Intelligent insulin closed-loop infusion system for type 1 diabetes

Abstract

The invention provides an intelligent insulin closed-loop infusion system for type 1 diabetes mellitus, and relates to the technical field of diabetes mellitus treatment equipment. The system comprises a blood glucose monitoring module, a control module, an insulin infusion module, a user interaction module, a wireless communication module and a power module, and closed-loop precise infusion of insulin is achieved through cooperation of the multiple modules. Wherein a closed-loop algorithm based on a dynamic basic rate formula is arranged in the control module, and the insulin infusion amount can be dynamically adjusted in combination with an LSTM blood glucose prediction model and a metabolic parameter least square method updating unit; the insulin infusion module adopts a micro peristaltic pump controlled by PWM (Pulse Width Modulation) and a current monitoring type blockage detection unit, so that the infusion precision and safety are guaranteed. According to the system, the blood glucose standard reaching time proportion can be increased to 78%, the hypoglycemia occurrence frequency is reduced to 0.8 times per week, the system is remarkably superior to a traditional insulin pump and a simple closed-loop system, meanwhile, the system has the advantages of being long in sensor service life and low in power consumption, and the treatment convenience and safety of type 1 diabetes patients are improved.

Description

Technical Field

[0001] This invention belongs to the field of diabetes treatment equipment technology, and more specifically, relates to an intelligent closed-loop insulin infusion system for type 1 diabetes. Background Technology

[0002] Type 1 diabetes is a chronic metabolic disease caused by the absolute deficiency of insulin due to the failure of pancreatic β-cell function. Patients need to rely on exogenous insulin infusion for life to maintain stable blood glucose levels. If blood glucose is poorly controlled, it can easily lead to serious complications such as retinopathy, nephropathy, and neuropathy, and even endanger life. Currently, the commonly used insulin infusion devices in clinical practice are mainly divided into two categories, but both have significant technical shortcomings and cannot meet the needs of patients for long-term precise treatment.

[0003] The first type is the traditional insulin pump, which only has the functions of fixed basal rate infusion and manual bolus infusion during meals, and lacks closed-loop control capability. Patients need to measure their blood glucose 4-6 times a day via finger prick (a cumbersome operation that is prone to infection), and then manually adjust the basal rate based on the measurement results. Not only can it not respond to blood glucose fluctuations in real time, but it is also prone to excessively high or low blood glucose due to human error. Clinical trials show that the blood glucose target is usually achieved only 60%-65% of the time with this type of device, and hypoglycemia occurs 2-3 times per week, seriously affecting the patient's quality of life and health safety.

[0004] The second type is the simplified closed-loop insulin system. While it can achieve closed-loop regulation of basal rate, it has three major shortcomings: First, it lacks blood glucose prediction capabilities, making it impossible to proactively avoid the risk of hypoglycemia; adjustments can only be made passively after blood glucose has dropped to a threshold. Second, metabolic parameters (such as basal rate) require manual calibration weekly and cannot be dynamically updated based on the user's daily diet, exercise, and metabolic status, leading to a decrease in regulation accuracy over time. Third, blockage detection relies on pressure sensors, with an accuracy rate of only 80%-85%, making it easy for catheter blockage to go undetected and cause hyperglycemia. Furthermore, the sensor replacement interval for both types of devices is generally 7-10 days, and the standby time is 3-5 days, with frequent device operation and charging further increasing the burden on patients.

[0005] In summary, existing insulin infusion devices have significant shortcomings in terms of blood glucose control accuracy, safety, intelligence, and user convenience. There is an urgent need for an intelligent infusion system with complete closed-loop control, early risk warning, automatic parameter updates, reliable safety protection, and low operational burden to address the long-term treatment challenges faced by patients with type 1 diabetes. Summary of the Invention

[0006] To address the aforementioned technical problems, this invention provides an intelligent closed-loop insulin infusion system for type 1 diabetes, which solves the technical problems of existing insulin infusion devices for type 1 diabetes, such as low blood glucose control accuracy, high risk of hypoglycemia, unreliable detection of infusion blockage, need for manual adjustment of metabolic parameters, frequent sensor replacement, and short standby time leading to a heavy user workload.

[0007] A smart insulin closed-loop infusion system for type 1 diabetes includes:

[0008] The blood glucose monitoring module is used to collect the user's blood glucose value G(t) in real time, with a sampling interval of 1-5 minutes and a collection accuracy of ≤±10mg / dL;

[0009] The control module has a built-in closed-loop control algorithm. The algorithm dynamically calculates the insulin infusion volume based on the real-time blood glucose value G(t), the user's basal metabolic parameters and historical blood glucose data, including the dynamic basal rate B(t) and the bolus dose D_meal.

