding is based on the physiological feedback relationship between blood glucose levels and insulin secretion. When blood glucose is in the hyperglycemic interval, the pancreatic β cells secrete more insulin to lower blood glucose. Therefore, the lower limit of the insulin threshold range corresponding to this interval must be significantly higher than that of the hypoglycemic interval. For example, within a 30-minute time window after drug administration, the lower limit of the insulin threshold range corresponding to the hyperglycemic interval is a certain value, while the lower limit of the insulin threshold range corresponding to the hypoglycemic interval is... The upper limit of the insulin threshold range is only half of this value to avoid the logical contradiction of high blood sugar corresponding to low insulin. At the same time, the upper and lower limits of each insulin threshold range are derived from the insulin test values ​​of normal subjects within that blood sugar sub-range, ensuring that the binding relationship conforms to the real physiological feedback law, rather than the theoretical derivation value. In order to adapt to the difference in insulin value caused by the plasma proportion correction coefficient in physiological indicator correction unit 6, the dynamic interval coupling mechanism will adjust the boundary value of the insulin threshold range in real time according to the plasma proportion correction coefficient. The plasma proportion correction coefficient is the core parameter used in physiological indicator correction unit 6 to correct the insulin value of finger-prick blood and venous blood. Since the baseline of the corrected insulin value has changed, using the original boundary value will lead to matching deviation. Therefore, the mechanism will automatically retrieve the plasma proportion correction coefficient corresponding to the current blood collection site and adjust the upper and lower limits of the insulin threshold range according to the coefficient ratio. For example, the plasma proportion correction coefficient corresponding to venous blood is 1 (no correction is needed), and its insulin threshold range is AB. If the plasma proportion correction coefficient of finger-prick blood is 0.9 (the corrected insulin value is 0.9 times the original value), then the insulin threshold range will be adjusted to 0.9A-0.9B ensures that the adjusted threshold range perfectly matches the corrected insulin baseline, avoiding verification deviations caused by differences in blood sampling sites. When blood glucose values ​​are near the boundaries of two adjacent sub-intervals, the mechanism uses a threshold gradient algorithm to achieve a smooth transition of the insulin threshold range, avoiding misjudgments caused by interval jumps. The core logic of the threshold gradient algorithm is to set a gradual transition zone centered on the blood glucose interval boundary. Within the transition zone, the insulin threshold range gradually transitions linearly from the upper limit of the lower interval to the lower limit of the higher interval, rather than abruptly jumping. For example, the upper limit of the insulin threshold corresponding to the upper limit of the "normal blood glucose interval" is C, and the lower limit of the insulin threshold corresponding to the lower limit of the "hyperglycemic interval" is D (D is greater than C). When the blood glucose value is within the transition zone from the upper limit of the "normal interval" to... When the "lower limit of the high blood sugar range" moves, the effective range of the insulin threshold gradually transitions from "less than or equal to C" to "greater than or equal to D." Blood sugar values ​​in the middle correspond to the "gradual threshold range between C and D." This transition method avoids situations where blood sugar values ​​only slightly fluctuate across the range boundary, but insulin values ​​that haven't reached the new range threshold are misjudged as abnormal. This ensures that the verification results more closely match the continuous changes in physiological indicators. The dynamic range coupling mechanism achieves scenario adaptability to the correlation between blood sugar and insulin thresholds and solves the problem of misjudgment of critical values. This makes the logic verification of the conditional rule engine more accurate and more consistent with the real laws of human physiological metabolism in clinical trials, providing core support for the reliable logical compliance results output by the dynamic correlation logic verification unit 7.

