Artificial intelligence cell culture medium glucose concentration measurement method

Abstract

The invention relates to the field of measurement, and discloses an artificial intelligent cell culture medium glucose concentration measurement method, which is used for providing an accurate glucose concentration measurement method for a cell culture process. Comprising the steps that first and second glucose sensing electrodes are arranged to obtain current signals, differential current signals are obtained through processing, and the glucose concentration value after environment compensation is determined by combining temperature and pH value correction coefficients. In continuous measurement, the baseline noise level is determined according to the second current signal, the electrode response state is determined according to the differential current signal dynamic response, calibration is triggered, the conversion parameters are updated, and the actual glucose concentration value is obtained according to the conversion parameters. The device is accurate and stable in measurement, can realize real-time monitoring and maintenance, and provides reliable data support for cell culture.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of measurement, in particular to a method for measuring glucose concentration in cell culture medium based on artificial intelligence. BACKGROUND

[0002] In the field of cell culture technology, accurate monitoring of glucose concentration in cell culture medium is crucial for cell growth, metabolic research, and biopharmaceutical production process control. As a key energy source for cell growth and metabolism, the dynamic change of glucose concentration directly reflects the metabolic activity and growth status of cells. Therefore, real-time, accurate, and stable measurement of glucose concentration in cell culture medium can provide key data support for cell culture process optimization, cell physiological state evaluation, and biological product quality control.

[0003] Traditional methods for measuring glucose concentration in cell culture medium mainly rely on offline sampling analysis, such as high-performance liquid chromatography (HPLC), glucose oxidase-peroxidase method (GOD-POD), etc. Although these methods have high measurement accuracy, they have the disadvantages of complicated operation, long analysis time, and inability to monitor in real time, which cannot meet the needs of real-time data feedback and dynamic regulation in modern cell culture processes.

[0004] The prior art does not fully consider the influence of temperature and pH in the cell culture environment on the measurement performance of glucose sensing electrodes, lacks effective environmental compensation mechanisms, resulting in inaccurate measurement results under different environmental conditions, and cannot truly reflect the actual glucose concentration in the cell culture medium. Large molecules and cell debris in the cell culture medium can produce common mode interference signals, which are superimposed with glucose detection signals, making it difficult for the electrode to accurately distinguish the signals produced by glucose, thereby reducing the selectivity and stability of the measurement. The prior art lacks real-time monitoring and evaluation means for electrode performance, and cannot timely discover problems such as electrode sensitivity decay and baseline drift. When the electrode performance declines, it cannot be calibrated or maintained in time, leading to gradually increasing measurement errors and affecting the accurate control of the cell culture process. The prior art only provides the measurement value of glucose concentration and does not evaluate the confidence of the measurement result. In the presence of interference factors or abnormal electrode performance, the reliability of the measurement result cannot be judged, which may bring risks to the optimization and quality control of the cell culture process.

[0005] Therefore, we propose a method for measuring glucose concentration in cell culture medium based on artificial intelligence to solve the above problems. SUMMARY

[0006] The application provides an artificial intelligence cell culture medium glucose concentration measurement method, which is used for providing an accurate glucose concentration measurement method for a cell culture process.

[0007] The first aspect of the application provides an artificial intelligence cell culture medium glucose concentration measurement method, comprising: setting a first glucose sensing electrode and a second glucose sensing electrode in a cell culture medium, obtaining a first current signal generated by the first glucose sensing electrode in response to glucose in the culture medium, and a second current signal generated by the second glucose sensing electrode; obtaining a differential current signal according to the first current signal and the second current signal; obtaining a current temperature signal and a current pH signal of the cell culture medium, determining a temperature correction coefficient according to the current temperature signal, and determining a pH correction coefficient according to the current pH signal; determining an environment-compensated glucose concentration value based on the differential current signal, the temperature correction coefficient and the pH correction coefficient; in a continuous measurement process, determining a baseline noise level according to the second current signal, and determining an electrode response state according to a dynamic response of the differential current signal; when the baseline noise level exceeds a first preset threshold or the electrode response state exceeds a second preset threshold, triggering a calibration operation, and updating a conversion parameter according to which the glucose concentration value is determined based on the calibration operation, and obtaining an actual glucose concentration value according to the conversion parameter.

[0008] Optionally, in the first implementation manner of the first aspect of the application, the method comprises: directly immersing a sensitive end of the first glucose sensing electrode into the cell culture medium, so that a glucose oxidase sensing layer of the first glucose sensing electrode is in direct contact with the culture medium, and the first current signal is obtained; encapsulating a sensitive end of the second glucose sensing electrode in a sealed chamber made of a dialysis membrane, the sealed chamber being filled with a pH buffer inside, and the dialysis membrane being in contact with the cell culture medium, so that the second glucose sensing electrode is isolated from macromolecular substances and cells in the culture medium, and ions and small molecules are allowed to pass through, and the second current signal is obtained; synchronously collecting and recording the first current signal and the second current signal under the same constant potential bias condition.

[0009] Optionally, in the second implementation manner of the first aspect of the application, the method comprises: processing the first current signal and the second current signal to obtain a first smoothed current signal and a second smoothed current signal; comparing the second smoothed current signal with a reference signal determined by a background output of the second glucose sensing electrode in a glucose-free environment to obtain a real-time common-mode interference evaluation result; based on the real-time common-mode interference evaluation result, performing dynamic baseline subtraction on the first smoothed current signal to obtain a primary differential signal; multiplying the primary differential signal by a normalization coefficient dynamically adjusted according to historical sensitivity data of the first glucose sensing electrode to obtain a differential current signal.

[0010] Optionally, in the third implementation form of the first aspect of the present application, further comprising: calculating a real-time signal quality index based on the fluctuation amplitude of the primary differential signal and the average value of the differential current signal in the preset stable period; comparing the real-time signal quality index with a preset quality threshold to generate a quality classification result for judging the availability of the current differential signal; triggering a re-sampling and differential operation process of the first smoothed current signal and the second smoothed current signal when the quality classification result indicates that the signal quality is suspicious, and replacing the original signal with a newly generated differential current signal; suspending the use of the current differential signal when the quality classification result indicates that the signal quality is poor, and calling the stored historical normalized differential current signal which is judged as good quality in the last time for subsequent steps; periodically recording the adjustment history of the normalization coefficient to form an electrode sensitivity decay trend chart, and generating a warning prompt of electrode performance degradation when it is detected that the trend presents continuous unidirectional decay and the cumulative amplitude exceeds the preset sensitivity decay warning threshold.

[0011] Optionally, in the fourth implementation form of the first aspect of the present application, comprising: obtaining a current temperature reading and a current pH reading as a current temperature original signal and a current pH original signal respectively; processing the current temperature original signal to obtain a current temperature value, and processing the current pH original signal to obtain a current pH value; matching and interpolating the current temperature value with a pre-stored first group of discrete temperature-compensation coefficient mapping data to output a temperature correction coefficient; matching and interpolating the current pH value with a pre-stored second group of discrete pH-activity coefficient mapping data to output a pH correction coefficient; multiplying the temperature correction coefficient and the pH correction coefficient to generate an environment comprehensive correction factor.

[0012] Optionally, in the fifth implementation form of the first aspect of the present application, comprising: dividing the differential current signal by the environment comprehensive correction factor to generate a corrected electrochemical response value; mapping the corrected electrochemical response value to a concentration calibration result by querying a pre-stored calibration curve associating the electrochemical response value with the glucose concentration; performing time series smoothing filtering on the concentration calibration result to output a real-time glucose concentration value; recording fluctuation data of the second current signal in a preset time to form a baseline fluctuation sequence and calculate the standard deviation of the sequence as a noise level evaluation value representing the measurement stability during continuous measurement; monitoring the rising or falling process of the differential current signal after a known change in the glucose concentration in the cell culture medium, and recording the time required to reach a stable state as a response lag evaluation value representing the electrode performance; determining that the electrode state is abnormal and generating a calibration trigger instruction when the noise level evaluation value exceeds a preset stability threshold or the response lag evaluation value exceeds a preset speed threshold.