[0010] The dynamic base rate B(t) is calculated according to the formula: B(t)=B0+k1·ΔG(t)+k2·dG / dt;

[0011] In the formula, B0 is the user's initial basal rate, ΔG(t)=G(t)-G_target (G_target is the target blood glucose value, which is 3.9-7.2mmol / L), dG / dt is the rate of change of blood glucose, and k1 and k2 are adaptive adjustment coefficients;

[0012] The insulin infusion module communicates with the control module and infuses insulin with an accuracy of 0.05-0.1 U / step according to the infusion volume command output by the control module.

[0013] Preferably, the blood glucose monitoring module includes a minimally invasive blood glucose sensor and a signal processing unit;

[0014] The minimally invasive blood glucose sensor has an insertion depth of 3-5mm and can work continuously for 7-14 days.

[0015] The signal processing unit performs noise reduction processing on the raw blood glucose signal collected by the sensor using a Kalman filter algorithm. The filtering formula is as follows:

[0016]

[0017] In the formula, Let t be the filtered blood glucose value, A be the state transition matrix, K(t) be the Kalman gain, Z(t) be the original blood glucose signal at time t, and H be the observation matrix.

[0018] Preferably, the control module further includes a blood glucose prediction unit, which is based on a long short-term memory network (LSTM) model. It takes as input the blood glucose sequence [G(t-29), G(t-28), ..., G(t)] within the last 30 minutes and outputs the predicted blood glucose value G_pred(t+T) for the next 15-30 minutes (T = 15, 30 minutes).

[0019] When G_pred(t+T) < 3.9 mmol / L, the control module triggers a hypoglycemia protection mechanism, reducing the insulin infusion rate by 30%-50%.

[0020] Preferably, the insulin infusion module includes a drug storage tank, a micro peristaltic pump, and an infusion tubing;

[0021] The infusion rate of the micro peristaltic pump is controlled by a pulse width modulation (PWM) signal. The pump body operates at a voltage of 3.7V and has a standby power consumption of ≤10μA.

[0022] The infusion catheter has an inner diameter of 0.2-0.3 mm and is equipped with an anti-backflow valve at the end of the catheter.

[0023] Preferably, the control module also includes a built-in metabolic parameter update unit. This unit adjusts the initial basal rate B0 daily based on the user's blood glucose data, insulin infusion data, and dietary records from the previous 24 hours using the least squares method. The adjustment formula is as follows:

[0024] B0_new=B0_old+λ·∑(t=1to24)[(G(t)-G_target)·B(t)] / 24, where λ is the correction coefficient, taking a value of 0.05-0.1, B0_old is the base rate before correction, and B0_new is the base rate after correction.

[0025] Preferably, the calculation of the high-dose D_meal during the meal satisfies the following formula:

[0026] D_meal=D0·(1+γ·Carb-δ·Exe), where D0 is the standard mealtime dose (calculated based on 0.5-1U insulin per 10g of carbohydrates), Carb is the current meal carbohydrate intake (unit: 10g), Exe is the user's exercise volume in the past hour (calculated based on exercise duration with a cadence >100 steps / minute, unit: 30 minutes), and γ and δ are adjustment coefficients, γ=0.1-0.2, δ=0.08-0.15.

[0027] Preferably, it also includes a user interaction module, which includes a display screen and physical buttons. The display screen can show real-time blood glucose values, predicted blood glucose curves and remaining insulin levels. The physical buttons allow users to manually input the amount of carbohydrates in their meals and the amount of exercise, and a long press for 3 seconds can trigger an emergency pause infusion function.

[0028] Preferably, the closed-loop control algorithm of the control module also introduces a blood glucose fluctuation weighting coefficient W(t), where W(t) = 1.5 when |G(t) - G_target| > 2.2 mmol / L;

[0029] When |G(t)-G_target|≤2.2mmol / L, W(t)=1.0;

[0030] At this point, the dynamic base rate B(t) is corrected to:

[0031] B(t) = W(t)·(B0+k1·ΔG(t)+k2·dG / dt) is used to improve the sensitivity of regulation when blood glucose deviates from the target value.

[0032] Preferably, it also includes a wireless communication module, which supports Bluetooth 5.0 or LoRa protocol and can upload real-time blood glucose data and insulin infusion data to a cloud server;

[0033] The cloud server can generate weekly blood glucose reports for users, including the percentage of time blood glucose is within the target range (target: >70%) and the number of times hypoglycemia occurs (target: <1 time / week).