[0063] The specific implementation of the three-dimensional decision matrix for adjusting the applicable conditions of the association logic: After the dynamic interval coupling mechanism achieves flexible adaptation of blood glucose and insulin thresholds, the dynamic association logic verification unit 7 still needs to solve the false pass / fail problem caused by the interaction effect of blood collection posture, site, and time window. For example, the blood glucose-insulin association pattern of the same subject in the scenario of "finger prick blood + standing posture + 30 minutes after drug administration" is significantly different from that in the scenario of "venous blood + sitting posture + 60 minutes after drug administration". If a single association rule is used, abnormal data is easily misjudged as pass / fail due to scenario interaction deviation. Therefore, the unit integrates three types of key variables through a three-dimensional decision matrix to accurately adjust the applicable conditions of the association logic, ensuring that the verification rule is fully matched with the actual operation scenario. The specific implementation method is as follows:

[0064] The core of the three-dimensional decision matrix is ​​to construct a multi-dimensional correlation logic adjustment framework based on blood sampling site, posture correction coefficient, and time window position. Each dimension relies on accurate data output by the unit to avoid subjective setting bias. The first dimension selects either a venous blood-specific or fingertip blood-specific correlation rule library based on the blood sampling site type. The blood sampling site type is obtained by the physiological indicator correction unit 6 through a human-machine collaborative triggering mechanism. The difference between the two types of specific rule libraries stems from the fundamentally different impacts of blood sampling sites on insulin values. In the venous blood-specific rule library, the correlation threshold between blood glucose and insulin is constructed based on data corrected from venous blood. Plasma components are stable, and insulin values ​​better reflect the overall metabolic state. The insulin threshold range in the rules is more broadly defined. The fingertip blood-specific rule library is designed specifically for the characteristics of fingertip blood after plasma osmotic compensation. Because fingertip blood insulin values ​​are more affected by tissue fluid reflux, the fluctuation tolerance of the insulin threshold range in the rules is stricter. Furthermore, the correlation logic adds a check for the synchronous change trend of insulin and blood glucose values ​​to avoid misjudgments caused by site characteristics. The first dimension of the matrix automatically retrieves the corresponding rule library as the verification basis through a one-to-one mapping between site type and rule library, ensuring the accuracy of the correlation. The second dimension of the logic is to adjust the boundary of the blood glucose correlation interval based on the posture correction coefficient. The posture correction coefficient is the core output parameter of the gravity effect compensation model in physiological indicator correction unit 6. This coefficient directly reflects the degree of influence of blood sampling posture on blood glucose value. Since posture differences can cause blood glucose benchmark values ​​to shift, if a fixed blood glucose correlation interval is still used, it is easy to have a situation where the corrected blood glucose value is at the boundary of the interval but is misjudged due to posture. For example, when sitting, the boundary of the "normal blood glucose interval" is a certain value. If the corrected blood glucose value in standing posture is close to this boundary, the actual blood glucose value before correction may be misjudged. The blood sugar level may have exceeded the normal range. Therefore, the second dimension of the matrix will adjust the upper and lower boundaries of the blood sugar correlation interval proportionally according to the magnitude of the posture correction coefficient. If the correction coefficient is greater than 1 (e.g., 1.05 for standing posture), it means that the original blood sugar value is too high due to posture, and the blood sugar interval boundary needs to be enlarged synchronously (e.g., the original boundary is multiplied by 1.05). If the correction coefficient is less than 1 (e.g., 0.98 for special sitting posture), the boundary will be reduced synchronously (e.g., the original boundary is multiplied by 0.98) to ensure that the adjusted blood sugar interval can accurately cover the normal blood sugar range under this posture and avoid posture deviation from interfering with the applicable conditions of the correlation logic.The third dimension dynamically loads a specific verification threshold for the time period based on the time window position of the judgment. The time window position is output by the dynamic judgment unit 5, which is the preset time period of the clinical trial to which the actual blood collection completion time belongs. The metabolic state of the human body varies significantly at different time periods, directly affecting the correlation between blood glucose and insulin. In the early stage of drug administration (15-30 minutes), the drug has not yet fully taken effect, blood glucose is mostly in the rising phase, and insulin secretion has just started. At this time, the correlation verification needs to focus more on whether insulin starts as blood glucose rises. In the middle stage of drug administration (30-60 minutes), the drug takes effect, blood glucose reaches its peak, and insulin secretion also reaches its peak. The verification needs to focus on whether the insulin peak matches the blood glucose peak. In the late stage of drug administration (60-90 minutes), blood glucose gradually decreases, and insulin secretion decreases accordingly. The verification needs to focus on whether the downward trends of the two are synchronized. Therefore, the third dimension of the matrix will load the specific verification threshold for the corresponding time period from the preset threshold library according to the time window position to ensure the time adaptability of the correlation logic. When the parameters of the three dimensions are all input into the three-dimensional decision matrix, the matrix outputs the most... The optimal association rule combination, for example, inputting "finger-prick blood + posture correction coefficient 1.05 + 30-60 minutes after administration", will first match the finger-prick blood-specific rule library, then multiply the blood glucose interval boundary in the rule library by 1.05, and finally load the peak matching threshold for 30-60 minutes after administration to form the optimal combination. Subsequently, the conditional rule engine calls this combination to perform multi-layer logic verification. The first layer verifies the basic association between insulin and blood glucose values ​​according to the finger-prick blood rule library. The second layer determines the blood glucose range according to the adjusted blood glucose interval. The third layer verifies whether the peak values ​​of the two match using the mid-term peak threshold. The three-layer verification is progressive, completely eliminating false pass / fail caused by interaction. Through the precise adjustment of the applicable conditions of the association logic by the three-dimensional decision matrix, the dynamic association logic verification unit 7 achieves comprehensive coverage of the multi-variable interaction effects of the blood collection scenario, ensuring that each set of corrected physiological indicators can be verified under the rules adapted to its own operating scenario. This further improves the accuracy and reliability of the logical quality control of clinical trial data, providing a key guarantee for the final output of high-quality clinical trial data.