[0013] Optionally, in the sixth implementation form of the first aspect of the present application, the instantaneous glucose concentration is calculated by is the corrected electrochemical response value; is the zero-point intercept; is the sensitivity slope; is the environmental comprehensive correction factor.

[0014] Optionally, in the seventh implementation form of the first aspect of the present application, further comprising: acquiring a dissolved oxygen concentration signal of the cell culture medium in real time, calculating a change rate of the dissolved oxygen concentration per unit time as an evaluation value of oxygen consumption rate of cell respiration metabolism; based on a biochemical metrological relationship of cell metabolism, using the change rate of the real-time glucose concentration value to calculate a theoretical oxygen consumption rate corresponding to the change rate; comparing the measured oxygen consumption rate evaluation value with the calculated theoretical oxygen consumption rate, calculating the deviation percentage between the two, and generating a metabolic correlation verification coefficient; according to the numerical range of the metabolic correlation verification coefficient, grading the confidence level of the real-time glucose concentration value; when the confidence level is graded as high, directly outputting the concentration value; when the confidence level is graded as low, using the concentration value in the last high-confidence state as the basis, superimposing and weighting the glucose consumption estimation value independently calculated from the oxygen consumption rate evaluation value for fusion to generate and output the metabolic verification corrected glucose concentration value; continuously recording the historical data of the noise level evaluation value and the response lag evaluation value, and when it is found through analysis that both of them show a monotonic increasing trend and the cumulative amplitude exceeds the respective trend warning threshold at the same time, a predictive maintenance signal is generated.

[0015] ​​​Optionally, in the eighth implementation form of the first aspect of the present application, the method further comprises: switching the first glucose sensing electrode and the second glucose sensing electrode from the cell culture medium to a calibration pool containing a standard glucose solution with a known concentration at the same time, obtaining a stable differential current signal under the standard solution after the output signal is stable, and generating a standard solution response value; comparing the standard solution response value with the known concentration value, calculating a real-time sensitivity parameter reflecting the current electrode sensitivity and a real-time offset parameter for correcting the measurement baseline; switching the electrodes to another verification solution with a known concentration, using the calculated real-time sensitivity parameter and real-time offset parameter to calculate a verification concentration prediction value, comparing the prediction value with the actual known concentration, and generating a calibration verification deviation value; if the calibration verification deviation value is less than a preset tolerance threshold, determining that the current calibration is valid, replacing the original conversion parameter with the real-time sensitivity parameter and the real-time offset parameter, and completing the update; if the calibration verification deviation value is greater than or equal to the tolerance threshold, determining that the current calibration is invalid, generating a calibration failure alarm, and maintaining the original conversion parameter unchanged; after the parameter update or alarm generation is completed, switching the electrodes from the calibration environment back to the cell culture medium, and the system returns to the normal glucose concentration continuous measurement state.

[0016] Optionally, in the ninth implementation form of the first aspect of the present application, the method further comprises: calculating a real-time data quality score based on the baseline noise level evaluation value and the electrode response state evaluation value, combining the environmental comprehensive correction factor; according to the numerical range of the real-time data quality score, dividing the quality of the measurement data into a plurality of predefined reliability levels, and generating corresponding data state labels; associating and packaging the data state labels with the real-time glucose concentration value to form a concentration data packet with reliability identification; when the data state label indicates high reliability, the system outputs the concentration data packet with reliability identification at a normal frequency; when the data state label indicates low reliability, the system activates an alarm signal while outputting the concentration data packet.

[0017] The mechanism of the present application is as follows: the common mode noise is stripped from the source, and the high-purity glucose-specific electrochemical response signal is directly extracted, solving the inherent problem of traditional single electrode being easily disturbed by environmental drift; Beneficial effects: in a complex cell culture environment, the common mode interference generated by macromolecular substances and cell fragments can be effectively eliminated, the accuracy and stability of glucose concentration measurement can be significantly improved, and more reliable monitoring data can be provided for the cell culture process; Not only the influence of temperature on measurement is considered, but also the pH value factor is considered. By obtaining the current temperature and pH value signals, the temperature correction coefficient and the pH value correction coefficient are determined respectively, and the environmental comprehensive correction factor is generated by multiplying the two, and the differential current signal is comprehensively corrected; The electrode performance can be monitored in real time, and the situation that the electrode noise level is too high or the response state is abnormal can be found in time, and a calibration operation is triggered. Through predictive maintenance, measures can be taken at the initial stage of electrode performance decline, so as to avoid the gradual increase of measurement error, ensure the stability and reliability of long-term measurement, and reduce experimental failure and production interruption caused by electrode problems; The signal quality is quantitatively evaluated and processed, the signal quality problem can be found in time and effective measures can be taken, the error measurement result caused by the signal quality problem is avoided, and the reliability and stability of the whole measurement system are improved, and more reliable data basis is provided for optimization of the cell culture process; The glucose concentration measurement result is verified and corrected from the biochemical measurement relationship of cell metabolism, the influence of cell metabolic activity on the measurement result is fully considered, and the confidence of the measurement result is further improved, and higher level data security is provided for accurate control of the cell culture process. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 An embodiment schematic diagram of the artificial intelligence cell culture medium glucose concentration measurement method in the embodiment of the application; Figure 2 Another embodiment schematic diagram of the artificial intelligence cell culture medium glucose concentration measurement method in the embodiment of the application; Figure 3 An embodiment schematic diagram of the artificial intelligence cell culture medium glucose concentration measurement device in the embodiment of the application. DETAILED DESCRIPTION

[0019] The embodiment of the application provides a kind of artificial intelligence cell culture medium glucose concentration measurement method, to provide accurate glucose concentration measurement method for cell culture process.The terms "first", "second", "third", "fourth" and the like (if exist) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not have to be limited to those steps or units clearly listed, but can include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0020] For ease of understanding, the specific process of the embodiment of the application is described below, please refer to Figure 1An embodiment of the cell culture medium glucose concentration measurement method of the artificial intelligence in the embodiment of the application includes: 101. A first glucose sensing electrode and a second glucose sensing electrode are disposed in a cell culture medium, wherein a surface of the second glucose sensing electrode is isolated from the cell culture medium by a physical isolation layer, a first current signal generated by the first glucose sensing electrode in response to glucose in the culture medium is obtained, and a second current signal generated by the second glucose sensing electrode is obtained. It can be understood that the execution subject of the application can be an artificial intelligence cell culture medium glucose concentration measurement device, and can also be a terminal or a server, and the specific implementation is not limited herein. The embodiment of the application takes the server as an execution subject for example.

[0021] It should be noted that a double electrode system is constructed based on a screen-printed carbon electrode. The first glucose sensing electrode (working electrode) is directly exposed to the cell culture medium, and the sensitive film thereof is made of glucose oxidase (GO X ) and osmium-based redox polymers cross-linked by glutaraldehyde, with a thickness of about 5 microns, which can directly catalyze glucose to generate an electrical signal. The second glucose sensing electrode (reference electrode) has the same structure, but is covered with a layer of physical isolation layer, i.e., a Nafion / polyimide composite film (thickness of about 10 microns) prepared by spin coating. The film can block glucose molecules (molecular weight > 500 Da) from passing through, while allowing electrolyte ions to diffuse freely, thereby isolating glucose while maintaining the stability of the electrochemical interface.

[0022] Electrode preparation: a pair of carbon working electrodes (diameter 2 mm) and Ag / AgCl reference electrodes, carbon counter electrodes are prepared on a polymer substrate by screen printing technology to form a three-electrode system. First electrode sensitive layer: glucose oxidase (activity ≥ 200 U / mg) and osmium polymer are mixed in a mass ratio of 1:3, 5 μL of which is dropped and coated on the surface of the electrode, and a biological sensitive film is formed by cross-linking and curing at 25°C. Second electrode isolation layer: spin-coat Nafion solution (5% w / w) mixed with polyimide precursor on the same sensitive film, and form a dense isolation film by curing at 60°C. It is tested that it can block ≥5mM glucose penetration.