[0034] Preferably, the insulin infusion module is further provided with a blockage detection unit, which determines whether the infusion catheter is blocked by monitoring the operating current I(t) of the micro peristaltic pump;

[0035] When I(t) > I0 + 2σ (I0 is the average normal operating current and σ is the standard deviation of current) and lasts for 5 seconds, a blockage alarm is triggered, and a fault message is displayed through the user interaction module.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] First, it significantly improves the accuracy and stability of blood glucose control. This system, through a complete closed-loop algorithm built into the control module, combined with a dynamic basal rate formula and a mealtime bolus adjustment formula, can respond to blood glucose changes in real time. Simultaneously, with the LSTM blood glucose prediction model, it can predict blood glucose trends 15-30 minutes in advance, proactively reducing the insulin dose before hypoglycemia occurs. Clinical trial data shows that this system achieves a blood glucose target achievement rate of 78%, an improvement of 16 percentage points compared to traditional insulin pumps and 10 percentage points compared to simplified closed-loop systems; the frequency of hypoglycemia is reduced to 0.8 times / week, a decrease of 1.5 times / week compared to traditional insulin pumps and 0.7 times / week compared to simplified closed-loop systems, effectively reducing the risk of complications.

[0038] Secondly, it significantly improves infusion safety and reliability. On the one hand, the insulin infusion module uses a miniature peristaltic pump controlled by PWM signals, achieving an infusion accuracy of 0.05-0.1U / step and an infusion error of ≤±5%, which is 5 percentage points higher than traditional insulin pumps, avoiding blood glucose fluctuations caused by inaccurate dosage. On the other hand, the system innovatively designs a current monitoring-type blockage detection unit, which triggers an alarm by judging the working current of the peristaltic pump, achieving a blockage detection accuracy of 100%. This solves the problem of traditional equipment having no blockage detection or inaccurate detection, while avoiding the defects of pressure sensors being easily affected by temperature and insulin viscosity, ensuring safe and uninterrupted infusion.

[0039] Furthermore, it significantly improves user convenience and tolerance. The system's minimally invasive blood glucose sensor can operate continuously for 14 days, doubling the usage time compared to traditional devices. The power module features a low-power design, providing up to 7 days of standby time, 4 days longer than traditional devices, greatly reducing the frequency of sensor replacement and device charging. In addition, the user interaction module supports one-click input of meal carbohydrates and exercise levels, with a 3-second long press triggering an emergency pause. The cloud module automatically generates weekly blood glucose reports, eliminating the need for manual recording and analysis by the user, lowering the operational threshold, and making it especially suitable for long-term use by elderly or mobility-impaired type 1 diabetes patients.

[0040] Finally, it possesses excellent adaptability and scalability. The metabolic parameter update unit of the control module automatically corrects the initial basal rate daily using the least squares method, eliminating the need for manual calibration by the user and adapting to differences in weight, disease duration, and metabolic status among different patients. The wireless communication module simultaneously supports Bluetooth 5.0 and LoRa protocols, enabling flexible integration with mobile apps and community healthcare cloud platforms. In the future, it can be further integrated with artificial intelligence diagnostic functions to provide patients with personalized treatment plans, demonstrating broad prospects for clinical application. Attached Figure Description

[0041] Figure 1 This is a schematic diagram of the system composition of the present invention. Detailed Implementation

[0042] The embodiments of the present invention will be described in further detail below with reference to the accompanying drawings and examples. The following examples are for illustrative purposes only and should not be construed as limiting the scope of the invention.

[0043] Please see Figure 1 This invention provides an intelligent insulin closed-loop infusion system for type 1 diabetes. The following detailed description of the intelligent insulin closed-loop infusion system is based on a specific implementation scenario using an adult type 1 diabetes patient (60kg, 5-year disease duration, no serious complications). The hardware selection, parameter settings, and algorithm execution flow of each module of the system all meet the reproducibility requirements, ensuring that those skilled in the art can implement this invention based on this embodiment.

[0044] The intelligent insulin closed-loop infusion system in this embodiment includes six core modules: a blood glucose monitoring module, a control module, an insulin infusion module, a user interaction module, a wireless communication module, and a power supply module. The modules communicate with each other via wired (FPC cable between the control module and the infusion module) or wireless (Bluetooth 5.0 between the control module and the cloud), forming a complete closed-loop control link. The core hardware selection and functional correspondence of each module are as follows:

[0045] The blood glucose monitoring module uses an enzyme electrode-type minimally invasive sensor combined with an STM32L431 signal processing chip. The control module is equipped with a Snapdragon W5100 processor (with a neural network acceleration unit). The insulin infusion module is equipped with a piezoelectric-driven micro peristaltic pump (model: PumpMini-05) and a 3mL drug reservoir. The user interaction module includes a 1.2-inch OLED display and three physical buttons. The wireless communication module integrates a Bluetooth 5.0 chip (nRF52840) and a LoRa module (SX1278). The power module uses a 3.7V / 1000mAh lithium polymer battery (supporting wireless charging) and can achieve a standby time of 7 days.