[0065] In this invention, the operation start time recording unit 1 acquires the start timestamp of the subject's disinfection operation; the operation sequence delay calculation unit 2 calculates the actual operation delay based on the standard time consumption; the transmission delay processing unit 3 separates the transmission and operation delays using dual timestamps; the actual collection time determination unit 4 uses a reverse time restoration algorithm to subtract the pure transmission delay time data and the operation delay time data from the device upload timestamp to obtain the true blood collection completion time; the time window dynamic judgment unit 5 compares the true time with the preset time window through dynamic offset boundaries to determine whether it is qualified; the physiological indicator correction unit 6 corrects blood glucose and insulin values ​​in real time using a hierarchical architecture and correction rule library; and the dynamic association logic verification unit 7 uses a conditional rule engine combined with a three-dimensional decision matrix to adjust the association rules and verify the indicator logic to ensure data reliability and provide support for clinical trials.

[0066] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A clinical trial data real-time acquisition and quality control management platform, characterized in that, Comprise: Operation start time recording unit (1) for real-time acquisition of the time stamp of the subject starting the disinfection operation sequence as the starting reference of the operation sequence; Operation sequence delay calculation unit (2) calculates the actual operation delay time based on the preset operation sequence standard time range; Transmission delay processing unit (3) is used for monitoring the transmission delay time of the device transmitting data to the platform through Bluetooth, and separating the transmission delay from the operation delay; Actual collection time determination unit (4) determines the actual blood collection completion time point by subtracting the output of operation sequence delay calculation unit (2) and transmission delay processing unit (3) from the time stamp uploaded from the device, so as to eliminate the superimposed effect of operation delay and transmission delay; Time window dynamic judgment unit (5) compares the output of actual collection time determination unit (4) with the preset collection time window, and judges as qualified when the actual blood collection completion time point falls within the time window, otherwise judges as overtime, so as to avoid false rejection or false retention caused by unrecognized operation start time; Physiological index correction unit (6) stores correction rule library for different blood collection posture and position, including correction coefficient of blood glucose value for sitting and standing posture and correction difference of insulin value for fingertip and vein, and corrects the original blood glucose and insulin value in real time based on the blood collection posture and position information input by the subject; The correction rule library of physiological index correction unit (6) contains multi-dimensional operation parameter mapping table: For blood collection posture dimension, store the numerical conversion relationship of blood glucose value for sitting and standing posture, and for blood collection position dimension, store the numerical conversion relationship of insulin value for fingertip blood and vein blood, wherein the conversion relationship of posture dimension contains blood glucose concentration gradient compensation rule caused by body circulation difference, and the conversion relationship of position dimension contains plasma separation coefficient compensation rule caused by capillary and vein blood composition difference, the mapping table is generated by clinical trial big data training, and supports dynamic loading according to research project type; The correction rule library adopts hierarchical correction architecture: The first layer is gravity effect compensation model based on posture dimension, which automatically matches the vertical displacement correction coefficient of blood glucose value according to the sitting and standing state input by the subject, and the second layer is plasma permeation compensation model based on position dimension, which automatically matches the plasma proportion correction coefficient of insulin value according to the input type, the two layers of correction coefficient are connected in parallel and superimposed through the mapping table to generate the final correction value to eliminate the physiological index deviation caused by micro operation difference; Dynamic association logic verification unit (7) verifies the output of physiological index correction unit (6) by applying dynamic association logic rule matched with the corrected value, including the association relationship of blood glucose threshold and insulin threshold, and adjusts the applicable conditions of association logic according to the posture and position difference in time window, so as to prevent false qualified judgment; Dynamic association logic verification unit (7) verifies through conditional rule engine: After receiving the corrected blood glucose value and insulin value, the corresponding association rule set is activated according to the current time window attribute, the rule set includes the logical matching condition of the blood glucose threshold interval and the insulin threshold interval, and the dynamic association threshold function of blood glucose and insulin, the engine calculates the blood glucose threshold interval based on the corrected value in real time, and calls the