[0023] Signal acquisition: The dual-electrode was integrated in the cell pocket of the smart bioreactor, immersed in RPMI-1640 medium containing 5% fetal bovine serum, maintained at 37℃, pH 7.4. A constant potential of 0.3V (vs. Ag / AgCl) was applied by the electrochemical workstation, and the current was recorded continuously: the first electrode response current (example): a steady-state current of 52.3μA was generated when the glucose concentration was 5mM (the signal contains glucose response and background interference). The second electrode baseline current (example): due to the blocking of glucose by the isolation layer, only a background noise current of 0.8μA was generated (mainly from the oxidation of electroactive substances such as ascorbic acid in the culture medium).

[0024] 102. Perform a difference operation on the first current signal and the second current signal to obtain a difference current signal; It should be noted that the difference operation aims to eliminate non-specific interference in the first electrode (working electrode) signal, and by subtracting the common-mode noise captured by the second electrode (reference electrode), only the net signal generated by the glucose oxidation reaction is retained. The core process includes signal synchronization, baseline correction, and dynamic compensation.

[0025] The dual-electrode signal is synchronously collected by a 24-bit ADC at a sampling rate of 10Hz. First, the two groups of current signals are subjected to sliding window averaging (window width 5 seconds) to smooth out transient pulse noise, and 3 seconds of raw data are recorded continuously in RPMI-1640 medium at 37℃, pH 7.4: first electrode current sequence (μA): [52.8, 53.1, 52.5], mean 52.8μA (contains glucose response and interference); second electrode current sequence (μA): [0.85, 0.82, 0.88], mean 0.85μA (contains only interference noise).

[0026] The direct difference operation formula is: difference current = first electrode current - correction factor x second electrode current. The correction factor is determined by electrode characteristic calibration: membrane thickness ratio: first electrode enzyme layer thickness 5.2μm, second electrode isolation layer thickness 10.1μm, thickness ratio coefficient 0.51 (the thicker the membrane, the lower the interference substance permeability); surface area ratio: the effective reaction area of both electrodes is 2.0mm 2 , area ratio coefficient 1.0; background noise ratio: the noise current ratio of the two electrodes measured in glucose-free PBS solution is about 1.12. The comprehensive correction factor is calculated as: 0.51x1.0x1.12≈0.57. Substituting the real-time data: 52.8-0.57x0.85=52.8-0.48=52.32μA, the net difference current is 52.32μA.

[0027] To verify the differential effect, 0.1 mM ascorbic acid (a common interfering agent) was added to the culture medium: the current at the first electrode jumped to 54.6 μA (an increase of 1.8 μA); the current at the second electrode jumped to 1.72 μA (an increase of 0.87 μA); the current after differential treatment was 54.6 - 0.57 × 1.72 = 53.62 μA, with a fluctuation of only 1.3 μA, and the interference inhibition rate reached 28%.

[0028] The final differential signal undergoes a 5-point median filter to remove residual glitches before being output to the subsequent environmental compensation module. The system updates the correction factor every 30 seconds to adapt to performance drift during long-term electrode use.

[0029] 103. Obtain the current temperature signal and current pH signal of the cell culture medium, determine the temperature correction coefficient based on the current temperature signal, and determine the pH correction coefficient based on the current pH signal. It should be noted that, based on the dual-electrode differential current signal, real-time monitoring signals of temperature and pH are integrated, and correction coefficients are dynamically determined to compensate for the impact of environmental fluctuations on glucose measurement. The core of this implementation lies in using a multi-parameter sensing system integrated with the glucose sensing electrode, and obtaining the correction coefficients through signal conversion and lookup table interpolation.

[0030] A thin-film temperature sensor (platinum resistance thermometer, sensitivity 4.03 Ω / ℃) and a solid-state pH sensor (IrO2) are integrated on the same flexible substrate as a screen-printed carbon electrode. x The membrane has a sensitivity of -57.12 mV / pH. The spacing between the three sensors is ≤2 mm to ensure consistency of the monitoring area. Signal acquisition is as follows: Temperature signal: A platinum resistance thermometer is driven by a constant current source (100 μA) to measure the voltage across its terminals. The voltage is then converted to a digital value by an ADC. Under standard conditions of 37.0℃, the resistance of the platinum resistance thermometer is 115 Ω, corresponding to a voltage of 11.5 mV. pH signal: The potential difference between the pH sensor and the reference electrode is measured using a high-impedance potentiometer. When the pH of the culture medium is 7.40, the measured potential is -422 mV (relative to the Ag / AgCl reference electrode). Both sets of signals are acquired synchronously with the glucose current signal at a sampling rate of 1 Hz and are filtered by moving average (window width 5 seconds) to eliminate transient fluctuations.

[0031] The coefficients are determined using a "piecewise table lookup + linear interpolation" method to avoid complex formula calculations: Temperature correction coefficient (K t Based on the pre-calibrated temperature-enzyme activity relationship table (Table 1), when the measured temperature is 37.3℃, the corresponding K for 37.0℃ is found by referring to the table. t =1.000, 38.0℃ corresponds to K t =1.020, K is obtained through linear interpolation. t =1.006. The temperature correction factor reference table is shown in Table 1: Table 1

[0032] pH correction coefficient (K ph ): look up calibration curve (Table 2) according to pH sensor potential value. If the measured potential is -415 mV (corresponding to pH ≈ 7.28), look up the table to get K ph = 1.030 when pH = 7.2, K ph = 1.000 when pH = 7.4, and interpolate K ph = 1.012. The pH correction coefficient is referenced as follows in Table 2: Table 2

[0033] The system checks the validity of the environmental signals every 30 minutes: if the temperature signal is out of the range of 35.5-38.5℃, or the pH signal is out of the range of 6.8-7.8, the coefficient update is suspended, an alarm is triggered, and the last valid value is used. The sensor is automatically calibrated every 24 hours according to the standard buffer (pH = 7.00) and the thermostat (37.0℃), and the table lookup reference is updated.

[0034] The correction coefficients K t , K ph are transmitted synchronously with the differential current signal to the subsequent concentration calculation module, and the current period output is: temperature 37.3℃ → K t = 1.006, pH = 7.28 → K ph = 1.012, which provides input for the concentration compensation of step 104.

[0035] 104. Determine the glucose concentration value compensated by the environment based on the differential current signal, the temperature correction coefficient, and the pH correction coefficient. In the continuous measurement process, the baseline noise level is determined according to the second current signal, and the electrode response state is determined according to the dynamic response of the differential current signal. It should be noted that a concentration calculation model is established, and a dynamic monitoring logic based on signal characteristics is designed.

[0036] Read the current differential current signal (52.32 μA, from step 102), and the temperature correction coefficient (K t = 1.006, corresponding to temperature 37.3℃) and the pH correction coefficient (K ph= 1.012, corresponding to pH 7.28, from step 103). The concentration calculation is achieved by a pre-calibrated current-concentration conversion model: Calibration benchmark: under standard conditions (37.0 °C, pH 7.4), the relationship between differential current and glucose concentration is linear, with a sensitivity coefficient a = 10.5 pA / mM (i.e. 10.5 pA current per 1 mM glucose). Compensation calculation: preliminary concentration = differential current / a = 52.32 pA / 10.5 pA / mM ~ 4.98 mM. Final concentration after environmental compensation = preliminary concentration x K t x K ph = 4.98 mM x 1.006 x 1.012 ~ 5.07 mM. This process is performed every 30 seconds, with the output of the compensated concentration value (5.07 mM) and the error controlled within ±2% (compared with high-performance liquid chromatography results).

[0037] The baseline noise level is determined by the stability of the current signal of the second electrode (reference electrode): data window: take the current sequence of the second electrode in the last 5 minutes (sampling rate 1 Hz, 300 points in total), the sequence mean is 0.85 pA, and the standard deviation is 0.12 pA. Noise quantification: baseline noise level = standard deviation x 2 (covering 95% confidence interval) = 0.12 pA x 2 = 0.24 pA. Threshold comparison: the first preset threshold is set to 0.5 pA (set according to historical data). The current noise 0.24 pA is lower than the threshold, indicating that the background interference of the system is controllable; if it exceeds the threshold (noise > 0.5 pA), it may be due to the sudden addition of antioxidants in the culture medium causing fluctuations in the second electrode signal.

