Insulin pump control method and system based on blood glucose data

Abstract

The invention discloses an insulin pump control method and system based on blood glucose data, and relates to the technical field of data control. Comprising the following steps: acquiring blood glucose data of a user; deducing the blood glucose data through a neural network model, and outputting model parameters of a preset time period most matched with the blood glucose data in the neural network model, the model parameters including a blood glucose fluctuation amplitude, an initial fluctuation time and a blood glucose fluctuation direction at a next moment; according to the model parameters, blood glucose change trends are judged, wherein the blood glucose change trends comprise the first blood glucose change trend, the second blood glucose change trend and the third blood glucose change trend; and adjusting the basic rate of the insulin pump according to the blood sugar change trend. The method has the beneficial effects that through a parameterized matching mechanism of the neural network model, the multi-dimensional characteristics of blood glucose change can be more accurately captured, the conversion from pure numerical prediction to mode recognition is realized, and the accuracy and reliability of blood glucose trend judgment are improved.

Description

Technical Field

[0001] The present invention relates to the field of data control technology, and in particular to a method and system for controlling an insulin pump based on blood sugar data. Background Art

[0002] Diabetic patients need to maintain blood sugar stability through insulin injections. Traditional insulin pump systems mainly rely on preset basal rates and large-dose injections before meals to control blood sugar. However, existing insulin pump control methods have many limitations. First, traditional methods often use simple feedback control mechanisms and only make adjustments based on current blood sugar values. They lack the ability to predict blood sugar change trends, resulting in a strong lag in control. Secondly, existing blood sugar prediction algorithms usually only focus on changes in blood sugar values, ignoring the time and direction characteristics of blood sugar fluctuations, making it difficult to fully grasp the complexity of blood sugar changes. Although the closed-loop control system in the existing technology has achieved automated regulation to a certain extent, it still has shortcomings in dealing with sharp blood sugar fluctuations, preventing hypoglycemia risks, and adapting to individual differences. Summary of the Invention

[0003] In view of the above-mentioned problems, the present invention is proposed.

[0004] Therefore, the technical problem solved by the present invention is that traditional methods mostly use a simple feedback control mechanism, which is adjusted only according to the current blood glucose value, lacks the ability to predict the trend of blood glucose changes, and ignores the time and direction characteristics of blood glucose fluctuations, making it difficult to fully grasp the complexity of blood glucose changes.

[0005] To solve the above technical problems, the present invention provides the following technical solution: a method for controlling an insulin pump based on blood glucose data, comprising the following steps: Get the user's blood sugar data; Inferring the blood glucose data using a neural network model, and outputting model parameters of the neural network model for a preset time period that best matches the blood glucose data, wherein the model parameters include the blood glucose fluctuation amplitude, the initial fluctuation time, and the blood glucose fluctuation direction at the next moment; determining a blood sugar change trend according to the model parameters, the blood sugar change trend including a first blood sugar change trend, a second blood sugar change trend, and a third blood sugar change trend; The basal rate of the insulin pump is adjusted according to the blood sugar change trend.

[0006] As a preferred embodiment of the insulin pump control method based on blood glucose data of the present invention, the step of outputting the model parameters includes: Inputting the blood glucose data into a pre-trained neural network model, wherein the pre-trained neural network model includes a plurality of model parameter groups corresponding to preset time periods, each model parameter group including a blood glucose fluctuation amplitude, an initial fluctuation time, and a blood glucose fluctuation direction at the next moment; Calculating the similarity between the blood glucose data and each model parameter group, wherein the similarity is determined by the Pearson correlation coefficient between the blood glucose data sequence and the blood glucose fluctuation pattern in the model parameter group; The model parameters with the highest similarity are selected as the model parameters that best match the blood glucose data.

[0007] The beneficial effects of this preferred technical solution are: selecting the optimal model parameters through a similarity matching mechanism, avoiding the numerical errors of traditional prediction methods, and improving the accuracy of blood sugar trend prediction.

[0008] As a preferred embodiment of the insulin pump control method based on blood sugar data of the present invention, the step of determining the blood sugar change trend according to the model parameters includes: Extracting the blood glucose fluctuation amplitude from the model parameters, setting multiple amplitude thresholds as a first judgment condition, and determining whether the blood glucose fluctuation amplitude is greater than any amplitude threshold. If it is greater than, determining it as a first blood glucose change trend; if it is not greater than, entering the initial fluctuation time judgment; Extract the initial fluctuation time from the model parameters, set multiple time thresholds as the second judgment condition, and judge whether the initial fluctuation time is greater than any time threshold. If it is greater than, it is determined as the second blood sugar change trend; if it is not greater than, it is determined as the third blood sugar change trend.