corresponding insulin threshold range for compliance matching, if the measured insulin value exceeds the association threshold range, it is marked as logical abnormality; The adjustment operation of the applicable condition of the association logic is realized by a three-dimensional decision matrix: The first dimension selects the venous blood exclusive or fingertip blood exclusive association rule base according to the blood sampling site type, the second dimension adjusts the division boundary of the blood glucose association interval according to the posture correction coefficient, and the third dimension dynamically loads the window period exclusive verification threshold combined with the determined time window position, and the optimal association rule combination is output by the matrix, and then the rule engine executes multi-layer logic verification, so as to eliminate the false qualified judgment caused by the interaction effect of posture, site and time window. 2.The clinical trial data real-time acquisition and quality control management platform of claim 1, wherein: The transmission delay processing unit (3) realizes the separation of transmission delay and operation delay by a double time stamp separation mechanism: When the device startup is completed, the first operation end time stamp is automatically generated and locally stored, and when the platform receives the data packet, the second platform receiving time stamp is generated, and the difference between the two is calculated as the pure transmission delay time, at the same time, the transmission delay processing unit (3) and the operation sequence delay calculation unit (2) interact in real time to obtain the actual delay time data of the operation sequence, and through the cooperative comparison of the time stamp difference calculation and the operation sequence delay data, the transmission delay component in the mixed delay is independently separated, and the device upload time stamp is ensured to only retain the operation end time information. 3.The clinical trial data real-time acquisition and quality control management platform of claim 1, wherein: The actual collection time determination unit (4) realizes the accurate positioning of the blood sampling completion time based on the reverse time restoration algorithm: Taking the device upload time stamp as the reference point, the pure transmission delay time data output by the transmission delay processing unit (3) and the operation delay time data output by the operation sequence delay calculation unit (2) are synchronously received, and the time sum of the pure transmission delay time data and the operation delay time data is subtracted in reverse through the time axis, to generate the actual blood sampling completion time point, wherein the operation delay time adopts a sequence accumulation model, including the measured time superposition value of the disinfection, blood sampling operation and device startup three stages, to ensure that the time restoration result is consistent with the actual operation node of the subject. 4.The clinical trial data real-time acquisition and quality control management platform of claim 3, wherein: The time window dynamic determination unit (5) performs qualifiedness verification through a dynamic offset window boundary determination mechanism: After receiving the actual blood sampling completion time point output by the actual collection time determination unit (4), it is compared with the start boundary and end boundary of the preset collection time window in real time, when the time point is located after the start boundary and before the end boundary, the qualified mark is triggered and output to the quality control database, when the time point is earlier than the start boundary or later than the end boundary, the timeout alarm is triggered and the data upload is frozen, the mechanism compensates for the systematic deviation between the restored time point and the original upload time stamp through the dynamic offset of the time window boundary. 5.The clinical trial data real-time acquisition and quality control management platform of claim 1, wherein: The real-time correction operation of the physiological index correction unit (6) is executed through a man-machine cooperative triggering mechanism: After the subject inputs the blood sampling posture and site information through the mobile terminal, the platform automatically collects the original blood glucose and insulin values uploaded by the equipment, calls the hierarchical correction architecture, corrects the original blood glucose value according to the posture dimension, corrects the original insulin value according to the site dimension, and stores the corrected index value in the original data, and synchronously marks the correction factor source for the dynamic correlation logic verification unit to trace back. 6.The clinical trial data real-time acquisition and quality control management platform of claim 1, wherein: The association relationship between the blood glucose threshold value and the insulin threshold value adopts a dynamic interval coupling mechanism: For blood glucose threshold interval division, each sub-interval is bound with an independent insulin threshold range, wherein the lower limit of the insulin threshold range corresponding to the high blood glucose interval is significantly higher than that of the low blood glucose interval, and the threshold range boundary value is dynamically adjusted according to the plasma proportion correction coefficient, when the blood glucose value is at the boundary of adjacent intervals, the threshold gradual change algorithm is used to smoothly transition the insulin threshold range, avoiding misjudgment caused by interval jumping.

Images