[0046] In the implementation of the blood glucose monitoring module, an enzyme electrode-type minimally invasive sensor is combined with an STM32L431 signal processing chip to achieve high-precision real-time acquisition and noise reduction. The sensor is inserted subcutaneously in the abdomen, with the insertion depth controlled at 3.5mm to avoid areas with dense blood vessels. It is fixed with medical adhesive and can work continuously for 14 days based on the principle of glucose oxidase without frequent replacement. The sampling interval is set to 3 minutes / sampling, with an acquisition accuracy of ±8mg / dL. At the same time, the signal processing unit (STM32L431) performs Kalman filtering on the raw blood glucose signal output by the sensor to eliminate interference from body temperature and movement. The filtering parameters are specifically set as follows: state transition matrix A = 1.02 (reflecting the slow change characteristics of blood glucose), observation matrix H = 1.0 (linear correlation between the raw signal and blood glucose value), and Kalman gain K(t) initially 0.3, which is dynamically adjusted with the number of samplings. The smaller the error, the closer K(t) is to 0.1.

[0047] For example, when t = 10 minutes, the original signal Z(t) = 8.5 mmol / L, and the filtered value at the previous moment... Substitute into the filtering formula The calculation yields:

[0048] It can effectively filter out interference fluctuations of ±0.3 mmol / L.

[0049] The control module, acting as the system's brain, runs a closed-loop algorithm based on the Snapdragon W5100 processor. The initial values ​​of the core parameters are set according to the user's basic information (weight 60kg, daily carbohydrate intake 200g / day). The specific algorithm execution flow is as follows:

[0050] In the dynamic basal rate calculation process, the dynamic basal rate B(t), as the "basal dose" of insulin infusion, needs to be adjusted according to the real-time blood glucose deviation, rate of change, and fluctuation weight. First, the initial parameters are set. The initial basal rate B0 is set to 0.5U / kg / day, or 30U / day, based on the user's weight, which is 1.25U / hour after being distributed over 24 hours.

[0051] The adjustment coefficients k1 = 0.02 U / (mmol·L) (used for adjusting the blood glucose level when it deviates from the target value), and k2 = 0.01 U / (mmol·L·min) (used for adjusting the rate of change of blood glucose).

[0052] The target blood glucose level G_target is set at 5.6 mmol / L, which falls within the range of 3.9-7.2 mmol / L;

[0053] The fluctuation weighting coefficient W(t) is adjusted according to the degree of blood glucose deviation from the target value. When |G(t)-5.6|>2.2mmol / L (i.e. G(t)<3.4 or>7.8mmol / L), W(t)=1.5, and otherwise W(t)=1.0.

[0054] Taking t=14:00 as an example, the real-time blood glucose G(t) at this time is 7.9 mmol / L (deviating from the target value of 2.3 mmol / L), the rate of change of blood glucose dG / dt is 0.1 mmol / L·min (showing an upward trend), and |G(t)-G_target|>2.2 mmol / L, so W(t)=1.5. Substituting into the formula B(t)=W(t)·(B0+k1·ΔG(t)+k2·dG / dt), where ΔG(t)=7.9-5.6=2.3mmol / L, we can calculate B(t)=1.5×(1.25+0.02×2.3+0.01×0.1)=1.5×1.297≈1.95U / hour, thus achieving precise adjustment of the basal rate when blood glucose is high.

[0055] For the functions of blood glucose prediction and hypoglycemia protection, the control module has a built-in LSTM blood glucose prediction model. The input layer of the model has 30 neurons (corresponding to the blood glucose sequence of the past 30 minutes, i.e., 10 sampling points, 1 every 3 minutes), the hidden layer has 2 layers with 64 neurons in each layer (the activation function is ReLU), and the output layer has 2 neurons (corresponding to the predicted blood glucose values ​​G_pred(15) and G_pred(30) for the next 15 minutes and 30 minutes). The model training data comes from the multi-dimensional data of blood glucose-insulin-diet-exercise of 1000 patients with type 1 diabetes (lasting 6 months).

[0056] When t = 22:00, if the model predicts that the blood glucose G_pred(15) will be 3.8 mmol / L (< 3.9 mmol / L) in the next 15 minutes, the control module will immediately trigger the hypoglycemia protection mechanism, reduce the current dynamic basal rate (originally 1.2 U / hour) by 40% to 0.72 U / hour, and display a "hypoglycemia risk" prompt through the interactive module to avoid the harm of hypoglycemia in advance.

[0057] Regarding the updating of metabolic parameters, the control module automatically updates the initial basal rate B0 at 2:00 AM every day (when the user is at rest). The correction formula parameter is set to a correction coefficient λ = 0.08 (to avoid parameter mutations). The data sample consists of 288 blood glucose data points (one every 5 minutes) and 24 basal rate data points (one every hour) from the previous 24 hours.

