Insulin pump control system
By using multimodal data acquisition and hierarchical predictive control, combined with intelligent infusion and safety redundancy mechanisms, the problems of incomplete parameter coverage and insufficient predictive ability in existing insulin pump systems have been solved, achieving accurate blood glucose prediction and dynamic adaptation, and improving the infusion accuracy and safety of the system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-23
- Publication Date
- 2026-03-27
AI Technical Summary
Existing insulin pump control systems rely on a single parameter, ignoring the dynamic influence of the environment and behavior, resulting in insufficient infusion accuracy, a lack of predictive control, and a lagging response mechanism.
A multimodal data acquisition module is used, which combines physiological, environmental and behavioral parameters, and uses an LSTM neural network for hierarchical predictive control. Combined with an intelligent infusion module and a safety redundancy module, it can dynamically adjust the insulin infusion rate and ensure infusion safety through dual controller cross-checking and sensor calibration.
It achieves accurate prediction and dynamic adaptation of blood glucose, reduces the incidence of post-exercise hypoglycemia, improves the system's fault safety and response speed, and meets clinical needs.
Smart Images
Figure CN121731597A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of medical equipment, in particular to an insulin pump control system. BACKGROUND
[0002] With the development of economy, population aging, improvement of living standards and lifestyle changes, diabetes has become a global public health problem affecting human health. Diabetes is a metabolic disease characterized by high blood sugar due to insulin secretion and action defects. Diabetes patients often have abnormal fat and protein metabolism, and long-term high blood sugar can cause multiple organ dysfunction or failure, especially heart, eye, blood vessels, kidney and organ function. Insulin therapy plays a very important role in the management of diabetes, and the insulin intensive treatment regimen mainly includes daily multiple subcutaneous insulin injection and continuous subcutaneous insulin infusion, i.e. insulin pump.
[0003] Insulin pump therapy is an insulin infusion device controlled by artificial intelligence, which continuously infuses insulin subcutaneously at a programmed rate to simulate the physiological secretion pattern of human insulin to a maximum extent, thereby achieving a better control of blood glucose.
[0004] Most of the existing insulin pump control systems only rely on blood glucose values or limited physiological parameters, ignoring the dynamic influence of environment and behavior, resulting in insufficient infusion accuracy; at the same time, the traditional closed-loop control system needs to wait for blood glucose abnormalities before adjusting the infusion, lacks predictive control, and the response mechanism is lagging. SUMMARY
[0005] The purpose of the present application is to provide an insulin pump control system, which aims to solve the technical problems of incomplete parameter coverage and weak prediction ability of the existing system, and to realize accurate prediction of blood glucose, dynamic adaptation of infusion and safe management and control of full-link data.
[0006] To achieve the above purpose, the present application provides an insulin pump control system, which comprises a multi-modal data acquisition module, a hierarchical prediction control module, an intelligent infusion module, a safety redundancy module, a data encryption transmission module and a user interaction module. The multi-modal data acquisition module simultaneously acquires physiological parameters, environmental parameters and behavior parameters, the hierarchical prediction control module outputs predicted blood glucose values based on LSTM neural network, the intelligent infusion module dynamically adjusts the insulin infusion rate according to the prediction results, the safety redundancy module performs mutual inspection through double controllers, and if the monitoring is abnormal, it triggers an audible and light alarm and stops insulin infusion.
[0007] Preferably, the hierarchical prediction control module comprises a short-term prediction layer and a medium-long term adjustment layer, the medium-long term adjustment layer stores 7-day history data through an embedded database (capacity ≥ 16 GB) and corrects the basal infusion rate (amplitude ± 5% to ± 15%) daily; The blood glucose prediction formula of the short-term prediction layer is as follows: ; Wherein, represents the predicted blood glucose value in the future t minutes; represents the current measured blood glucose value; represents the physiological parameter contribution value ( = 1: insulin residual amount influence coefficient, = 2: heart rate fluctuation coefficient, i = 3: cortisol concentration coefficient, i = 4: blood glucose history trend coefficient); i represents the environmental parameter contribution value ( = 1: temperature correction coefficient, i = 2: altitude and pressure coefficient, = 3: humidity interference coefficient); j represents the behavior parameter contribution value ( = 1: exercise intensity metabolic coefficient, j = 2: food nutrition absorption coefficient); j 、 、 k All represent parameter weights, satisfying k . . .