[0009] The beneficial effects of this preferred technical solution are: the hierarchical judgment mechanism avoids the limitations of single-dimensional judgment, and through progressive judgment of fluctuation amplitude and initial fluctuation time, it can more accurately distinguish different types of blood sugar change trends and improve the reliability and accuracy of trend classification.

[0010] As a preferred embodiment of the insulin pump control method based on blood sugar data of the present invention, determining the blood sugar change trend according to the model parameters further comprises: extracting a blood sugar fluctuation direction at a next moment from the model parameters, and performing directional correction on the first blood sugar variation trend, the second blood sugar variation trend, and the third blood sugar variation trend according to the blood sugar fluctuation direction at the next moment; When the next moment's blood sugar fluctuation direction is rising, increase the trend strength coefficient; when the next moment's blood sugar fluctuation direction is falling, decrease the trend strength coefficient; The trend strength coefficient is used to calculate the adjustment amplitude of the insulin pump basal rate.

[0011] The beneficial effects of this preferred technical solution are: the introduction of the trend intensity coefficient realizes the quantitative control of the basal rate adjustment, and performs dynamic correction according to the direction of blood sugar fluctuation, making the adjustment of the insulin pump more precise and avoiding the problem of over-adjustment or under-adjustment.

[0012] As a preferred embodiment of the blood glucose data-based insulin pump control method of the present invention, the preset time period includes four hourly time periods within a day, and determining the blood glucose change trend based on the model parameters further includes: Obtaining the current time point corresponding to the blood glucose data, and determining the hourly time period to which the current time point belongs; According to the hourly time period, the thresholds in the first judgment condition and the second judgment condition are adjusted in a time period-specific manner.

[0013] As a preferred embodiment of the insulin pump control method based on blood sugar data of the present invention, the step of adjusting the basal rate of the insulin pump according to the blood sugar change trend includes: When the blood sugar change trend is the first blood sugar change trend, increasing the basal rate of the insulin pump according to the trend intensity coefficient; When the blood sugar change trend is the second blood sugar change trend, reducing the basal rate of the insulin pump according to the trend intensity coefficient; When the blood sugar change trend is the third blood sugar change trend, the current basal rate of the insulin pump is maintained.

[0014] The beneficial effects of this preferred technical solution are: the differentiated basal rate adjustment strategy adopts corresponding control schemes for different blood sugar change trends, thereby achieving personalized and precise blood sugar management.

[0015] As a preferred embodiment of the insulin pump control method based on blood glucose data of the present invention, the step of increasing the basal rate of the insulin pump according to the trend intensity coefficient includes: determining whether the blood glucose data is in a hypoglycemic state, and if so, fine-tuning the basal rate according to the initial fluctuation time in the model parameters; and if not, increasing the basal rate of the insulin pump according to the trend strength coefficient; The steps for fine-tuning the base rate include: If the initial fluctuation time is greater than the preset time threshold, the base rate adjustment range is multiplied by the mitigation coefficient; if the initial fluctuation time is not greater than the preset time threshold, the base rate adjustment range is multiplied by the trend intensity coefficient.

[0016] The beneficial effects of this preferred technical solution are: a safety protection mechanism for hypoglycemia, risk assessment through initial fluctuation time, avoiding improper adjustments in hypoglycemia, and enhancing system safety.

[0017] As a preferred embodiment of the insulin pump control method based on blood glucose data of the present invention, the step of reducing the basal rate of the insulin pump according to the trend intensity coefficient includes: determining whether the blood glucose data is in a hyperglycemic state, and if so, adjusting the basal rate according to the blood glucose fluctuation direction at the next moment in the model parameters; if not, reducing the basal rate of the insulin pump according to the trend strength coefficient; The steps for fine-tuning the base rate include: If the blood sugar fluctuation direction at the next moment is upward, the basal rate adjustment amplitude is multiplied by the trend intensity coefficient to reduce the basal rate of the insulin pump. If the blood sugar fluctuation direction at the next moment is downward, the basal rate adjustment amplitude is multiplied by the relief coefficient to reduce the basal rate of the insulin pump.

[0018] The beneficial effects of this preferred technical solution are: refined control of hyperglycemia, differentiated treatment according to the direction of blood sugar fluctuation, ensuring the control effect while avoiding the risk of excessive blood sugar reduction.