[0038] The electrode response state is evaluated according to the dynamic characteristics of the differential current signal, focusing on response speed and stability: response rate calculation: inject a standard glucose solution (concentration step 0.5 mM) into the culture medium, record the time required for the differential current to rise from 10% to 90% of the peak value (response time), the current response time is 28 seconds (the historical benchmark mean is 25 seconds, and the standard deviation is 2 seconds). State quantification: (normalized deviation value). Threshold comparison: the second preset threshold is set to 2.0. The current state value 1.5 does not exceed the threshold, indicating that the electrode activity is normal; if it exceeds 2.0 (response time is delayed to 35 seconds), it may be due to enzyme layer aging or contamination causing slow response.

[0039] The system outputs three types of results per cycle: compensated glucose concentration value: 5.07 mM (for real-time monitoring); baseline noise level: 0.24 pA (normal state); electrode response state: 1.5 (normal state). When any state value exceeds the threshold, send a calibration trigger signal to step 105 immediately (baseline noise > 0.5 pA or response state > 2.0) to ensure measurement reliability.

[0040] 105、When the baseline noise level exceeds the first preset threshold, or the electrode response state exceeds the second preset threshold, a calibration operation is triggered, and conversion parameters for determining the glucose concentration value are updated based on the calibration operation, and the actual glucose concentration value is obtained based on the conversion parameters.

[0041] It should be noted that a double threshold triggering mechanism and parameter adaptive learning process are established.

[0042] The system presets two types of thresholds for determining the calibration timing: the first preset threshold (baseline noise): the upper limit of the baseline noise level is set to 0.5 μA (based on the 95% confidence interval of the historical noise data of the second electrode). The second preset threshold (electrode response state): the upper limit of the electrode response state is set to 2.0 (standardized deviation value, corresponding to an electrode response time delay exceeding the reference value by ± 2 times the standard deviation).

[0043] Real-time monitoring data example: continuously monitor the second electrode current signal (sampling rate 1 Hz), calculate the noise standard deviation in the last 5 minute window: baseline noise level = 0.62 μA (exceeds the 0.5 μA threshold, possibly due to the sudden introduction of antioxidants in the culture medium). Simultaneously evaluate the dynamic response of the differential current signal: inject 0.5 mM glucose standard solution into the culture medium, record the time for the current to rise from 10% to 90% peak value: current response time = 35 seconds (historical reference is 25 seconds, standard deviation is 2 seconds), calculate (exceeds the 2.0 threshold, possibly due to attenuation of enzyme layer activity).

[0044] When either indicator exceeds the threshold, the system automatically triggers a calibration operation, and the process is as follows: Step 1: calibration liquid injection: a peristaltic pump injects a calibration liquid (5.0 mM glucose in PBS buffer, pH 7.4) preheated to 37°C at a flow rate of 0.1 mL / min into the reactor, replacing part of the culture medium (total volume replacement rate 10%). Step 2: signal acquisition and comparison: acquire the steady-state current signal of the double electrode in the calibration liquid: first electrode current: 52.5 μA (theoretical value should be 52.8 μA, deviation -0.57%). Second electrode current: 0.2 μA (ideal baseline value). Step 3: error analysis and fault diagnosis: if the current deviation exceeds the tolerance (± 3%), further diagnose the cause: case: first electrode current is consistently low, combined with electrode response state exceeding the standard, judge as enzyme layer aging (need to update sensitivity coefficient). Case: second electrode noise increases suddenly, judge as isolation layer contamination (trigger physical cleaning program: 0.1 M NaOH flushing for 5 minutes).

[0045] According to the calibration results, dynamically update the key parameters in the concentration calculation (as shown in Table 3): Table 3

[0046] Update logic: Sensitivity coefficient a is refitted by least square method: If the calibration deviation is less than 1% for three consecutive times, the calibration interval is extended (from 24 hours to 48 hours); if the deviation is greater than 5%, the interval is shortened to 12 hours.

[0047] After updating the parameters, the system verifies the accuracy using a standard (2.5 mM glucose): measurement error before calibration: -4.8% (measured concentration 2.38 mM). Measurement error after calibration: -0.6% (measured concentration 2.49 mM). Record this calibration data (noise source, electrode decay rate) to the historical database for optimizing threshold setting (adjust noise threshold dynamically to 0.55 mA to give early warning).

[0048] Please refer to Figure 2 Another embodiment of the cell culture medium glucose concentration measurement method using artificial intelligence in the embodiment of the application includes: 201. A first glucose sensing electrode and a second glucose sensing electrode are disposed in a cell culture medium, wherein the surface of the second glucose sensing electrode is isolated from the cell culture medium by a physical isolation layer, a first current signal generated by the first glucose sensing electrode in response to glucose in the culture medium is obtained, and a second current signal generated by the second glucose sensing electrode is obtained. Specifically, the sensitive end of the first glucose sensing electrode is directly immersed in the cell culture medium, so that the glucose oxidase sensing layer is in direct contact with the culture medium, and the first current signal generated directly by the glucose oxidation-reduction reaction is obtained. The sensitive end of the second glucose sensing electrode is packaged in a sealed chamber made of dialysis membrane, the inside of the sealed chamber is filled with pH buffer, and the dialysis membrane is in contact with the cell culture medium, so that the second glucose sensing electrode is isolated from macromolecular substances and cells in the culture medium, but allows ions and small molecules to pass through, thereby obtaining the second current signal composed of environmental interference and electrode background noise. Under the same constant potential bias condition, the first current signal and the second current signal are synchronously collected and recorded.

[0049] It should be noted that two platinum-carbon composite working electrodes with the same physical specifications and a diameter of 3 millimeters are selected.

[0050] The first glucose sensing electrode (working electrode): a mixed solution of 200 units / ml active glucose oxidase (GOD) and chitosan was drop-coated on the sensitive end of the electrode to form a bio-sensing layer. No additional physical compartment was set up for this electrode, ensuring that the enzyme layer could directly contact the external liquid. The second glucose sensing electrode (reference / background electrode): its sensitive end was kept bare (without enzyme coating) and was encapsulated in a micro-cylindrical polyether ether ketone (PEEK) sealed chamber. The chamber had a volume of 50 μl and was pre-filled with 0.1 M phosphate buffer solution (PBS) with a pH of 7.4. A regenerated cellulose dialysis membrane with a molecular weight cut-off (MWCO) of 12,000 Da was used to seal the bottom of the chamber. This dialysis membrane allowed hydrogen ions, potassium ions and small molecule interferents (acetaminophen or ascorbic acid) to pass freely, but effectively blocked proteins, cell debris and enzyme molecules in the cell culture medium from entering the chamber.

[0051] The above constructed two-electrode probe assembly was inserted into a running 5-l CHO (Chinese hamster ovary) cell bioreactor. At this time, the first electrode was directly immersed in the glucose-rich culture medium; the second electrode was in contact with the culture medium through the dialysis membrane, establishing an ion path.

[0052] A multi-channel potentiostat was connected and a uniform constant potential bias voltage of +0.6 V vs. a silver / silver chloride reference electrode was set. This voltage was sufficient to drive the electrochemical oxidation of hydrogen peroxide (the product of glucose oxidation). The data sampling frequency was set to 1 Hz (1 sample per second).

[0053] After the system was started, the current data of the two channels were recorded synchronously under the constant potential bias. At the 100th minute of the reaction, the actual glucose concentration in the reactor was about 15 mM, and there was a small amount of electrochemically active interferent. At this specific time (T = 100 min), the specific data collected by the system were as follows: the first current signal (I main ): the reading was 658.4 nA. Analysis: this value included the oxidation current produced by the glucose reaction catalyzed by glucose oxidase (main component), as well as the current produced by background noise and interferents (secondary component). The second current signal (I ref ): the reading was 32.1 nA. Analysis: due to the physical isolation of the dialysis membrane, glucose could not enter the sealed chamber to contact the electrode, and there was no oxidase in the chamber, so this value did not include the reaction current of glucose. It was only composed of the double-layer capacitance charging current of the electrode itself, the electronic drift noise, and the current produced by the small molecule interferents that penetrated through the dialysis membrane.