[0008] Preferably, the intelligent infusion module comprises an insulin pump and a single-cavity infusion pipeline, the insulin pump adopts a piezoelectric ceramic micropump, and the single-cavity infusion pipeline is suitable for a medical standard infusion needle, so as to reduce the discomfort of the patient during puncture.
[0009] Preferably, the control process of the intelligent infusion module is as follows: Step S201, receiving the predicted blood glucose value in the future 10 minutes output by the hierarchical prediction control module . Step S202, calculating the insulin infusion rate based on the predicted blood glucose value and the clinical safety standard IR . Step S203, driving the insulin pump to accurately infuse according to the calculation result, and monitoring the infusion pressure of the single-cavity infusion pipeline in real time through a pressure sensor P tude . Step S204, if the infusion pressure is abnormal P tude If the pressure value is greater than 50 kPa and lasts for 5 seconds, the insulin infusion is immediately suspended and the safety redundancy module alarm is triggered.
[0010] Preferably, the insulin infusion rate in step S202 is: ; wherein, represents the basal insulin infusion rate; represents the basal rate correction coefficient; k represents the hyperglycemia infusion adjustment coefficient.
[0011] Preferably, the safety redundancy module comprises a dual-controller mutual inspection unit, a sensor double calibration unit, and an infusion blockage unit. The dual-controller mutual inspection unit comprises a main controller and a backup controller, and uses STM32 main / backup chips, which are inspected once every 100 ms. The mutual inspection formula is specifically: ; wherein, represents the main controller output value; represents the backup controller output value; the main controller output value and the backup controller output value both comprise an insulin infusion rate IR and a predicted blood glucose value ; DR represents the deviation rate; If DR≤2%, the system is normally operated, and the mutual inspection is continued at an interval of 100 ms; If 2%≤DR≤5%, a yellow audible and light alarm is triggered, the deviation log is automatically recorded, the mutual inspection frequency is increased to once every 10 ms, and the system is continued to be operated; If DR>5%, a red audible and light alarm is triggered, the insulin infusion is immediately stopped, the backup controller is switched to operate alone, and the user is prompted to contact the maintenance through the user interaction module.
[0012] Preferably, the sensor double calibration unit cross-calibrates every 15 minutes through near-infrared spectroscopy (wavelength 1550 nm) and electrochemical sensing. When the blood glucose detection deviation is greater than 0.2 mmol / L, the correction is made (the infrared spectroscopy detection value is , and the electrochemical sensing detection value is ); the infusion blockage detection unit monitors the infusion pressure of the single-lumen infusion pipeline through a pressure sensor (accuracy ±1 kPa) P tude , and when the pressure value is greater than 50 kPa and lasts for 5 seconds, P tude the pipeline is determined to be blocked, the insulin infusion is immediately stopped, and the alarm is triggered.
[0013] Preferably, the data encryption transmission module adopts the Bluetooth 5.3 protocol, with a transmission rate of 2Mbps and a communication distance of ≤10m. It integrates the AES-256 encryption algorithm (the key is automatically updated every 24 hours). At the same time, it presets parameter thresholds (blood glucose <1.1 or blood glucose >33.3mmol / L, temperature <-10 or >40℃). Abnormal data exceeding the threshold is intercepted in real time and triggers a local alarm. After interception, manual confirmation is required before transmission can be resumed.
[0014] Preferably, the user interaction module includes a 2.4-inch IPS touch screen (320×240 resolution, visible in sunlight), an audible and visual alarm unit (80dB buzzer + tri-color LED), a historical data query unit (supports exporting data from the past 30 days to PDF / Excel), and an emergency control button (press and hold for 2 seconds to pause infusion / start emergency rescue); the audible and visual alarm unit supports 3 tones (hypoglycemia: low-frequency buzzer + flashing green LED; hyperglycemia: medium-frequency buzzer + flashing yellow LED; device malfunction: high-frequency buzzer + solid red LED), and automatically reduces alarm sensitivity during nighttime sleep (triggered only when blood glucose < 3.5 mmol / L).
