Intelligent real-time blood detection system based on intravascular indwelling sensor

Abstract

According to the intelligent real-time blood detection system based on the intravascular indwelling sensors, through an intravascular indwelling multi-sensor integrated system (electrochemical / optical coaxial layered design), continuous monitoring of one-time puncture for more than or equal to 7 days is achieved, in a clinical test, puncture within 7 days can be reduced from 28 times to 3 times, an improved Kalman filtering algorithm eliminates signal crosstalk, and the detection accuracy is improved. The D-S evidence theory is combined to dynamically fuse multi-modal data, so that the blood glucose detection MARD is reduced to 4.1%, the blood potassium error is less than 0.08 mM, and the motion artifact residual error is 1t; and 0.5%. A real-time data transmission and intelligent alarm system significantly improves clinical response efficiency, self-adaptively calibrates individual differences, and realizes high-precision long-acting monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of blood detection, and specifically to an intelligent real-time blood detection system based on an intravascular indwelling sensor. Background Art

[0002] Existing blood detection relies on repeated intravenous punctures, resulting in a high risk of infection and patient pain. Laboratory tests are time-consuming and it is difficult to capture acute changes such as sudden intraoperative potassium elevation in a timely manner. Continuous Glucose Monitoring (CGM) only monitors subcutaneous tissue fluid glucose, which has a lag with blood glucose. Therefore, an intelligent real-time blood detection system based on an intravascular indwelling sensor is proposed. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide an intelligent real-time blood detection system based on an intravascular indwelling sensor, which can effectively solve the problems raised in the above background art.

[0004] To solve the above problems, the technical solution adopted by the present invention is: an intelligent real-time blood detection system based on an intravascular indwelling sensor, comprising: An intravascular sensing unit, which includes an electrochemistry sensor, an optical sensor and a catheter integrated coaxially and hierarchically, and is indwelled in the blood vessel through intravenous puncture. The electrochemistry sensor obtains the Faraday current corresponding to the blood biochemical index in real time through redox reaction, and the optical sensor detects the blood component concentration through the dual-wavelength near-infrared spectrum absorption characteristic; A data processing unit, which performs a detection method on the electrochemistry current signal and the optical absorbance signal, including the following steps: S1. Establish a three-dimensional state space model including blood glucose concentration G, potassium ion concentration K and motion artifact M, and fuse multi-source signals through an improved Kalman filtering algorithm. The state vector value is: Where is the blood glucose concentration, is the potassium ion concentration, is the motion artifact interference amount; The observation vector value is: Where, , is the current output of the electrochemistry sensor, is the absorbance value of the optical sensor, is the combined vector amplitude of the triaxial accelerometer; The update law is: Where represents the Kalman gain matrix, is the prior estimation error covariance matrix, is the observation noise covariance matrix, is the observation matrix. , is the sensor noise variance, is the blood glucose-current conversion coefficient, is the blood potassium-absorbance conversion coefficient; S2. Use the D-S evidence theory for diagnostic decision-making, and the combined confidence level is: where represents the hyperglycemic value, H is the hypothesis to be verified, is the value of the electrochemical sensor, is the value of the optical sensor, and the conflict coefficient is , and when is satisfied, an alarm is triggered, is the preset confidence threshold, = 0.85.

[0005] As a further preferred embodiment of the present invention, the state transition matrix is: where is the glucose metabolism rate constant; is the blood potassium excretion rate constant; is the motion attenuation coefficient; is the modulus of the acceleration vector.

[0006] As a further preferred embodiment of the present invention, the inner wall of the catheter includes a piezoelectric drive brush structure, and its amplitude control function is: is the maximum amplitude, is the damping coefficient, is the drive frequency, is the phase compensation angle.

[0007] As a further preferred embodiment of the present invention, the state equation is further included in the step S1: Predict physiological parameters, where is the state transition matrix, including glucose metabolism kinetics, the renal excretion rate of blood potassium; the time decay characteristics of motion artifacts; Observation equation: , is the observation matrix, generating a signal estimate.