[0058] For example, before the correction, B0_old = 1.25 U / hour, and for the first 24 hours, ∑[(G(t)-G_target)·B(t)] = 12.6. Substituting into the formula B0_new = B0_old + λ·∑(...) / 24, we can calculate B0_new = 1.25 + 0.08 × 12.6 / 24 ≈ 1.29 U / hour, making the basal rate more closely match the user's daily metabolic state.

[0059] In the bolus meal calculation step, the user inputs the amount of carbohydrates in their meal via an interactive module before eating. The control module then calculates the bolus meal dose (D_meal) based on a preset formula. Initial parameters are set as follows: standard bolus dose D0 = 0.8U / 10g carbohydrates (based on the user's daily insulin sensitivity), adjustment coefficients γ = 0.15 (correction margin for carbohydrate intake) and δ = 0.12 (correction margin for exercise intensity). Exercise intensity (Exe) is measured in 30-minute increments of 100 steps / minute. If the user walks for 30 minutes (110 steps / minute) one hour before a meal and inputs a carbohydrate intake of "50g" (i.e., 5 × 10g), then Exe = 1. Substituting this into the formula D_meal = D0·(1 + γ·Carb - δ·Exe), the result is:

[0060] D_meal=0.8×(1+0.15×5-0.12×1)=0.8×1.63≈1.3U. This dose is infused in three separate infusions with meals (60% before meals, and 20% at 15 minutes and 30 minutes after meals) to avoid a sudden rise in blood glucose after meals.

[0061] The insulin infusion module uses a piezoelectrically driven micro-peristaltic pump as its core, combined with a 3mL reservoir and infusion tubing, to achieve precise infusion and blockage detection. The reservoir contains rapid-acting insulin (such as aspart insulin), and the remaining amount is monitored in real time by a resistive sensor.

[0062] Infusion accuracy is achieved by controlling the pump speed with a PWM signal. The PWM frequency is set to 1kHz, and the duty cycle is 0-100%, corresponding to an infusion volume of 0-0.1U / step. The minimum infusion unit reaches 0.05U, which meets the accuracy requirements.

[0063] The infusion catheter has an inner diameter of 0.25 mm and a built-in one-way silicone anti-backflow valve at the end, allowing only insulin to flow out and effectively preventing blood backflow from clogging the catheter. For clogging detection, the clogging detection unit determines the catheter status by monitoring the peristaltic pump's operating current. The normal operating current parameters are set as follows: mean I0 = 50 mA, standard deviation σ = 5 mA. When the operating current I(t) > I0 + 2σ = 60 mA and lasts for 5 seconds (to avoid transient current interference), catheter clogging is detected. At this point, the pump immediately stops infusion, the interactive module displays a red "Catheter Clogging" warning, and a buzzer sounds an alarm at 1 kHz / 0.5-second intervals until the user replaces the catheter, ensuring infusion safety and conforming to the design of claim 10.

[0064] The user interaction module consists of a 1.2-inch OLED display and three physical buttons. The display is divided into three areas to show information:

[0065] The upper section displays the real-time blood glucose value (e.g., 7.3 mmol / L, in green font), the middle section displays the predicted blood glucose curve for the next 30 minutes (horizontal axis is time, vertical axis is blood glucose value), and the lower section displays the remaining insulin amount (e.g., 2.1 mL) and infusion status (e.g., "normal").

[0066] The physical buttons have clearly defined functions: the "confirm" button is used to confirm the input amount of carbohydrates or exercise, the "input" button is used to switch between digital input of carbohydrates and exercise, and the "emergency" button can be pressed and held for 3 seconds to trigger the pause function and pressed briefly to cancel the alarm, thus meeting the interactive needs.

[0067] The wireless communication and cloud module supports Bluetooth 5.0 and LoRa protocols. Bluetooth 5.0 enables short-range communication between the control module and the mobile APP (communication distance ≤ 10 meters), while the LoRa module enables long-range communication between the control module and the community medical station cloud (communication distance ≤ 1 km). Real-time blood glucose data and insulin infusion data can be uploaded to the cloud server.

[0068] The cloud server generates a weekly blood glucose report for users. The report includes the percentage of time that blood glucose is within the target range (target > 70%) and the number of times hypoglycemia occurs (target < 1 time / week). Users can view the report through a mobile app to facilitate timely adjustments to their treatment plan.

[0069] Taking a typical user's day of "breakfast-work-dinner" as an example, the complete system workflow is as follows:

[0070] At 6:00 AM, the user inserts the blood glucose sensor into the abdomen. The system automatically calibrates the sensor (compared to the finger prick blood glucose value, with an error of ≤5%), loads the initial basal rate B0 = 1.29 U / hour updated the previous day, enters standby mode, and completes the power-on initialization.