[0015] Preferably, the LSTM neural network of the short-term prediction layer includes 12 neurons in the input layer, 64 neurons in the hidden layer, and 1 neuron in the output layer, with 500 iterations, a learning rate of 0.001, and a loss function of MSE; the accuracy is verified using the MARD accuracy verification formula, specifically: ; in, express t The blood glucose value was measured after minutes; n represents the number of test samples, n≥100, and the samples must cover the blood glucose range of 1.1~33.3mmol / L and daily, exercise, high temperature and sleep scenarios; MARD is automatically calculated once every 24 hours. If MARD > 6.8%, a sensor calibration reminder is triggered. If it is still > 8% after calibration, it is determined that the prediction module is faulty, automatic infusion is stopped, and the system is switched to manual infusion mode, prompting the user to contact technical personnel.
[0016] Therefore, the present invention employs the above-mentioned insulin pump control system, and the specific technical effects are as follows: Breakthrough in Innovative Architecture: The "Physiological-Environmental-Behavioral-Safety" four-dimensional data fusion architecture breaks through the limitations of single parameters, integrates a safety mechanism through a closed-loop design across the entire chain, and achieves multi-dimensional technological innovation; Performance metrics leap: The hierarchical prediction algorithm extends the blood glucose prediction cycle to 30 minutes with MARD≤6.8%, and the 5-second response in extreme scenarios reduces the incidence of post-exercise hypoglycemia to 5%. The safety redundancy design reduces the failure probability to 0.1%, and the core performance far exceeds that of existing technologies. Application compatibility: The core components are mass-produced and controllable, compatible with existing production lines (yield ≥ 98%), suitable for patients with type 1 / 2 diabetes and with a TIR of 89%, convenient operation and wide environmental adaptability (-10℃~40℃, -500m~5000m) meet the needs of clinical and practical use, and the single-charge battery life ≥ 72 hours improves the user experience.
[0017] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0018] Figure 1 This is a diagram showing the components of an embodiment of an insulin pump control system according to the present invention. Detailed Implementation
[0019] The technical solution of the present invention will be further described below with reference to the accompanying drawings and embodiments.
[0020] Unless otherwise defined, the technical or scientific terms used in this invention shall have the ordinary meaning understood by one of ordinary skill in the art to which this invention pertains. The terms "first," "second," and similar terms used in this invention do not indicate any order, quantity, or importance, but are merely used to distinguish different components. Terms such as "comprising" or "including" mean that the element or object preceding the word encompasses the elements or objects listed following the word and their equivalents, without excluding other elements or objects. Terms such as "connected" or "linked" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. Terms such as "upper," "lower," "left," and "right" are used only to indicate relative positional relationships; when the absolute position of the described object changes, the relative positional relationship may also change accordingly.
[0021] like Figure 1 As shown, the present invention provides an insulin pump control system, including a multimodal data acquisition module, a hierarchical predictive control module, an intelligent infusion module, a safety redundancy module, a data encryption transmission module, and a user interaction module; the multimodal data acquisition module simultaneously acquires physiological parameters, environmental parameters, and behavioral parameters.
[0022] The hierarchical prediction and control module outputs predicted blood glucose values based on an LSTM neural network. The hierarchical prediction and control module includes a short-term prediction layer and a medium- and long-term adjustment layer. The medium- and long-term adjustment layer stores 7 days of historical data through an embedded database (capacity ≥16GB) and adjusts the basic infusion rate daily (range ±5%~±15%). The formula for predicting blood glucose in the short-term prediction layer is: ; in, Indicates the future tminute Predicted blood glucose values; This indicates the current measured blood glucose level; Indicates the contribution value of physiological parameters ( i =1: Influence coefficient of residual insulin level, i =2: Heart rate variability coefficient, i =3: Cortisol concentration coefficient i =4: Historical trend coefficient of blood glucose). Indicates the contribution value of environmental parameters ( j =1: Temperature correction factor j =2: Altitude air pressure coefficient j =3: Humidity interference coefficient); Indicates the contribution value of behavioral parameters ( k =1: Exercise intensity metabolic coefficient k =2: Nutrient absorption coefficient (e.g., nutrient absorption coefficient). , , All represent parameter weights, satisfying .