[0008] Compared with the prior art, the present invention provides an intelligent real-time blood detection system based on an intravascular indwelling sensor, having the following beneficial effects: The present invention realizes continuous monitoring for ≥7 days with a single puncture through an intravascular indwelling multi-sensor integration system (electrochemical / optical coaxial hierarchical design). In clinical trials, the number of punctures within 7 days can be reduced from 28 to 3. The improved Kalman filtering algorithm eliminates signal crosstalk, and combines the D-S evidence theory to dynamically fuse multi-modal data, reducing the MARD of blood glucose detection to 4.1%, the error of blood potassium <0.08 mM, and the residual error of motion artifacts <0.5%. The real-time data transmission and intelligent alarm system significantly improve the clinical response efficiency, adaptively calibrate individual differences, and achieve high-precision long-term monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] Figure 1 It is a flowchart of real-time signal processing of the present invention; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0010] It should be noted that if "and / or" or "and / or" appears throughout the text, its meaning includes three parallel solutions. Taking "A and / or B" as an example, it includes solution A, or solution B, or the solution where A and B are satisfied simultaneously.

[0011] In addition, the technical solutions between various embodiments can be combined with each other, but it must be based on the ability of those of ordinary skill in the art to implement. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the protection scope required by the present invention.

[0012] As a specific embodiment 1 of the present invention: The working process of the present invention: 1. Puncture and implant the sensing unit → 2. Magnetically fix the external relay unit → 3. The sensor automatically samples every 5 minutes → 4. The data is uploaded to the cloud via LoRaWAN → 5. Abnormal values trigger an alarm at the medical staff end; Reference Figure 1 , the real-time signal processing process is: graph TD A[Original signal sampling] --> B[Digital filtering] B --> C[Motion artifact elimination] C --> D[Multi-sensor data fusion] D --> E[Concentration calculation] E --> F[Abnormal value marking] F --> G[Cloud transmission] Digital filtering stage: Adaptive Kalman filtering eliminates 50 / 60 Hz power frequency interference Wavelet transform (db4 wavelet) separates high-frequency noise and physiological signals Motion artifact elimination (for signal drift caused by patient movement): Input: Accelerometer data (XYZ three axes) Algorithm: RLS (recursive least squares) algorithm to establish a motion-signal interference model Multi-sensor data fusion: Using D-S evidence theory to integrate conflicting data (when the results of the blood glucose electrode and the optical sensor are inconsistent). Output confidence score (calibration is triggered when the threshold ≥ 0.85). Machine learning calibration algorithm Training data: 10,000 clinical blood samples (covering different BMIs and hematocrits). Online learning: Automatically update the model parameters after receiving the laboratory blood draw results (federated learning protects privacy). As a specific embodiment 2 of the present invention: The catheter is made of biocompatible polyurethane material, with an outer diameter of 1.5 mm. An electrochemical sensor is coaxially integrated inside the catheter. A platinum-silver chloride electrode pair is used, and glucose oxidase is immobilized on the surface. The working voltage is set to +0.35V vs Ag / AgCl. By detecting the Faraday current generated by the glucose oxidation reaction, the current-concentration conversion relationship is Where is the blood glucose concentration, is Gaussian white noise, with a standard deviation .

[0013] The outer layer is coaxially coated with an optical sensor module. A dual-wavelength near-infrared light source of 850 nm and 1300 nm is used. The characteristic absorbance of potassium ions at 1300 nm is detected by a photodiode. The concentration conversion formula is: Where is the potassium ion concentration, is optical noise, with a standard deviation .

[0014] A three-axis MEMS accelerometer is integrated at the end of the catheter. The motion artifact quantization formula is: The data fusion algorithm performs the following steps: 1. State space modeling. The state vector is defined as , where is the motion artifact interference amount The state transition equation uses a time-varying matrix as The parameters are set as (glucose metabolism rate), (potassium excretion rate), , the process noise covariance .

[0015] 2. Improved Kalman filter The observation equation is defined as: The observation noise covariance , and the Kalman gain update formula is: Real-time output of the posterior prediction , the data update frequency is 100 Hz.

[0016] When the input electrochemical signal is interfered by the optical sensor (±10% fluctuation), the residual error after filtering is <0.5% (3.2% for the traditional Kalman filter).

[0017] 3. D-S Evidence Decision Set the hypothesis set , the basic probability assignment of the electrochemical sensor Assignment of the optical sensor The calculated confidence after combination is , When , an alarm is triggered, and the corresponding blood glucose concentration threshold is .

[0018] 4. Antithrombotic Control The amplitude function of the piezoelectric bristles on the inner wall of the catheter is: where the damping coefficient , the driving frequency .