[0019] The present invention provides an insulin pump control system based on blood sugar data.

[0020] To solve the above technical problems, the present invention provides the following technical solutions: an insulin pump control system based on blood glucose data, comprising a data acquisition module, a neural network model module, a blood glucose change trend judgment module, and an insulin pump basal rate adjustment module; The data acquisition module is used to obtain the user's blood sugar data; The neural network model module is used to infer the acquired blood glucose data; The blood sugar change trend judgment module judges the blood sugar change trend according to the model parameters output by the neural network model; The insulin pump basal rate adjustment module adjusts the basal rate of the insulin pump accordingly according to the determined blood sugar change trend.

[0021] As a preferred solution of the insulin pump control system based on blood glucose data described in the present invention, it also includes a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the insulin pump control method based on blood glucose data are implemented.

[0022] Beneficial effects of the present invention: This method utilizes a three-dimensional parameter combination: blood glucose fluctuation amplitude, initial fluctuation time, and fluctuation direction. This system constructs a description system for blood glucose variation. This overcomes the limitations of traditional methods that focus solely on changes in blood glucose values, and better reflects the temporal dynamics and trends of blood glucose variation. By introducing a hierarchical judgment mechanism and a trend intensity coefficient, it enables refined regulation of the insulin pump basal rate, avoiding the over- or under-regulation issues inherent in traditional methods and improving the stability and safety of blood glucose control.

[0023] By considering the physiological specificity of different time periods and adjusting the time period-specific thresholds, it better adapts to the influence of the human body's circadian rhythm, makes the control strategy of the insulin pump more in line with physiological laws, and improves the physiological adaptability of blood sugar management.

[0024] By selecting the optimal model parameters through the similarity matching algorithm, it can better adapt to individual differences and blood sugar change patterns under different physiological states, realize personalized blood sugar management, and improve the accuracy and applicability of insulin pump control. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0026] Figure 1 This is an overall flow chart of a blood glucose data-based insulin pump control method provided by one embodiment of the present invention. DETAILED DESCRIPTION

[0027] To make the above-mentioned objects, features, and advantages of the present invention more clearly understood, the following detailed description of the specific embodiments of the present invention is given in conjunction with the accompanying drawings. It is obvious that the described embodiments are only part of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in this field without creative work should fall within the scope of protection of the present invention.

[0028] Example 1, reference Figure 1 , is an embodiment of the present invention, which provides an insulin pump control method based on blood glucose data, comprising the following steps: S1. Obtain the user's blood sugar data.

[0029] In this embodiment, obtaining the user's blood glucose data is achieved through a continuous glucose monitoring system (CGM). Specifically, obtaining the blood glucose data includes the following steps: First, a blood glucose sensor implanted in the user's subcutaneous tissue monitors interstitial fluid glucose concentration in real time. The sensor uses enzyme electrode technology, where glucose oxidase reacts with glucose to generate a current signal whose intensity is proportional to the blood glucose concentration. The sensor collects blood glucose data every 15 seconds, and the raw current signal is converted to a digital signal via an analog-to-digital converter.

[0030] Secondly, the collected blood glucose data is preprocessed. This includes: removing obviously abnormal blood glucose values. Specifically, blood glucose values ​​outside the physiological range of 1.1-33.3 mmol / L are considered abnormal data; performing a sliding average filter on continuous blood glucose data with a filter window of three data points to reduce the impact of sensor noise; and supplementing missing blood glucose data. When the data is missing for less than 5 minutes, linear interpolation is used to complete the data.

[0031] Next, construct a blood glucose data sequence. Arrange the preprocessed blood glucose data in chronological order to form a blood glucose data sequence. The blood glucose data sequence includes the blood glucose value, the corresponding timestamp, and the data quality indicator. In this embodiment, the blood glucose data within the last two hours is selected as the input data sequence, with 480 data points (one data point every 15 seconds).

[0032] Finally, blood glucose data is normalized. The blood glucose value is converted to a deviation from the individual's baseline blood glucose level using the formula: Normalized blood glucose value = (Current blood glucose value - Individual baseline blood glucose value) / Individual blood glucose standard deviation. The individual baseline blood glucose value is calculated using the user's blood glucose data from the past seven days, and the individual blood glucose standard deviation is also calculated based on blood glucose variability over the past seven days.