[0023] The LSTM neural network with short-term prediction layer contains 12 neurons in the input layer, 64 neurons in the hidden layer, and 1 neuron in the output layer. It iterates 500 times, has a learning rate of 0.001, and uses MSE as the loss function. Accuracy is verified using the MARD accuracy verification formula, specifically: ; in, express t The blood glucose value was measured after minutes; n represents the number of test samples, n≥100, and the samples must cover the blood glucose range of 1.1~33.3mmol / L and daily, exercise, high temperature and sleep scenarios; MARD is automatically calculated once every 24 hours. If MARD > 6.8%, a sensor calibration reminder is triggered. If it is still > 8% after calibration, it is determined that the prediction module is faulty and automatic infusion is stopped.
[0024] The intelligent infusion module includes an insulin pump and a single-lumen infusion line. The insulin pump uses a piezoelectric ceramic micropump, and the single-lumen infusion line is compatible with standard medical infusion needles, which can reduce patient discomfort during puncture.
[0025] The control flow of the intelligent infusion module is as follows: Step S201: Receive the predicted blood glucose value for the next 10 minutes output by the hierarchical prediction control module. ; Step S202: Calculate the insulin infusion rate based on the predicted blood glucose level and clinical safety standards. IR ; ; in, Indicates the basal insulin infusion rate; This represents the base rate correction factor; k This represents the hyperglycemic infusion adjustment factor.
[0026] Step S203: Drive the insulin pump to deliver insulin precisely according to the calculation results, and monitor the infusion pressure of the single-lumen infusion tubing in real time through a pressure sensor. P tude ; Step S204: If the infusion pressure is abnormal ( P tude If the pressure exceeds 50 kPa for 5 seconds, immediately stop the infusion and trigger the safety redundancy module alarm.
[0027] The safety redundancy module performs mutual checks through dual controllers. If an anomaly is detected, an audible and visual alarm is triggered, and insulin infusion is stopped. The safety redundancy module includes a dual controller mutual check unit, a sensor dual calibration unit, and an infusion blockage unit. The dual controller mutual check unit consists of a main controller and a backup controller, using an STM32 main / backup chip. Mutual checks are performed every 100ms, and the specific mutual check formula is as follows: ; in, This indicates the output value of the main controller; This indicates the output value of the standby controller; both the main controller output value and the standby controller output value include the insulin infusion rate. IR and predicted blood glucose levels ; DR Indicates the deviation rate; If DR≤2%: Normal operation, continue mutual testing at 100ms intervals; If 2%≤DR≤5%: trigger a yellow audible and visual alarm, automatically record the deviation log, continue running but increase the mutual inspection frequency to once every 10ms; If DR > 5%: trigger a red audible and visual alarm, immediately stop insulin infusion, switch to standby controller for independent operation, and prompt the user to contact maintenance via the user interaction module.
[0028] The sensor's dual calibration unit performs cross-calibration with the electrochemical sensor every 15 minutes via near-infrared spectroscopy (wavelength 1550nm). This is done when there is a deviation in blood glucose detection. Correction for >0.2 mmol / L ( These are infrared spectral detection values. (The value is detected by electrochemical sensing); the infusion blockage detection unit monitors the tubing pressure through a pressure sensor (accuracy ±1kPa). When the tubing pressure is greater than 50kPa and lasts for 5 seconds, it is determined to be a tubing blockage, and insulin infusion is immediately stopped and an alarm is triggered.