[0019] Animal experiment (porcine model) Index This system Laboratory test Error rate p-value Blood glucose (mmol / L) 6.2±0.3 6.1±0.2 1.64% 0.12 Potassium (mM) 4.0±0.2 3.9±0.1 2.56% 0.08 Comparison of the number of punctures: 21 punctures in 7 days in the traditional group vs 1 implantation in the system group (p < 0.001) Response time in human trials: After 7 days of indwelling in 30 patients, the thrombus coverage rate on the catheter surface was <5% (more than 30% for the control traditional catheter), the platelet adsorption amount was reduced by 82% (p < 0.01), from the increase in blood potassium to the system alarm: an average of 2.3 minutes (laboratory report delay ≥ 45 minutes); reduction of punctures: in diabetic patients: the number of punctures within 7 days decreased from 28 to 3 (89.3% reduction).

[0020] As a specific embodiment 3 of the present invention: Example of step-by-step calculation in a medical scenario Let the recognition framework Θ = {Normal, Hyper, Hypo}, and the data of two groups of sensors are as follows: Evidence source 1 (electrochemical)

[0021] Evidence source 2 (optical)

[0022] Calculate the fusion confidence of "hyperglycemia (Hyper)" 1. Find all combinations:

[0023] 2. Calculate the numerator: m1(Hyper)m2(Hyper) + m1(Hyper)m2(Θ) + m1(Θ)m2(Hyper) = 0.75×0.6 + 0.75×1 + 1×0.6 = 1.8 3. Calculate the conflict amount K: m1(Normal)m2(Hyper) + m1(Hyper)m2(Normal) +... = 0.1×0.6 + 0.75×0.2 +... = 0.27 4. Final result: (Hyper) = \frac{1.8}{1 - 0.27} = 0.85 Blood glucose diagnosis threshold: The judgment threshold for Hyper is set to 7.8 mmol / L (corresponding to the calculation method of m;(Hyper)).

[0024] Confidence threshold: An alarm is triggered when m_comb(H) > 0.85.

[0025] The above are only the preferred embodiments of the present invention, and do not limit the patent scope of the present invention. Any equivalent structural transformation made under the inventive concept of the present invention, or direct / indirect application in other related technical fields, is included in the patent protection scope of the present invention.

Claims

1. An intelligent real-time blood detection system based on an intravascular indwelling sensor, characterized in that, Including: An intravascular sensing unit, which includes an electrochemical sensor, an optical sensor and a catheter indwelled in a blood vessel through venous puncture. The electrochemical sensor obtains the Faraday current corresponding to blood biochemical indexes in real time through redox reactions, and the optical sensor detects the concentration of blood components through the absorption characteristics of dual-wavelength near-infrared spectroscopy; A data processing unit, which performs a detection method on the electrochemical current signal and the optical absorbance signal, including the following steps: S1. Establish a three-dimensional state space model including blood glucose concentration G, blood potassium ion concentration K and motion artifact M, and fuse multi-source signals through the Kalman filtering algorithm, satisfying the following improved expression: The state vector value is: where is the blood glucose concentration, is the blood potassium ion concentration, is the amount of motion artifact interference; The observed vector value is: Among them, , is the current output of the electrochemical sensor, is the absorbance value of the optical sensor, is the magnitude of the combined vector of the triaxial accelerometer; The update law is as follows: where represents the Kalman gain matrix, is the prior estimation error covariance matrix, is the observation noise covariance matrix, is the observation matrix, , is the sensor noise variance, is the blood glucose - current conversion coefficient, is the blood potassium - absorbance conversion coefficient; S2. The D - S evidence theory is used for diagnostic decision - making, and the combined confidence level is: where represents the hyperglycemic value, H is the hypothesis to be verified, is the value of the electrochemical sensor, is the value of the optical sensor, and the conflict coefficient is , when is satisfied, an alarm is triggered, is the preset confidence threshold.

2. The intelligent real-time blood detection system based on an intravascular indwelling sensor according to claim 1, wherein The state transition matrix is: wherein is the glucose metabolic rate constant; is the potassium excretion rate constant; is the exercise attenuation coefficient; is the modulus of the acceleration vector.

3. The intelligent real-time blood detection system based on an intravascular indwelling sensor according to claim 1, characterized in that, The inner wall of the catheter includes a piezoelectric drive bristle structure, and its amplitude control function is: is the maximum amplitude, is the damping coefficient, is the driving frequency, is the phase compensation angle.

4. The intelligent real-time blood detection system based on an intravascular indwelling sensor according to claim 1, wherein The state equation is also included in the step S1: Predicting physiological parameters, where is a state transition matrix, including glucose metabolism kinetics, the renal excretion rate of potassium in the blood; the time decay characteristics of motion artifacts; Observation equation: , is the observation matrix to generate the signal estimation value.

Images