[0033] Through the above steps, a preprocessed and standardized blood glucose data sequence was obtained. This sequence contains the numerical changes, time information and individual characteristics of blood glucose, providing high-quality input data for subsequent neural network model inference.

[0034] S2. Infer the blood glucose data through the neural network model and output the model parameters of the preset time period in the neural network model that best match the blood glucose data, wherein the model parameters include the blood glucose fluctuation amplitude, the initial fluctuation time, and the blood glucose fluctuation direction at the next moment.

[0035] The output steps of model parameters include A1~A3: A1. Input the blood glucose data into a pre-trained neural network model. The pre-trained neural network model includes model parameter groups corresponding to multiple preset time periods. Each model parameter group includes the blood glucose fluctuation amplitude, initial fluctuation time, and blood glucose fluctuation direction at the next moment. First, the standardized blood glucose data sequence obtained in step S1 is input into the neural network model. In this embodiment, the pre-trained neural network model contains 96 model parameter groups corresponding to preset time periods. These time periods cover 24 hours a day, and each time period is 15 minutes. Each model parameter group contains three key parameters: The blood glucose fluctuation amplitude parameter indicates the intensity of blood glucose changes during this time period, with a numerical range of 0.5-8.0 mmol / L; the initial fluctuation time parameter indicates the time point when blood glucose begins to change significantly, with a numerical range of 0-30 minutes; the blood glucose fluctuation direction parameter at the next moment uses a numerical coding method, where 1 indicates an increase, -1 indicates a decrease, and 0 indicates stability.

[0036] The neural network model uses forward propagation to map the 480-dimensional blood glucose data sequence into a feature space of 96 model parameter groups. Each model parameter group corresponds to a specific blood glucose variation pattern, such as postprandial blood glucose rise, nighttime blood glucose drop, and stress-induced blood glucose fluctuation.

[0037] A2. Calculate the similarity between the blood glucose data and each model parameter set. The similarity is determined by the Pearson correlation coefficient between the blood glucose data sequence and the blood glucose fluctuation pattern in the model parameter set. Similarity was calculated using the Pearson correlation coefficient. Specifically, we first reconstructed the corresponding blood glucose fluctuation pattern sequence based on each model parameter set. The reconstruction process was as follows: Using the blood glucose fluctuation amplitude as the intensity of change, the initial fluctuation time as the starting point of the change, and the fluctuation direction as the trend of change, we generated a theoretical blood glucose sequence of length 480.

[0038] The calculation formula is: reconstructed blood glucose value (t) = baseline blood glucose value + blood glucose fluctuation amplitude × direction coefficient × time decay function (t-initial fluctuation time), where the time decay function adopts exponential decay form: , τ is the time constant, which is set to 60 minutes.

[0039] Then, calculate the Pearson correlation coefficient between the actual blood glucose data sequence and each reconstructed blood glucose sequence. The calculation formula is: ,in, is the actual blood sugar data, To reconstruct blood glucose data, and are the corresponding means respectively.

[0040] In this embodiment, 96 correlation coefficient values ​​are calculated, with a numerical range of -1 to 1, wherein the closer the correlation coefficient is to 1, the higher the similarity.

[0041] A3. Select the model parameters with the highest similarity as the model parameters that best match the blood glucose data.

[0042] By comparing the 96 Pearson correlation coefficients, the model parameter group corresponding to the largest correlation coefficient is selected. In this example, assume that the 42nd model parameter group has the highest correlation coefficient of 0.87. This parameter group corresponds to the post-lunch time period (12:00-12:15). Its model parameters are: blood glucose fluctuation amplitude of 3.2mmol / L, initial fluctuation time of 8 minutes, and blood glucose fluctuation direction of 1 (increasing) at the next moment.

[0043] To ensure the reliability of the selection, the system also sets a similarity threshold of 0.6. When the highest correlation coefficient is lower than this threshold, the system will select multiple model parameter groups with high similarity for weighted averaging, with weights allocated based on similarity.

[0044] Through the above steps, the model parameter group that best matches the current blood sugar data is successfully output. This parameter group contains the key characteristic information of blood sugar fluctuations and provides accurate input data for subsequent judgment of blood sugar change trends.

[0045] S3. Determine a blood sugar change trend according to the model parameters. The blood sugar change trend includes a first blood sugar change trend, a second blood sugar change trend, and a third blood sugar change trend.