[0071] At 7:00, the user prepares breakfast and inputs the carbohydrate amount "60g" through the interactive module. The system calculates D_meal = 1.5U according to the large dose calculation formula for meals and infuses it in 3 times (60% before the meal, and 20% each 15 minutes and 30 minutes after the meal).

[0072] Meanwhile, the blood glucose monitoring module collects blood glucose every 3 minutes, which is then transmitted to the control module after being filtered by Kalman. At 7:30, the blood glucose rises to 7.5 mmol / L, and the control module increases the dynamic basal rate to 1.8 U / hour to complete the breakfast meal control.

[0073] The user's working hours are from 9:00 to 17:00. At 10:00, the LSTM blood glucose prediction model built into the control module predicts that the blood glucose will drop to 3.8 mmol / L (<3.9 mmol / L) in 15 minutes. The hypoglycemic protection mechanism is immediately triggered, the current basal rate is reduced by 40%, and the interactive module prompts "replenish a small amount of carbohydrates".

[0074] At 2:00 PM, if a user walks for 20 minutes (step frequency 105 steps / minute) and manually inputs the exercise volume "0.67" (i.e., 20 minutes = 0.67 × 30 minutes), the system will lower the basal rate by 10% to achieve dynamic adjustment during the work period.

[0075] At 18:30, the user prepared dinner and entered the carbohydrate amount "40g". The system calculated D_meal = 1.1U and infused it with the meal. At 23:00, the infusion tubing became slightly blocked. The pump current rose to 62mA and lasted for 5 seconds. The system triggered a blockage alarm. After the user replaced the tubing, the system returned to normal.

[0076] At 2:00 AM the following day (when the user is at rest), the system automatically updates the initial basal rate B0 to 1.31 U / hour, completing parameter iteration and ensuring the accuracy of subsequent infusion doses.

[0077] The system described in this embodiment underwent a 4-week clinical trial in 20 adult patients with type 1 diabetes to verify its effectiveness. In terms of blood glucose control, the percentage of patients achieving target blood glucose levels increased from 62% before the trial to 78%, and the frequency of hypoglycemia (<3.9 mmol / L) decreased from 2.3 times / week to 0.8 times / week, both superior to similar products in the industry (typically achieving target blood glucose levels ≈70%, and hypoglycemia frequency ≈1.5 times / week).

[0078] In terms of infusion accuracy and safety, the insulin infusion error was ≤±5%, meeting the accuracy requirement of 0.05U / step, the blockage detection accuracy rate reached 100%, and there were no cases of severe hypoglycemia or hyperglycemia due to system failure during the trial.

[0079] From a user experience perspective, the sensor needs to be replaced every 14 days, and the system has a standby time of 7 days, which reduces the frequency of user operation. User satisfaction surveys show that the scores for "ease of operation" and "stability of blood sugar control" are both >4.5 / 5, indicating a good overall user experience.

[0080] To further verify the superiority of the system of the present invention, the following embodiments and comparative examples were set up, all of which were applied to 20 adult type 1 diabetic patients (weight 55-65kg, disease duration 3-7 years, no serious complications) in the same batch, with a trial period of 4 weeks, and the key performance indicators were compared and analyzed.

[0081] Example:

[0082] The intelligent insulin closed-loop infusion system described in this invention has the same hardware selection, parameter settings, and algorithm execution process as described in the specific embodiments above, and has functions such as complete closed-loop control, blood glucose prediction, automatic updating of metabolic parameters, and blockage detection.

[0083] Comparative Example 1:

[0084] It uses a traditional insulin pump (model: Minimed670G basic version), which only has the functions of basal rate infusion and manual mealtime bolus infusion. It has no closed-loop control and requires users to manually adjust the basal rate according to the fingertip blood glucose value (measured 4-6 times a day). It also has no blockage detection function.

[0085] Comparative Example 2:

[0086] The simplified closed-loop insulin system (model: Dexcom G7+Tandemt:sl imX2 basic closed-loop version) only achieves closed-loop control of basal rate, without blood glucose prediction function or automatic update function of metabolic parameters. The basal rate needs to be manually calibrated every week, and the blockage detection relies on a pressure sensor (with low accuracy).

[0087] Performance Indicator Comparison Analysis Table:

[0088]

[0089] As can be seen from the table above, the system of the present invention is significantly superior to traditional insulin pumps and simple closed-loop systems in terms of blood glucose control effect, safety, infusion accuracy, and user convenience: First, through a complete closed-loop control algorithm (including dynamic basal rate and mealtime bolus adjustment) and blood glucose prediction function, it greatly improves the blood glucose target achievement rate and reduces the risk of hypoglycemia; Second, the high-precision infusion pump and reliable blockage detection function ensure infusion safety and accuracy; Third, the long-life sensor and low power consumption design reduce the frequency of user operation, improve ease of use, and better meet the long-term use needs of type 1 diabetes patients.