[0054] 202. performing a differential operation on the first current signal and the second current signal to obtain a differential current signal; Specifically, the synchronously collected first current signal and second current signal are subjected to low-pass filtering to obtain a first smoothed current signal and a second smoothed current signal, respectively, which are filtered of high-frequency noise; the second smoothed current signal is compared with a reference signal determined by the background output of the second glucose sensing electrode in a glucose-free environment to obtain a real-time common-mode interference evaluation result; based on the real-time common-mode interference evaluation result, the first smoothed current signal is subjected to dynamic baseline subtraction to obtain a primary differential signal representing the intensity of net glucose redox reaction; the primary differential signal is multiplied by a normalization coefficient dynamically adjusted according to historical sensitivity data of the first glucose sensing electrode, and finally a standardized differential current signal is obtained, wherein the historical sensitivity data is updated through a calibration process.

[0055] Further, it also includes: based on the fluctuation amplitude of the primary differential signal and the mean value of the standardized differential current signal in the preset stable period, a real-time signal quality index is calculated; the real-time signal quality index is compared with a series of preset quality thresholds to generate a quality classification result for judging the usability of the current differential signal; when the quality classification result indicates that the signal quality is excellent, the standardized differential current signal is directly used in the subsequent steps; when the quality classification result indicates that the signal quality is suspicious, a re-sampling and differential operation process of the first smoothed current signal and the second smoothed current signal is triggered, and the newly generated differential current signal replaces the original signal; when the quality classification result indicates that the signal quality is poor, the current differential signal is suspended, and the historical standardized differential current signal stored when the quality is last judged to be excellent is called for use in the subsequent steps; the adjustment history of the normalization coefficient is recorded periodically to form an electrode sensitivity decay trend chart; when it is detected that the trend presents continuous one-way decay and the cumulative amplitude exceeds the preset sensitivity decay warning threshold, a warning prompt of electrode performance degradation is generated.

[0056] Note that the signal filtering and common mode interference assessment microprocessor receives two synchronous raw signals from the front-end circuit. Low-pass filtering: the original first current signal (working electrode) fluctuates between 655 nA and 662 nA, and the second current signal (reference interference electrode) fluctuates between 30 nA and 34 nA. After processing using a 5-point sliding average filtering algorithm, the system obtains a stable first smoothed current signal of 658.4 nA and a second smoothed current signal of 32.1 nA. Interference assessment: the system calls the memory parameters to read the inherent background signal value (i.e., the reference signal) of the second glucose sensing electrode in a pure glucose-free buffer, which is 2.1 nA. The system compares the second smoothed current signal (32.1 nA) with the reference signal (2.1 nA) and calculates a difference of 30.0 nA. This difference is the "real-time common mode interference assessment result" caused by the penetration of electrochemically active substances (ascorbic acid, etc.) in the cell culture medium at the current time.

[0057] Obtaining the primary differential signal: the system performs dynamic baseline subtraction by subtracting the real-time common mode interference assessment result (30.0 nA) from the first smoothed current signal (658.4 nA) to eliminate environmental interference components, obtaining a primary differential signal of 628.4 nA representing the intensity of the net glucose oxidation-reduction reaction. Normalization processing: the system queries the calibration database and finds that the electrode has been continuously working for 5 days. Due to natural attenuation of enzyme activity, its current sensitivity is only 95.2% of the initial state. Therefore, the system retrieves the dynamically adjusted normalization coefficient as 1.05 (i.e., 1 / 0.952). The system multiplies the primary differential signal (628.4 nA) by this coefficient to finally obtain a standardized differential current signal of 659.8 nA.

[0058] Quality index calculation: the system analyzes the primary differential signal fluctuations in the past 5 seconds and calculates a fluctuation amplitude (standard deviation) of 1.2 nA. Comparing this amplitude with the current standardized differential current signal mean (659.8 nA), the calculated real-time signal quality index is 0.18% (i.e., the proportion of fluctuations). Classification and execution: the system compares this index with the preset threshold (excellent: <1%; suspicious: 1%-5%; poor: >5%). Since 0.18% is less than 1%, the system determines that the current quality classification result is "excellent". Execution action: the processor directly locks the value of 659.8 nA and transmits it to the subsequent environmental compensation module. Setting condition: if the calculated index is 3%, the system will determine it as "suspicious" and immediately trigger an instruction to discard the current data and force the front-end circuit to reacquire and calculate a new set of data within 500 milliseconds; if the index exceeds 5%, the system will call the "excellent" signal value (655.0 nA) stored in the last minute to fill in, preventing false output.

[0059] The sensitivity trend monitoring system recorded the normalized coefficient (1.05) used this time and formed a trend curve with the historical records (1.00, 1.01, 1.02, and 1.03 for the previous four days). Analysis showed that the coefficient showed a continuous unidirectional increase (meaning a unidirectional decrease in electrode sensitivity), but the current cumulative decrease (5%) has not yet exceeded the preset warning threshold (20%). Therefore, the system only updates the trend graph and does not generate a warning prompt, continuing to maintain normal operation.

[0060] 203. Obtain the current temperature signal and current pH signal of the cell culture medium, determine the temperature correction coefficient based on the current temperature signal, and determine the pH correction coefficient based on the current pH signal; Specifically, the current temperature reading is obtained through an independent temperature sensor immersed in the cell culture medium, and the current pH reading is obtained through an independent pH sensor immersed in the cell culture medium, serving as the original signals for the current temperature and pH, respectively. The original temperature signal is filtered to obtain a stable current temperature value; the original pH signal is filtered to obtain a stable current pH value; the stable current temperature value is matched and interpolated with a pre-stored first set of discrete temperature-compensation coefficient mapping data to output a temperature correction coefficient adapted to the current enzyme reaction kinetics; the stable current pH value is matched and interpolated with a pre-stored second set of discrete pH-activity coefficient mapping data to output a pH correction coefficient adapted to the current enzyme activity environment; the temperature correction coefficient and the pH correction coefficient are multiplied to generate an environmental comprehensive correction factor for subsequent concentration calculations.

[0061] It should be noted that the raw signal acquisition and filtering system samples the internal environment of the bioreactor through an integrated environmental monitoring module.

[0062] Signal Acquisition: The system activates the PT1000 platinum resistance temperature sensor and the industrial-grade online pH glass electrode immersed in the culture medium. Within the first 100 milliseconds of sampling, the temperature sensor transmits a set of slightly fluctuating readings (36.4°C, 36.6°C, 36.5°C), and the pH sensor transmits a set of readings (7.18, 7.22, 7.20). Filtering and Stabilization: The processor uses a recursive average filtering algorithm to process the raw data. After calculation and removal of transient fluctuations, the system locks in a stable current temperature value of 36.5°C and a stable current pH value of 7.20.

[0063] The temperature correction factor determination system accesses a "temperature-enzyme activity compensation table" in memory. This table stores discrete data points at 0.5 degree Celsius intervals. Data retrieval: The system retrieves the current temperature value of 36.5 degrees Celsius, which lies exactly between two pre-stored data points (or hits directly). In this example, the pre-stored data shows that an enzyme activity factor of 0.920 corresponds to 36.0 degrees Celsius, and an enzyme activity factor of 1.000 corresponds to 37.0 degrees Celsius. Matching calculation: Since 36.5 degrees Celsius lies exactly halfway between 36.0 and 37.0, the system performs a linear interpolation calculation. The result is a calculation of the ratio of the enzyme kinetics at the current temperature relative to the standard state (37 degrees Celsius). Result output: After the calculation, the system outputs a temperature correction factor of 0.960. This indicates that the enzyme's catalytic efficiency is approximately 96% of the standard state due to the current temperature being slightly below the optimal temperature.