[0046] The steps of judging the blood sugar change trend based on the model parameters include B1 and B2: B1. Extracting the blood glucose fluctuation amplitude from the model parameters, setting multiple amplitude thresholds as the first judgment condition, and determining whether the blood glucose fluctuation amplitude is greater than any of the amplitude thresholds. If it is greater than, determining it as the first blood glucose change trend; if it is not greater than, entering the initial fluctuation time judgment; In this embodiment, the blood glucose fluctuation amplitude of 3.2 mmol / L from the model parameters obtained in step S2 is extracted. This value is compared with the adjusted amplitude thresholds: compared with the first amplitude threshold of 4.5 mmol / L, 3.2 < 4.5, which does not meet the condition; compared with the second amplitude threshold of 6.0 mmol / L, 3.2 < 6.0, which does not meet the condition; compared with the third amplitude threshold of 7.5 mmol / L, 3.2 < 7.5, which does not meet the condition.

[0047] Since the blood sugar fluctuation amplitude of 3.2mmol / L is not greater than any amplitude threshold, the system determines that it does not meet the conditions of the first blood sugar change trend and enters the initial fluctuation time judgment process.

[0048] B2. Extract the initial fluctuation time from the model parameters, set multiple time thresholds as the second judgment condition, and judge whether the initial fluctuation time is greater than any time threshold. If it is greater than, it is determined as the second blood glucose change trend; if it is not greater than, it is determined as the third blood glucose change trend.

[0049] Extract the initial fluctuation time of 8 minutes from the model parameters. Compare this value with multiple adjusted time thresholds: Compared with the first time threshold of 8 minutes, 8 = 8, meeting the condition; compared with the second time threshold of 15 minutes, 8 < 15, failing the condition; compared with the third time threshold of 25 minutes, 8 < 25, failing the condition.

[0050] Since the initial fluctuation time of 8 minutes is equal to the first time threshold, the system determines it as the second blood sugar change trend. The second blood sugar change trend indicates that the blood sugar change has a moderate degree of urgency and requires moderate insulin adjustment.

[0051] The blood sugar trend can also be determined based on the model parameters, including B3 and B4: B3. extracting the next moment's blood sugar fluctuation direction from the model parameters, and performing directional corrections on the first blood sugar variation trend, the second blood sugar variation trend, and the third blood sugar variation trend according to the next moment's blood sugar fluctuation direction; The next-moment blood sugar fluctuation direction parameter in the model parameters is set to 1 (increasing). Based on this direction information, the system performs a directional correction on the second blood sugar trend. Because the blood sugar fluctuation direction is increasing and it is currently early morning, the system adjusts the urgency level of the second blood sugar trend from "medium" to "moderately high" to account for the physiological characteristics of early morning blood sugar rises.

[0052] B4. When the next blood sugar fluctuation direction is upward, increase the trend strength coefficient. When the next blood sugar fluctuation direction is downward, decrease the trend strength coefficient. The trend strength coefficient is used to calculate the adjustment range of the insulin pump basal rate.

[0053] Based on the next moment's upward blood sugar fluctuation direction, increase the trend strength coefficient. The specific calculation formula is: Trend Strength Coefficient = Base Coefficient × Direction Correction Factor × Time Period Correction Factor. In this embodiment, the base coefficient is set to 1.0, the upward correction factor is 1.2, and the early morning correction factor is 1.1. Therefore, the trend strength coefficient = 1.0 × 1.2 × 1.1 = 1.32.

[0054] This trend strength coefficient will be used to calculate the adjustment range of the insulin pump basal rate in the subsequent steps. A larger value indicates that a larger insulin adjustment is required.

[0055] Through the above steps, the system successfully determines that the current blood sugar change trend is the second blood sugar change trend after direction correction, and calculates the trend intensity coefficient of 1.32, providing an accurate parameter basis for subsequent insulin pump control.

[0056] It is also important to note that the preset time period includes the four hourly periods of the day, and the blood sugar trend determined by the model parameters also includes: Obtain the current time point corresponding to the blood glucose data and determine the hourly time period to which the current time point belongs; In this example, assuming the current time is 7:30 a.m., the system divides the 24-hour day into four main time periods: morning (6:00-12:00), afternoon (12:00-18:00), evening (18:00-24:00), and night (12:00-6:00).

[0057] Based on the current time of 7:30, the system determines that the current time point is in the early morning time period. Each time period corresponds to different physiological states: the early morning time period considers the effects of dawn and breakfast; the afternoon time period considers blood sugar changes after lunch; the evening time period considers the effects of dinner and activity; and the night time period considers basal metabolic state.