[0029] The data encryption transmission module adopts the Bluetooth 5.3 protocol, with a transmission rate of 2Mbps and a communication distance of ≤10m. It integrates the AES-256 encryption algorithm (the key is automatically updated every 24 hours). At the same time, it presets parameter thresholds (blood glucose <1.1 or blood glucose >33.3mmol / L, temperature <-10 or >40℃). Abnormal data exceeding the threshold is intercepted in real time and triggers a local alarm. After interception, manual confirmation is required before transmission can be resumed.
[0030] The user interaction module includes a 2.4-inch IPS touchscreen display (320×240 resolution, sunlight visible), an audible and visual alarm unit (80dB buzzer + tri-color LED), a historical data query unit (supports exporting data from the past 30 days to PDF / Excel), and an emergency control button (press and hold for 2 seconds to pause infusion / start emergency rescue). The audible and visual alarm unit supports three tones (hypoglycemia: low-frequency buzzer + flashing green LED; hyperglycemia: medium-frequency buzzer + flashing yellow LED; device malfunction: high-frequency buzzer + solid red LED), and automatically reduces alarm sensitivity during nighttime sleep (triggered only when blood glucose < 3.5 mmol / L).
[0031] The workflow is as follows: System initialization (after boot): Step 1: Self-test of each module (sensor, controller, pump body, communication). The self-test takes ≤10 seconds. If any module fails, a red alarm will be triggered and the faulty module will be displayed. Step 2: The user enters basic information (weight, medication concentration, target blood glucose range), and the system loads historical data (last 7 days). Step 3: Sensor preheating and calibration (5 minutes for electrochemical sensor preheating, 1 minute for near-infrared spectroscopy calibration). After calibration, the sensor enters standby mode, waiting for the user to start the infusion.
[0032] Standard blood glucose management procedure: Step 1 (Data Acquisition and Encryption): The multimodal data acquisition module acquires 12 multimodal parameters per second. After being encrypted by the data encryption transmission module AES-256, the data is transmitted to the hierarchical predictive control module via the CAN bus. At the same time, the module monitors in real time whether the data exceeds the abnormal threshold. If it does, the module intercepts the data and issues an alarm. Step 2 (Hierarchical Prediction): The short-term prediction layer performs a blood glucose prediction every minute and outputs the blood glucose curve for the next 30 minutes; the medium- and long-term adjustment layer adjusts the basal infusion rate every hour based on historical data; and identifies the current scenario type every 500ms to determine whether to trigger extreme scenario logic. Step 3 (Intelligent Infusion): The intelligent infusion module calculates the insulin infusion rate based on the layer scenario type and the short-term prediction layer prediction results, and controls the precise infusion of insulin. The response time is ≤2 seconds in normal scenarios and ≤5 seconds in extreme scenarios. Step 4 (Safety Monitoring): The safety redundancy module works synchronously, the dual controller mutual inspection unit checks the controller every 100ms, the sensor dual calibration unit calibrates the sensor every 15 minutes, the infusion blockage unit monitors the pipeline pressure in real time, and any abnormality in any link is handled according to the preset logic (alarm / stop infusion). Step 5 (User Interaction): The user interaction module displays blood glucose and infusion status in real time, supports users to query historical data and manually adjust the scene, triggers audible and visual alarms in case of abnormality, and allows users to pause the infusion or start emergency rescue via the emergency button.
[0033] Example 1: Routine eating scenario (adult type 1 diabetic patient, weight 60kg).
[0034] Scenario parameters: Breakfast intake of 50g carbohydrates, 10g protein, and 5g fat; ambient temperature of 25℃, humidity of 50%, altitude of 100m, resting state (MET=1.2); user's target blood glucose range of 4.4~10.0mmol / L; initial basal infusion rate set to 0.9U / h.
[0035] Module operation process: Multimodal data acquisition: The multimodal data acquisition module collects parameters every second, among which the physiological parameter unit measures the current blood glucose level. =6.5mmol / L, residual insulin 22U, heart rate 75 bpm, cortisol concentration 9μg / dL; environmental parameter unit measured temperature 24℃, altitude 150m, humidity 55%; behavioral parameter unit obtained the nutritional composition ratio of food intake (carbohydrate 70.6%, protein 17.6%, fat 11.8%) through the nutritional composition input interface, exercise intensity MET=1.3.