[0090] In summary, this specific implementation method uses an adult type 1 diabetic patient weighing 60kg, with a disease duration of 5 years and no serious complications as the typical application target, fully presenting the implementation path of the intelligent insulin closed-loop infusion system of this invention. From the perspective of the overall system composition, the hardware selection (such as the SensPro-14 enzyme electrode minimally invasive sensor, Snapdragon W5100 processor, and PumpMini-05 piezoelectric driven micro-peristaltic pump, etc.) and functional division of the six core modules of blood glucose monitoring, control, insulin infusion, user interaction, wireless communication, and power supply are clearly defined. A complete closed-loop link of "acquisition-processing-decision-execution-feedback" is constructed through a combination of wired and wireless communication methods to ensure the coordinated operation of each module.

[0091] In terms of implementation details for each module, not only were the core technical parameters quantitatively defined (such as sensor insertion depth of 3.5mm, sampling interval of 3 minutes, infusion accuracy of 0.05-0.1U / step, and standby time of 7 days), but also reproducible execution logic was provided for key algorithms and functions: Kalman filtering achieves blood glucose signal noise reduction by dynamically adjusting the state transition matrix A and Kalman gain K(t); dynamic basal rate calculation introduces a fluctuation weight coefficient W(t) to improve the adjustment sensitivity when deviating from the target value; the LSTM blood glucose prediction model achieves 15-30 minute blood glucose prediction through a clear network structure (30 input neurons, 2 layers of 64-neuron hidden layers) and training strategy (Adam optimizer, loss function with low blood glucose risk weight); the metabolic parameter update unit automatically corrects the initial basal rate B0 daily based on the least squares method; and the blockage detection unit achieves 100% accuracy in blockage determination by monitoring the peristaltic pump operating current (I(t) > 60mA for 5 seconds). These detailed designs ensure the feasibility of the invention.

[0092] From the perspective of typical scenario workflows and clinical trial validation, taking the user's "breakfast-work-dinner" daily cycle as an example, the system's entire operation process, from initial startup, mealtime dose calculation, dynamic basal rate adjustment, hypoglycemia risk warning to nighttime parameter updates, is clearly demonstrated. A 4-week clinical trial data and comparative analysis of the example-comparison model further confirm that this system significantly outperforms traditional insulin pumps and simple closed-loop systems in key indicators such as the percentage of time blood glucose reaches target levels (78%), the number of hypoglycemic events (0.8 times / week), infusion accuracy (error ≤ ±5%), blockage detection accuracy (100%), and user convenience (sensor replacement every 14 days, 7-day standby). This fully verifies the technical advantages of this invention in improving blood glucose control accuracy, ensuring infusion safety, and reducing operational burden in patients with type 1 diabetes.

[0093] The value of this specific implementation method lies in providing detailed support for the implementation of the technical solution and providing a reproducible and adjustable implementation reference for those skilled in the art. For type 1 diabetes patients with different weights and disease courses, personalized adaptation can be achieved by adjusting parameters such as the initial basal rate B0 and adjustment coefficients k1 / k2 / γ / δ. At the same time, the system retains room for expansion in terms of hardware upgrades (such as replacing with higher precision sensors) and algorithm optimization (such as integrating AI personalized treatment models), laying the foundation for subsequent technology iteration and clinical promotion.

[0094] The embodiments of the present invention are given for illustrative and descriptive purposes only, and are not intended to be exhaustive or to limit the invention to the forms disclosed. Many modifications and variations will be apparent to those skilled in the art. The embodiments were chosen and described in order to better illustrate the principles and practical application of the invention, and to enable those skilled in the art to understand the invention and to design various embodiments with various modifications suitable for a particular purpose.

Claims

1. A smart insulin closed-loop infusion system for type 1 diabetes, characterized in that, include: The blood glucose monitoring module collects the user's blood glucose value G(t) in real time, with a sampling interval of 1-5 minutes; The control module has a built-in closed-loop control algorithm. The algorithm dynamically calculates the insulin infusion volume based on the real-time blood glucose value G(t), the user's basal metabolic parameters and historical blood glucose data, including the dynamic basal rate B(t) and the bolus dose D_meal. The dynamic base rate B(t) is calculated according to the formula: B(t)=B0+k1·ΔG(t)+k2·dG / dt; In the formula, B0 is the user's initial basal rate, ΔG(t)=G(t)-G_target, dG / dt is the rate of change in blood glucose, and k1 and k2 are adaptive adjustment coefficients; The insulin infusion module communicates with the control module and infuses insulin with an accuracy of 0.05-0.1 U / step according to the infusion volume command output by the control module.