[0058] According to the hourly time period, the thresholds in the first judgment condition and the second judgment condition are adjusted time period-specifically. The threshold in the early morning time period is set higher than the thresholds in other time periods to adapt to the blood sugar fluctuation characteristics of the dawn phenomenon.

[0059] In this embodiment, the thresholds in the judgment criteria are adjusted based on the early morning time period. Specifically, the amplitude thresholds for the early morning time period are set to: a first amplitude threshold of 4.5 mmol / L, a second amplitude threshold of 6.0 mmol / L, and a third amplitude threshold of 7.5 mmol / L. These values ​​are all 1.0 mmol / L higher than the thresholds for other time periods to accommodate the natural rise in blood sugar caused by the dawn phenomenon.

[0060] At the same time, the time thresholds for the early morning time period are adjusted to: 8 minutes for the first time threshold, 15 minutes for the second time threshold, and 25 minutes for the third time threshold, which are 3-5 minutes longer than other time periods to adapt to the lower insulin sensitivity in the early morning.

[0061] S4. Adjust the basal rate of the insulin pump according to the trend of blood sugar changes.

[0062] C1. When the blood sugar change trend is the first blood sugar change trend, the basal rate of the insulin pump is increased according to the trend intensity coefficient; The steps for increasing the basal rate of an insulin pump based on the trend strength factor include: Determine whether the blood sugar data is in a hypoglycemic state. If so, fine-tune the basal rate according to the initial fluctuation time in the model parameters; if not, increase the basal rate of the insulin pump according to the trend intensity coefficient; The steps for fine-tuning the base rate include: If the initial fluctuation time is greater than the preset time threshold, the base rate adjustment range is multiplied by the relief coefficient. If the initial fluctuation time is not greater than the preset time threshold, the base rate adjustment range is multiplied by the trend intensity coefficient.

[0063] When the blood sugar change trend is the first blood sugar change trend, it indicates that the blood sugar fluctuation amplitude is large and requires significant adjustment of insulin infusion. Assume in another embodiment that the user's blood sugar fluctuation amplitude is 5.5mmol / L, exceeding the adjusted first amplitude threshold of 4.5mmol / L, it is determined to be the first blood sugar change trend, and the trend intensity coefficient is 1.45.

[0064] First, determine the blood sugar status: the current blood sugar value is 8.2mmol / L. The hypoglycemia threshold is set at 3.9mmol / L. Since 8.2 > 3.9, it is determined to be non-hypoglycemic. Therefore, the basal rate of the insulin pump is directly increased based on the trend strength coefficient of 1.45.

[0065] Basal rate adjustment calculation: The current basal rate is 1.2 units / hour, and the basal adjustment range is set to 0.3 units / hour. The adjusted basal rate = 1.2 + (0.3 × 1.45) = 1.2 + 0.435 = 1.635 units / hour.

[0066] Treatment for hypoglycemia: Assuming the current blood glucose level is 3.5 mmol / L, indicating hypoglycemia, fine-tune the insulin dose based on the initial fluctuation time. The initial fluctuation time is 12 minutes, and the preset time threshold is 10 minutes. Since 12 > 10, multiply the basal rate adjustment by the relief factor of 0.5. The adjusted basal rate = 1.2 + (0.3 × 0.5) = 1.35 units / hour. This prevents excessive insulin infusion during hypoglycemia.

[0067] C2. When the blood sugar change trend is the second blood sugar change trend, the basal rate of the insulin pump is reduced according to the trend intensity coefficient; The steps to reduce the basal rate of an insulin pump based on the trend strength factor include: Determine whether the blood sugar data is in a hyperglycemic state. If so, adjust the basal rate according to the next moment's blood sugar fluctuation direction in the model parameters; if not, appropriately reduce the basal rate of the insulin pump according to the trend strength coefficient; The steps for fine-tuning the base rate include: If the blood sugar fluctuation direction at the next moment is upward, the basal rate adjustment amplitude will be multiplied by the trend intensity coefficient to reduce the basal rate of the insulin pump. If the blood sugar fluctuation direction at the next moment is downward, the basal rate adjustment amplitude will be multiplied by the relief coefficient to reduce the basal rate of the insulin pump.

[0068] According to the judgment result of step S3, the current blood sugar change trend is the second blood sugar change trend, and the trend intensity coefficient is 1.32. According to the processing logic of the second blood sugar change trend, the basal rate of the insulin pump needs to be reduced.