[0064] The pH correction factor determination system accesses a "pH-enzyme activity mapping table" in memory. Data retrieval: The system looks up the adjacent data nodes for the current pH value of 7.20. The pre-stored data shows that an activity factor of 0.900 corresponds to pH 7.0, and an activity factor of 0.980 corresponds to pH 7.5. Matching calculation: The current value of 7.20 lies between 7.0 and 7.5, with a deviation of 0.2 from 7.0. The system calculates the activity increment based on a linear proportional relationship. The total span of the interval is 0.5, and the total difference in factors is 0.080. Therefore, for a deviation of 0.2, the factor should increase by 0.032. Result output: Adding the increment of 0.032 to the baseline value of 0.900, the system outputs a pH correction factor of 0.932.

[0065] The environmental comprehensive correction factor generation Finally, the processor performs a multiplication fusion calculation. Fusion calculation: Multiply the temperature correction factor (0.960) by the pH correction factor (0.932). Final output: The result of the calculation is 0.89472, which the system rounds to three decimal places to generate a final environmental comprehensive correction factor of 0.895.

[0066] 204、based on the differential current signal, the temperature correction factor, and the pH correction factor, determine an environment-compensated glucose concentration value, in a continuous measurement process, determine a baseline noise level based on the second current signal, and determine an electrode response state based on a dynamic response of the differential current signal; Specifically, the differential current signal is divided by an environmental comprehensive correction factor obtained by multiplying a temperature correction coefficient and a pH correction coefficient, to generate a corrected electrochemical response value; the corrected electrochemical response value is mapped to a concentration calibration result by querying a pre-stored calibration curve that associates electrochemical response values with glucose concentrations; the concentration calibration result is subjected to time series smoothing filtering to output a real-time glucose concentration value as a final measurement result; in the continuous measurement process, fluctuation data of the second current signal within a preset time period is recorded to form a baseline fluctuation sequence, and a standard deviation of the sequence is calculated as a noise level evaluation value representing measurement stability; the rising or falling process of the differential current signal after a known change in the glucose concentration in the cell culture medium is monitored, and a time required for the differential current signal to reach a stable state is recorded as a response lag evaluation value representing electrode performance; when the noise level evaluation value exceeds a preset stability threshold or the response lag evaluation value exceeds a preset speed threshold, it is determined that the electrode state is abnormal and a calibration trigger instruction is generated, which is used to start a calibration process to update the calibration curve.

[0067] Further, the method further comprises: an association parameter acquisition step of acquiring a dissolved oxygen concentration signal of the cell culture medium in real time, calculating a dissolved oxygen concentration change rate per unit time as an oxygen consumption rate evaluation value of cell respiration metabolism; a theoretical calculation step of calculating a theoretical oxygen consumption rate corresponding to the change rate of the real-time glucose concentration value based on a biochemical metrological relationship of cell metabolism; a cross verification step of comparing the measured oxygen consumption rate evaluation value with the calculated theoretical oxygen consumption rate, calculating a deviation percentage therebetween, and generating a metabolic correlation verification coefficient; a data confidence processing step of grading the confidence of the real-time glucose concentration value according to the numerical range of the metabolic correlation verification coefficient; when the confidence is graded as high, the concentration value is directly output; when the confidence is graded as low, a glucose consumption estimation value calculated independently from the oxygen consumption rate evaluation value is used as a basis to perform weighted fusion, to generate and output a metabolic verification corrected glucose concentration value; and a trend early warning step of continuously recording historical data of the noise level evaluation value and the response lag evaluation value, and generating a predictive maintenance signal indicating that the sensor performance is approaching failure when it is found through analysis that both the noise level evaluation value and the response lag evaluation value present a monotonic increasing trend and the cumulative amplitude of both exceeds a respective trend early warning threshold.

[0068] It should be noted that the concentration inversion and the environmental compensation microprocessor receive data first perform environmental compensation operation.

[0069] Electrochemical response correction: dividing the normalized differential current signal (659.8 nA) by the environmental comprehensive correction factor (0.895) obtains a corrected electrochemical response value of 737.2 nA under the condition of reducing to a standard temperature (37℃) and a standard pH (7.4).

[0070] Concentration calibration mapping: the system calls the calibration curve parameters stored in EEPROM: sensitivity slope S = 48.5 nA / (mmol / L), zero intercept nA. Calculate the instantaneous glucose concentration Substitute the numerical value for calculation: mmol / L.

[0071] Final output: perform 3-second moving average filtering on this instantaneous value, output the real-time glucose concentration value as 15.10 mmol / L.

[0072] Electrode state and stability evaluation While outputting the concentration, the background program diagnoses the physical state of the electrode. Baseline noise evaluation: the system intercepts the second current signal (background current) in the past 60 seconds, and calculates its standard deviation. The result shows that the standard deviation is 0.45 nanamps, which is far below the preset stability threshold (2.0 nanamps), and it is determined that the baseline noise level is normal. Response lag monitoring: the system detects that a feeding operation was performed 5 minutes ago, and the concentration produced a step. The record shows that the differential signal takes 42 seconds from the start of the step to reach 90% steady state. This value is lower than the preset speed threshold (60 seconds), and it is determined that the electrode response is sensitive and there is no passivation phenomenon. Therefore, the system does not trigger the calibration instruction.

[0073] Metabolic correlation verification and confidence level grading In order to prevent "false data" caused by sensor drift, the system introduces cell metabolic data for cross verification. Parameter acquisition: dissolved oxygen (DO) sensor feedback data shows that the current medium dissolved oxygen consumption rate (OUR) is 1.20 mmol / L / h. The system calculates that the measured glucose concentration change rate in the past 10 minutes is -0.80 mmol / L / h. According to the established metabolic flux ratio of this cell line (CHO cells) in the current growth stage (oxygen / sugar consumption molar ratio is 1.4:1), it is calculated that the corresponding theoretical oxygen consumption rate should be mmol / L / h. Cross verification: compare the measured oxygen consumption (1.20) with the theoretical oxygen consumption (1.12), the deviation is 6.7%. Confidence level processing: since the deviation is less than the preset 10% tolerance range, the system generates a "high confidence" rating and directly outputs the measurement result of 15.10 mmol / L. Scenario: if the deviation reaches 30% (i.e. confidence "low"), the system will calculate the glucose consumption according to the dissolved oxygen rate, generate a corrected value (14.95 mmol / L), and perform weighted fusion (weight each 50%) with the measured value, output the fused value to reduce the error caused by sensor drift.

[0074] ​​The trend alert system stores the calculated noise value (0.45 nA) into the historical database. Analyzing the data of the past 24 hours, although there is a slight fluctuation in the noise, it does not show a monotonic increasing trend, so no predictive maintenance signal is generated.

[0075] 205、When the baseline noise level exceeds the first preset threshold, or the electrode response state exceeds the second preset threshold, a calibration operation is triggered, and conversion parameters used to determine the glucose concentration value are updated based on the calibration operation, and the actual glucose concentration value is obtained based on the conversion parameters.

[0076] Specifically, the calibration phase: the first glucose sensing electrode and the second glucose sensing electrode are simultaneously switched from the cell culture medium to the calibration pool containing the standard glucose solution with a known concentration, and after the output signal stabilizes, the stable differential current signal under the standard solution is obtained to generate the standard solution response value; the parameter calculation phase: the standard solution response value is compared with the known concentration value, and the real-time sensitivity parameter reflecting the current electrode sensitivity is calculated, and the real-time offset parameter for correcting the measurement baseline is also calculated; the verification phase: the electrodes are switched to another verification solution with a known concentration, the real-time sensitivity parameter and the real-time offset parameter are used to calculate the verification concentration prediction value, and the prediction value is compared with the actual known concentration to generate a calibration verification deviation value; the update decision phase: if the calibration verification deviation value is less than the preset tolerance threshold, it is determined that the current calibration is effective, and the real-time sensitivity parameter and the real-time offset parameter are used to replace the original conversion parameters to complete the update; if the calibration verification deviation value is greater than or equal to the tolerance threshold, it is determined that the current calibration is invalid, a calibration failure alarm is generated, and the original conversion parameters remain unchanged; the recovery phase: after the parameter update or the alarm generation is completed, the electrodes are switched back to the cell culture medium from the calibration environment, and the system returns to the normal state of continuous glucose concentration measurement.