[0036] Hierarchical predictive control: The short-term prediction layer calculates the predicted blood glucose value based on an LSTM neural network (12 neurons in the input layer and 64 neurons in the hidden layer). Substituting this value into the blood glucose prediction formula of the short-term prediction layer, the final output is: G pred (30) = 10.2 mmol / L, G pred (10) = 8.8 mmol / L; The mid-to-long-term adjustment layer calls the blood glucose data after lunch for the past 7 days from the embedded database (capacity 16GB) and corrects the baseline infusion rate to 0.95 U / h (correction magnitude +5.6%).
[0037] Intelligent infusion: The intelligent infusion module receives the prediction results and calculates the infusion rate. , IR = ; Safety monitoring: The mutual inspection deviation obtained from the dual controller mutual inspection unit is 2% ≤ 5%, which is normal; The dual calibration unit of the sensor indicates a calibration deviation of 0.08 mmol / L ≤ 0.1 mmol / L, which is normal. The infusion blockage unit indicates a pipeline pressure of 28 kPa ≤ 50 kPa, which is normal. Result: After 30 minutes G actual =9.5mmol / L, prediction error is 3.1%, blood glucose is within the target range (4.4~10.0mmol / L), no abnormal alarm.
[0038] Example 2: Strenuous exercise scenario (adult type 2 diabetic patient, weight 70kg).
[0039] The difference from Example 1 is: The stratified prediction unit triggers extreme motion scenarios; the intelligent infusion module determines the insulin infusion rate to be 0; the dual-controller mutual detection unit increases the shoulder support frequency to once every 10 mms, with a deviation of 3% ≤ 5%; the infusion blockage unit determines the tubing pressure to be 32 kPa ≤ 50 kPa; after 30 minutes... G actual =4.2mmol / L, successfully avoiding hypoglycemia (target >3.9mmol / L). 20 minutes after exercise ended, the scenario was resolved, and regular infusion was gradually resumed.
[0040] Therefore, the present invention adopts the above-mentioned insulin pump control system, which aims to solve the technical problems of incomplete parameter coverage and weak prediction ability of existing systems, and realize accurate blood glucose prediction, dynamic adaptive infusion and full-link data security management.
[0041] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.
Claims
1. An insulin pump control system, characterized in that: It includes a multimodal data acquisition module, a hierarchical predictive control module, an intelligent infusion module, a security redundancy module, a data encryption transmission module, and a user interaction module; The multimodal data acquisition module simultaneously collects physiological parameters, environmental parameters, and behavioral parameters. The hierarchical prediction control module outputs predicted blood glucose values based on an LSTM neural network. The intelligent infusion module dynamically adjusts the insulin infusion rate according to the prediction results. The safety redundancy module performs mutual checks through dual controllers. If an abnormality is detected, an audible and visual alarm is triggered, and insulin infusion is stopped.
2. The insulin pump control system according to claim 1, characterized in that: The hierarchical prediction and control module includes a short-term prediction layer and a medium-to-long-term adjustment layer. The medium-to-long-term adjustment layer stores 7 days of historical data in an embedded database and adjusts the basic infusion rate daily. The blood glucose prediction formula for the short-term prediction layer is as follows: ； in, Indicates the future t Predicted blood glucose levels within minutes; This indicates the current measured blood glucose level; Indicates the contribution value of physiological parameters; Indicates the contribution value of environmental parameters; This represents the contribution value of the behavioral parameter; , , All represent parameter weights, satisfying .
3. The insulin pump control system according to claim 2, characterized in that: The intelligent infusion module includes an insulin pump and a single-chamber infusion line. The insulin pump is a piezoelectric ceramic micropump, and the single-chamber infusion line is compatible with standard medical infusion needles.