[0069] First, determine the blood sugar status: the current blood sugar value is 7.8mmol / L. The hyperglycemia threshold is set at 10.0mmol / L. Since 7.8 < 10.0, it is determined to be non-hyperglycemic. Therefore, the basal rate of the insulin pump is reduced based on the trend strength coefficient of 1.32.

[0070] Basal rate adjustment calculation: The current basal rate is 1.2 units / hour, and the basal adjustment range is set to 0.2 units / hour. Since it is a reduction operation, the adjusted basal rate = 1.2-(0.2×1.32) = 1.2-0.264 = 0.936 units / hour.

[0071] Treatment for hyperglycemia: Assuming the current blood glucose level is 12.5 mmol / L, indicating hyperglycemia, adjustments are made based on the direction of blood glucose fluctuation at the next moment. The fluctuation direction, as determined in step S3, is rising (parameter value 1). Since the blood glucose level is rising and hyperglycemia is present, the basal rate adjustment is multiplied by the trend strength coefficient of 1.32. The adjusted basal rate = 1.2 - (0.2 × 1.32) = 0.936 units / hour, achieving more aggressive glucose-lowering treatment.

[0072] If the direction of blood sugar fluctuation at the next moment is downward: Assume that the fluctuation direction parameter is -1 (downward), indicating that blood sugar has a natural downward trend, then use the relief coefficient 0.7 for mild adjustment. After adjustment, the basal rate = 1.2-(0.2×0.7) = 1.06 units / hour to avoid excessive blood sugar lowering.

[0073] C3. When the blood sugar change trend is the third blood sugar change trend, maintain the current basal rate of the insulin pump.

[0074] When the blood sugar change trend is the third blood sugar change trend, it means that the blood sugar fluctuation is small and the initial fluctuation time is short, and the blood sugar state is relatively stable. In this case, the system maintains the current basal rate of the insulin pump unchanged.

[0075] Implementation: The system maintains the current basal rate of 1.2 units / hour and activates continuous monitoring mode, reassessing blood sugar trends every 5 minutes. If three consecutive assessments show the same trend, the monitoring interval is extended to 10 minutes to reduce unnecessary system resource consumption.

[0076] In this embodiment, the non-hyperglycemic state processing in C2 is finally executed, and the basal rate of the insulin pump is adjusted from 1.2 units / hour to 0.936 units / hour, with an adjustment range of -22%. This adjustment will be gradually implemented over the next 15 minutes to ensure a smooth transition of blood sugar control.

[0077] Example 2 is an embodiment of the present invention, which provides an insulin pump control system based on blood glucose data, including a data acquisition module, a neural network model module, a blood glucose change trend judgment module, and an insulin pump basal rate adjustment module; The data acquisition module is used to obtain the user's blood sugar data; The neural network model module is used to infer the acquired blood glucose data; The blood sugar change trend judgment module judges the blood sugar change trend based on the model parameters output by the neural network model; The insulin pump basal rate adjustment module adjusts the basal rate of the insulin pump accordingly based on the determined blood sugar change trend.

[0078] The invention also includes a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of a method for controlling an insulin pump based on blood glucose data.

[0079] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for controlling an insulin pump based on blood glucose data, characterized in that: The following steps are involved: Get the user's blood sugar data; Inferring the blood glucose data using a neural network model, and outputting model parameters of the neural network model for a preset time period that best matches the blood glucose data, wherein the model parameters include the blood glucose fluctuation amplitude, the initial fluctuation time, and the blood glucose fluctuation direction at the next moment; determining a blood sugar change trend according to the model parameters, the blood sugar change trend including a first blood sugar change trend, a second blood sugar change trend, and a third blood sugar change trend; The basal rate of the insulin pump is adjusted according to the blood sugar change trend.

2. The insulin pump control method based on blood sugar data according to claim 1, characterized in that: The output step of the model parameters includes: Inputting the blood glucose data into a pre-trained neural network model, wherein the pre-trained neural network model includes a plurality of model parameter groups corresponding to preset time periods, each model parameter group including a blood glucose fluctuation amplitude, an initial fluctuation time, and a blood glucose fluctuation direction at the next moment; Calculating the similarity between the blood glucose data and each model parameter group, wherein the similarity is determined by the Pearson correlation coefficient between the blood glucose data sequence and the blood glucose fluctuation pattern in the model parameter group; The model parameters with the highest similarity are selected as the model parameters that best match the blood glucose data.