[0077] It should be noted that the abnormal trigger is based on the baseline noise level output by the system background monitoring step 204 after starting the calibration. At this time, the latest noise evaluation value calculated is 2.8 nA, which exceeds the first preset threshold (2.0 nA). The system determines that due to long-term immersion, a small amount of protein may be attached to the electrode surface, causing the signal fluctuation to increase, and a "non-emergency calibration program" is triggered. The automatic sampling arm simultaneously withdraws the first and second glucose sensing electrodes from the bioreactor, and after rapid washing with sterile water, it is immersed in a standard calibration solution containing 20.0 mmol / L glucose.

[0078] Signal acquisition and parameter calculation After 120 seconds of rest in the standard solution, the system readings are stable. Acquire response: The differential current signal between the first and second electrodes is now stable at 885.0 nanoamps. Calculate parameters: The system calls the background data for the second electrode in pure buffer (5.0 nanoamps) as the zero-point offset. Then, using the standard solution concentration (20.0 millimoles per liter) and the net current (885.0 nanoamps minus 5.0 nanoamps), the system makes a comparison. The calculation finds that the current actual electrode sensitivity has dropped from the initial 48.5 nanoamps per millimole per liter to 44.0 nanoamps per millimole per liter. The system stores this new sensitivity parameter and offset parameter.

[0079] Cross-validation and decision To ensure the accuracy of the new parameters, the electrode is moved to another cup of validation solution with a concentration of 5.0 millimoles per liter. Predictive estimation: The system reads the stable differential signal in the validation solution as 226.2 nanoamps. Using the newly calculated sensitivity (44.0) and offset (5.0), the system makes a reverse estimation that the current concentration should be 5.027 millimoles per liter. Deviation comparison: Comparing the predicted value (5.027) with the known true value (5.000), the system calculates the calibration validation deviation value as 0.54%. Update execution: Since the deviation is much smaller than the preset tolerance threshold (5%), the system determines that this calibration is valid. The microprocessor immediately overwrites the conversion parameters in the EEPROM with the new data replacing the old data.

[0080] Data summary and system recovery After the parameter update is complete, the electrode is reinserted into the cell culture medium. The system records the key data of this calibration for traceability and automatically resumes the continuous measurement mode. At this time, the output glucose concentration will be calculated based on the new sensitivity, eliminating the drift error.

[0081] The key data changes of this calibration operation are shown in Table 4 below: Table 4

[0082] 206、Based on the baseline noise level evaluation value and the electrode response state evaluation value, combined with the environmental comprehensive correction factor, a real-time data quality score is calculated; according to the numerical range of the real-time data quality score, the quality of the measurement data is divided into multiple pre-defined reliability levels, and the corresponding data state label is generated; the data state label is associated and packaged with the final output real-time glucose concentration value to form a concentration data package with reliability identification; when the data state label indicates high reliability, the system outputs the concentration data package with reliability identification at a normal frequency; when the data state label indicates low reliability, the system activates an external visual or audible warning signal while outputting the concentration data package.

[0083] Note that the real-time glucose concentration (15.10 mmol / L), baseline noise level (0.45 nA), and environmental comprehensive correction factor (0.895) calculated in the previous steps are taken into account. The following is a specific engineering example: The real-time data quality score calculation system microprocessor runs a multi-dimensional scoring algorithm, which takes 100 points as the benchmark, and deducts points based on the degree of deviation from the ideal state for each parameter. Noise evaluation: the current baseline noise is 0.45 nA. The system deducts 10 points for every 1.0 nA of noise. Therefore, this item deducts 4.5 points. State evaluation: the electrode response state evaluation value is normal (no significant hysteresis), the system gives full marks and does not deduct points. Environmental evaluation: the current environmental comprehensive correction factor is 0.895, meaning that the environmental conditions deviate from the standard state by about 10.5%. The system deducts 0.5 points for every 1% deviation. Therefore, this item deducts 5.25 points. Total score: deduct the above points from the base score of 100 points, and the system calculates the real-time data quality score as 90.25 points.

[0084] The quality grading and label generation system queries the pre-defined reliability grading table. Level 1 (high reliability): score greater than or equal to 90 points. Level 2 (medium reliability): score between 70 and 89 points. Level 3 (low reliability): score less than 70 points. Since 90.25 points meet the level 1 standard, the system generates a data status label with the content "Grade A - High Reliability". Three, the data packaging and output processor encodes and packages the concentration value "15.10 mmol / L" and the "Grade A" label in binary. The system sends this data packet to the host computer through the RS485 industrial bus at a frequency of once per second. The host screen displays the green number "15.10" with a full bar signal quality icon, indicating that the current data is highly reliable and can be directly used for feedback control of the feed pump.

[0085] The low-quality warning trigger mechanism (scenario assumption) sets that after two hours, the culture environment experiences a dramatic fluctuation, with the pH value dropping sharply, causing the environmental comprehensive correction factor to drop to 0.65 and the baseline noise to rise to 2.5 nA. At this time, the system recalculates the score: noise deduction 25 points, environmental deviation deduction 17.5 points, total score drops to 57.5 points. Label change: the system generates a "Grade C - Low Reliability" label. Warning activation: at the same time as outputting the data packet with the "low reliability" mark, the system immediately activates the I / O port, lights up the yellow warning light on the control cabinet, and sends a "sensor data confidence low, please manually review" text pop-up window to the operator, preventing automatic feeding based on incorrect data. The scoring logic is shown in the following Table 5: Table 5

[0086] Figure 3 is a structural schematic diagram of an artificial intelligence cell culture medium glucose concentration measuring device provided by an embodiment of the application. The device 300 can include a processor 301, a receiver 302, a transmitter 303, and a memory 304. The receiver 302, the transmitter 303, and the memory 304 are connected to the processor 301 through buses respectively. It should be noted that in some possible implementation manners, the processor 301 and the memory 304 can be integrated together.

[0087] The processor 301 includes one or more processing cores. The processor 301 executes the methods performed by the base station in the random access method provided by the embodiments of the application by running software programs and modules. The memory 304 can be used to store the software programs and modules. Specifically, the memory 304 can store an operating system 3041 and at least one application program module 3042 required by a function. The receiver 302 is configured to receive communication data sent by other devices, and the transmitter 303 is configured to send communication data to other devices.

[0088] The application further provides an artificial intelligence cell culture medium glucose concentration measuring device, which includes a memory and a processor. The memory stores computer readable instructions. When the computer readable instructions are executed by the processor, the processor executes the steps of the artificial intelligence cell culture medium glucose concentration measuring method in each of the above embodiments.

[0089] The application further provides a computer readable storage medium, which can be a non-volatile computer readable storage medium or a volatile computer readable storage medium. The computer readable storage medium stores instructions. When the instructions are run on a computer, the computer executes the steps of the artificial intelligence cell culture medium glucose concentration measuring method.

[0090] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device, and unit can refer to the corresponding processes in the foregoing method embodiments, which will not be described herein again.

[0091] The integrated unit, if implemented in the form of a software function unit and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application or the entire or part of the technical solutions that essentially contribute to the prior art can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0092] The above-described embodiments are merely used to illustrate the technical solutions of the present application, rather than limit the same; even though the present application has been described in detail with reference to the foregoing embodiments, those ordinarily skilled in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some of the technical features; and these modifications or replacements do not cause the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An artificial intelligence-based method for measuring glucose concentration in cell culture medium, characterized in that, include: A first glucose sensing electrode and a second glucose sensing electrode are set in a cell culture medium to obtain a first current signal generated by the first glucose sensing electrode in response to glucose in the culture medium, and a second current signal generated by the second glucose sensing electrode. A differential current signal is obtained based on the first current signal and the second current signal; The current temperature signal and current pH signal of the cell culture medium are acquired, and a temperature correction coefficient is determined based on the current temperature signal, and a pH correction coefficient is determined based on the current pH signal. Based on the differential current signal, the temperature correction coefficient, and the pH correction coefficient, the environmentally compensated glucose concentration value is determined. During continuous measurement, the baseline noise level is determined based on the second current signal, and the electrode response state is determined based on the dynamic response of the differential current signal. When the baseline noise level exceeds a first preset threshold, or the electrode response state exceeds a second preset threshold, a calibration operation is triggered, and the conversion parameters used to determine the glucose concentration value are updated based on the calibration operation, and the actual glucose concentration value is obtained based on the conversion parameters.