2. The intelligent insulin closed-loop infusion system for type 1 diabetes as described in claim 1, characterized in that, The blood glucose monitoring module includes a minimally invasive blood glucose sensor and a signal processing unit; The insertion depth of the minimally invasive blood glucose sensor is 3-5 mm; The signal processing unit performs noise reduction processing on the raw blood glucose signal collected by the sensor using a Kalman filter algorithm. The filtering formula is as follows: In the formula, Let t be the filtered blood glucose value, A be the state transition matrix, K(t) be the Kalman gain, Z(t) be the original blood glucose signal at time t, and H be the observation matrix.

3. The intelligent insulin closed-loop infusion system for type 1 diabetes as described in claim 2, characterized in that, The control module also includes a blood glucose prediction unit, which is based on a long short-term memory network (LSTM) model. It takes the blood glucose sequence [G(t-29), G(t-28), ..., G(t)] within the last 30 minutes as input and outputs the predicted blood glucose value G_pred(t+T) for the next 15-30 minutes (T = 15, 30 minutes). When G_pred(t+T) < 3.9 mmol / L, the control module triggers a hypoglycemia protection mechanism, reducing the insulin infusion rate by 30%-50%.

4. The intelligent insulin closed-loop infusion system for type 1 diabetes as described in claim 3, characterized in that, The insulin infusion module includes a drug storage tank, a micro peristaltic pump, and an infusion tubing; The infusion rate of the micro peristaltic pump is controlled by a pulse width modulation (PWM) signal. The pump body operates at a voltage of 3.7V and has a standby power consumption of ≤10μA. The infusion catheter has an inner diameter of 0.2-0.3 mm and is equipped with an anti-backflow valve at the end of the catheter.

5. The intelligent insulin closed-loop infusion system for type 1 diabetes as described in claim 4, characterized in that, The control module also includes a built-in metabolic parameter update unit. This unit adjusts the initial basal rate B0 daily based on the user's blood glucose data, insulin infusion data, and dietary records from the previous 24 hours using the least squares method. The adjustment formula is as follows: B0_new=B0_old+λ·∑(t=1to24)[(G(t)-G_target)·B(t)] / 24, where λ is the correction coefficient, taking a value of 0.05-0.1, B0_old is the base rate before correction, and B0_new is the base rate after correction.

6. The intelligent insulin closed-loop infusion system for type 1 diabetes as described in claim 5, characterized in that, The calculation of the high-dose D_meal during meals satisfies the following formula: D_meal=D0·(1+γ·Carb-δ·Exe), where D0 is the standard meal dose, Carb is the current carbohydrate intake, Exe is the user's exercise volume in the past hour, and γ and δ are adjustment coefficients, γ=0.1-0.2, δ=0.08-0.

15.

7. The intelligent insulin closed-loop infusion system for type 1 diabetes as described in claim 6, characterized in that, It also includes a user interaction module, which includes a display screen and physical buttons. The display screen shows real-time blood glucose levels, predicted blood glucose curves, and remaining insulin levels. The physical buttons allow users to manually input the amount of carbohydrates in their meals and the amount of exercise. Pressing and holding the button for 3 seconds can trigger an emergency pause infusion function.

8. The intelligent insulin closed-loop infusion system for type 1 diabetes as described in claim 7, characterized in that, The closed-loop control algorithm of the control module also introduces a blood glucose fluctuation weighting coefficient W(t), which is 1.5 when |G(t)-G_target|>2.2mmol / L; When |G(t)-G_target|≤2.2mmol / L, W(t)=1.0; At this point, the dynamic base rate B(t) is corrected to: B(t) = W(t)·(B0+k1·ΔG(t)+k2·dG / dt) is used to improve the sensitivity of regulation when blood glucose deviates from the target value.

9. The intelligent insulin closed-loop infusion system for type 1 diabetes as described in claim 8, characterized in that, It also includes a wireless communication module that supports Bluetooth 5.0 or LoRa protocol to upload real-time blood glucose data and insulin infusion data to a cloud server; The cloud server generates weekly blood glucose reports for users, including the percentage of time blood glucose levels are within the target range and the number of times hypoglycemia occurs.

10. The intelligent insulin closed-loop infusion system for type 1 diabetes as described in claim 9, characterized in that, The insulin infusion module is also equipped with a blockage detection unit, which determines whether the infusion catheter is blocked by monitoring the operating current I(t) of the micro peristaltic pump. When I(t) > I0 + 2σ and lasts for 5 seconds, a blockage alarm is triggered, and a fault message is displayed through the user interaction module.

Images