3. The insulin pump control method based on blood sugar data according to claim 2, characterized in that: The step of determining the blood sugar change trend according to the model parameters includes: Extracting the blood glucose fluctuation amplitude from the model parameters, setting multiple amplitude thresholds as a first judgment condition, and determining whether the blood glucose fluctuation amplitude is greater than any amplitude threshold. If it is greater than, determining it as a first blood glucose change trend; if it is not greater than, entering the initial fluctuation time judgment; Extract the initial fluctuation time from the model parameters, set multiple time thresholds as the second judgment condition, and judge whether the initial fluctuation time is greater than any time threshold. If it is greater than, it is determined as the second blood sugar change trend; if it is not greater than, it is determined as the third blood sugar change trend.

4. The insulin pump control method based on blood sugar data according to claim 3, characterized in that: Determining the blood sugar change trend according to the model parameters also includes: extracting a blood sugar fluctuation direction at a next moment from the model parameters, and performing directional correction on the first blood sugar variation trend, the second blood sugar variation trend, and the third blood sugar variation trend according to the blood sugar fluctuation direction at the next moment; When the next moment's blood sugar fluctuation direction is rising, increase the trend strength coefficient; when the next moment's blood sugar fluctuation direction is falling, decrease the trend strength coefficient; The trend strength coefficient is used to calculate the adjustment amplitude of the insulin pump basal rate.

5. The insulin pump control method based on blood sugar data according to claim 4, characterized in that: The preset time period includes four hourly time periods in a day, and determining the blood sugar change trend according to the model parameters further includes: Obtaining the current time point corresponding to the blood glucose data, and determining the hourly time period to which the current time point belongs; According to the hourly time period, the thresholds in the first judgment condition and the second judgment condition are adjusted in a time period-specific manner.

6. The insulin pump control method based on blood sugar data according to claim 5, characterized in that: The steps of adjusting the basal rate of the insulin pump according to the blood sugar change trend include: When the blood sugar change trend is the first blood sugar change trend, increasing the basal rate of the insulin pump according to the trend intensity coefficient; When the blood sugar change trend is the second blood sugar change trend, reducing the basal rate of the insulin pump according to the trend intensity coefficient; When the blood sugar change trend is the third blood sugar change trend, the current basal rate of the insulin pump is maintained.

7. The insulin pump control method based on blood sugar data according to claim 6, characterized in that: The step of increasing the basal rate of the insulin pump according to the trend strength coefficient comprises: determining whether the blood glucose data is in a hypoglycemic state, and if so, fine-tuning the basal rate according to the initial fluctuation time in the model parameters; and if not, increasing the basal rate of the insulin pump according to the trend strength coefficient; The steps for fine-tuning the base rate include: If the initial fluctuation time is greater than the preset time threshold, the base rate adjustment range is multiplied by the mitigation coefficient; if the initial fluctuation time is not greater than the preset time threshold, the base rate adjustment range is multiplied by the trend intensity coefficient.

8. The insulin pump control method based on blood sugar data according to claim 7, characterized in that: The step of reducing the basal rate of the insulin pump according to the trend strength coefficient comprises: determining whether the blood glucose data is in a hyperglycemic state, and if so, adjusting the basal rate according to the blood glucose fluctuation direction at the next moment in the model parameters; if not, reducing the basal rate of the insulin pump according to the trend strength coefficient; The steps for fine-tuning the base rate include: If the blood sugar fluctuation direction at the next moment is upward, the basal rate adjustment amplitude is multiplied by the trend intensity coefficient to reduce the basal rate of the insulin pump. If the blood sugar fluctuation direction at the next moment is downward, the basal rate adjustment amplitude is multiplied by the relief coefficient to reduce the basal rate of the insulin pump.

9. An insulin pump control system based on blood glucose data, applying the insulin pump control method based on blood glucose data according to any one of claims 1 to 8, characterized in that: It includes data acquisition module, neural network model module, blood sugar change trend judgment module, and insulin pump basal rate adjustment module; The data acquisition module is used to obtain the user's blood sugar data; The neural network model module is used to infer the acquired blood glucose data; The blood sugar change trend judgment module judges the blood sugar change trend according to the model parameters output by the neural network model; The insulin pump basal rate adjustment module adjusts the basal rate of the insulin pump accordingly according to the determined blood sugar change trend.

10. An insulin pump control system based on blood sugar data, characterized in that: The invention also includes a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the steps of the insulin pump control method based on blood glucose data according to any one of claims 1 to 8.

Images