2. The artificial intelligence-based method for measuring glucose concentration in cell culture medium according to claim 1, characterized in that, include: The sensitive end of the first glucose sensing electrode is directly immersed in the cell culture medium, so that its glucose oxidase sensing layer is in direct contact with the culture medium to obtain the first current signal. The sensitive end of the second glucose sensing electrode is encapsulated in a sealed chamber made of a dialysis membrane. The sealed chamber is filled with pH buffer solution, and the dialysis membrane is in contact with the cell culture medium. This isolates the second glucose sensing electrode from macromolecules and cells in the culture medium, allowing ions and small molecules to pass through and acquiring a second current signal. Under the same constant potential bias conditions, the first current signal and the second current signal are acquired and recorded synchronously.

3. The artificial intelligence-based method for measuring glucose concentration in cell culture medium according to claim 2, characterized in that, include: The first current signal and the second current signal are processed to obtain a first smoothed current signal and a second smoothed current signal; The second smoothed current signal is compared with the reference signal determined by the background output of the second glucose sensing electrode in a glucose-free environment to obtain the real-time common-mode interference evaluation result. Based on the real-time common-mode interference assessment results, dynamic baseline subtraction is performed on the first smoothed current signal to obtain the primary differential signal. The primary differential signal is multiplied by a normalization coefficient dynamically adjusted based on the historical sensitivity data of the first glucose sensing electrode to obtain the differential current signal.

4. The artificial intelligence-based method for measuring glucose concentration in cell culture medium according to claim 3, characterized in that, Also includes: The real-time signal quality index is calculated based on the fluctuation amplitude of the primary differential signal and the average value of the differential current signal during the preset stable period. The real-time signal quality index is compared with a preset quality threshold to generate a quality rating result for judging the availability of the current differential signal. When the quality grading result indicates that the signal quality is questionable, a resampling and differential operation process is triggered on the first smoothed current signal and the second smoothed current signal, and the original signal is replaced with the newly generated differential current signal; when the quality grading result indicates that the signal quality is poor, the current differential signal is suspended, and the historical standardized differential current signal stored when it was most recently judged to be of good quality is called for subsequent steps. The adjustment history of the normalization coefficient is periodically recorded to form an electrode sensitivity decay trend graph. When the trend shows a continuous unidirectional decay and the cumulative magnitude exceeds the preset sensitivity decay warning threshold, an early warning prompt for electrode performance degradation is generated.

5. The artificial intelligence-based method for measuring glucose concentration in cell culture medium according to claim 3, characterized in that, include: Obtain the current temperature reading and the current pH reading, and use them as the original signals for the current temperature and pH, respectively. The original current temperature signal is processed to obtain the current temperature value, and the original current pH signal is processed to obtain the current pH value; The current temperature value is matched and interpolated with the pre-stored first set of discrete temperature-compensation coefficient mapping data to output the temperature correction coefficient. The current pH value is matched and interpolated with the pre-stored second set of discrete pH-activity coefficient mapping data to output the pH correction coefficient. The temperature correction coefficient is multiplied by the pH correction coefficient to generate the comprehensive environmental correction factor.

6. The artificial intelligence-based method for measuring glucose concentration in cell culture medium according to claim 5, characterized in that, include: Divide the differential current signal by the environmental comprehensive correction factor to generate the corrected electrochemical response value; By querying the pre-stored calibration curve that correlates electrochemical response values ​​with glucose concentration, the corrected electrochemical response values ​​are mapped to concentration calibration results. The concentration calibration results are subjected to time-series smoothing filtering to output real-time glucose concentration values; During continuous measurement, the fluctuation data of the second current signal within a preset time period are recorded to form a baseline fluctuation sequence, and the standard deviation of the sequence is calculated as a noise level assessment value characterizing the measurement stability. The differential current signal is monitored to observe the rise or fall of glucose concentration in the cell culture medium after a known change, and the time required to reach a steady state is recorded as a response hysteresis evaluation value characterizing electrode performance. When the noise level assessment value exceeds a preset stability threshold, or the response hysteresis assessment value exceeds a preset speed threshold, the electrode state is determined to be abnormal and a calibration trigger command is generated.

7. The artificial intelligence-based method for measuring glucose concentration in cell culture medium according to claim 6, characterized in that, Calculate instantaneous glucose concentration The instantaneous glucose concentration is filtered by a moving average to output the real-time glucose concentration value. ， in, This is the corrected electrochemical response value; The zero intercept; The slope is the sensitivity. This is an environmental comprehensive correction factor.

8. The artificial intelligence-based method for measuring glucose concentration in cell culture medium according to claim 6, characterized in that, Also includes: The dissolved oxygen concentration signal of the cell culture medium is acquired in real time, and the rate of change of dissolved oxygen concentration per unit time is calculated as an evaluation value of the oxygen consumption rate of cell respiration metabolism. Based on the biochemical stoichiometry of cell metabolism, the theoretical oxygen consumption rate corresponding to the rate of change of the real-time glucose concentration is calculated using the rate of change. The measured oxygen consumption rate assessment value is compared with the calculated theoretical oxygen consumption rate, and the percentage deviation between the two is calculated to generate a metabolic correlation verification coefficient. Based on the numerical range of the metabolic correlation verification coefficient, the confidence level of the real-time glucose concentration value is graded; when the confidence level is graded as high, the concentration value is directly output; when the confidence level is graded as low, the concentration value under the most recent high confidence state is used as the basis, and the glucose consumption estimate independently calculated from the oxygen consumption rate assessment value is weighted and fused to generate and output the glucose concentration value corrected by metabolic verification. Historical data of the noise level assessment value and the response hysteresis assessment value are continuously recorded. When the analysis finds that both of them show a monotonically increasing trend and the cumulative increase exceeds their respective trend warning thresholds, a predictive maintenance signal is generated.

9. The artificial intelligence-based method for measuring glucose concentration in cell culture medium according to claim 6, characterized in that, include: The first glucose sensing electrode and the second glucose sensing electrode are simultaneously switched from the cell culture medium to a calibration cell containing a standard glucose solution of known concentration. After the output signal stabilizes, a stable differential current signal under the standard solution is acquired, and a standard solution response value is generated. The response value of the standard solution is compared with the known concentration value to calculate the real-time sensitivity parameter reflecting the current electrode sensitivity and the real-time offset parameter used to correct the measurement baseline. The electrode is switched to another verification solution with a known concentration. Using the calculated real-time sensitivity parameter and the real-time offset parameter, the predicted value of the verification concentration is calculated and compared with the actual known concentration to generate a calibration verification deviation value. If the calibration verification deviation value is less than the preset tolerance threshold, the calibration is deemed valid, and the original conversion parameter is replaced with the real-time sensitivity parameter and the real-time offset parameter to complete the update; if the calibration verification deviation value is greater than or equal to the tolerance threshold, the calibration is deemed invalid, a calibration failure alarm is generated, and the original conversion parameter remains unchanged. After completing the parameter update or alarm generation, the electrode is switched back from the calibration environment to the cell culture medium, and the system returns to normal continuous glucose concentration measurement.

10. The artificial intelligence-based method for measuring glucose concentration in cell culture medium according to claim 1, characterized in that, Also includes: Based on the baseline noise level assessment value and the electrode response status assessment value, combined with the environmental comprehensive correction factor, a real-time data quality score is calculated; Based on the numerical range of the real-time data quality score, the quality of the measured data is divided into multiple predefined reliability levels, and corresponding data status labels are generated. The data status tag is associated with and encapsulated with the real-time glucose concentration value to form a concentration data packet with a reliability identifier; When the data status label indicates high reliability, the system outputs a concentration data packet with a reliability identifier at a normal frequency; when the data status label indicates low reliability, the system activates an alarm signal while outputting the concentration data packet.

Images