4. An insulin pump control system according to claim 3, characterized in that: The control flow of the intelligent infusion module is as follows: Step S201: Receive the predicted blood glucose value for the next 10 minutes output by the hierarchical prediction control module. ; Step S202: Calculate the insulin infusion rate based on the predicted blood glucose level and clinical safety standards. IR ; Step S203: Drive the insulin pump to deliver insulin precisely according to the calculation results, and monitor the infusion pressure of the single-lumen infusion tubing in real time through a pressure sensor. P tude ; Step S204: If the infusion pressure is abnormal, immediately stop the infusion and trigger the safety redundancy module alarm.
5. An insulin pump control system according to claim 4, characterized in that: The insulin infusion rate in step S202 is: ； in, This indicates the basal insulin infusion rate; This represents the base rate correction factor; k This represents the hyperglycemic infusion adjustment factor.
6. An insulin pump control system according to claim 5, characterized in that: The safety redundancy module includes a dual-controller mutual detection unit, a sensor dual calibration unit, and an infusion blocking unit. The dual-controller mutual detection unit includes a main controller and a backup controller, using an STM32 main / backup chip, and performs mutual detection every 100ms. The specific mutual detection formula is as follows: ； in, This indicates the output value of the main controller; This indicates the output value of the standby controller; both the main controller output value and the standby controller output value include the insulin infusion rate. IR and predicted blood glucose levels ; DR Indicates the deviation rate; If DR≤2%: Normal operation, continue mutual testing at 100ms intervals; If 2%≤DR≤5%: trigger a yellow audible and visual alarm, automatically record the deviation log, continue running but increase the mutual inspection frequency to once every 10ms; If DR > 5%: trigger a red audible and visual alarm, immediately stop insulin infusion, switch to standby controller for independent operation, and prompt the user to contact maintenance via the user interaction module.
7. An insulin pump control system according to claim 6, characterized in that: The sensor dual calibration unit performs cross-calibration every 15 minutes using near-infrared spectroscopy and electrochemical sensing, correcting when the deviation between the two exceeds 0.2 mmol / L; the infusion obstruction detection unit monitors the infusion pressure of the single-lumen infusion tubing using a pressure sensor. P tude ,when P tude If the pressure is greater than 50 kPa and lasts for 5 seconds, it is considered a tubing blockage, and insulin infusion should be stopped immediately and an alarm should be triggered.
8. An insulin pump control system according to claim 7, characterized in that: The data encryption transmission module adopts the Bluetooth 5.3 protocol, with a transmission rate of 2Mbps and a communication distance of ≤10m, and integrates the AES-256 encryption algorithm; at the same time, it presets parameter thresholds, and abnormal data exceeding the thresholds are intercepted in real time and trigger local alarms.
9. An insulin pump control system according to claim 8, characterized in that: The user interaction module includes a 2.4-inch IPS touchscreen display, an audible and visual alarm unit, a historical data query unit, and an emergency control button; the audible and visual alarm unit supports three tones: Hypoglycemia: Low-frequency beep + flashing green LED; High blood sugar: Medium-frequency beeping + flashing yellow LED; Equipment malfunction: High-frequency buzzer + constantly lit red LED; During nighttime sleep, the alarm sensitivity is automatically reduced, and the hypoglycemia alarm is only triggered when blood glucose is <3.5mmol / L.
10. An insulin pump control system according to claim 9, characterized in that: The LSTM neural network of the short-term prediction layer contains 12 neurons in the input layer, 64 neurons in the hidden layer, and 1 neuron in the output layer. It iterates 500 times, has a learning rate of 0.001, and uses MSE as the loss function. Accuracy is verified using the MARD accuracy verification formula, specifically: ； in, express t The blood glucose value was measured after minutes; n represents the number of test samples, n≥100, and the samples must cover the blood glucose range of 1.1~33.3mmol / L and daily, exercise, high temperature and sleep scenarios; MARD is automatically calculated once every 24 hours. If MARD > 6.8%, a sensor calibration reminder is triggered. If it is still > 8% after calibration, it is determined that the prediction module is faulty, automatic infusion is stopped, and the system is switched to manual infusion mode, prompting the user to contact technical personnel.