Optional sensor calibration in continuous glucose monitoring
A method for optional external calibration of a calibration-free glucose sensor uses values of measured working electrode current (Isig) and EIS data to calculate a final sensor glucose (SG) value. Counter electrode voltage (Vcntr) may also be used as an input. Raw Isig and Vcntr values may be preprocessed, and low-pass filtering, averaging, and/or feature generation may be applied. SG values may be generated using one or more models for predicting SG calculations. When an external blood glucose (BG) value is available, the BG value may also be used in calculating the SG values. A SG variance estimate may be calculated for each predicted SG value and modulated, with the modulated SG values then fused to generate a fused SG. A Kalman filter, as well as error detection logic, may be applied to the fused SG value to obtain a final SG, which is then displayed to the user.
1. A method for optional external calibration of a calibration-free glucose sensor for measuring a level of glucose in a body of a user, the glucose sensor including physical sensor electronics, a microcontroller, and a working electrode, the method comprising:
periodically measuring, by the physical sensor electronics, electrode current (Isig) signals for the working electrode;
performing, by the microcontroller, an Electrochemical Impedance Spectroscopy (EIS) procedure to generate EIS-related data for the working electrode;
based on the Isig signals and EIS-related data and a plurality of calibration-free sensor glucose (SG)-predictive models, calculating, by the microcontroller, a respective electrochemical SG value for each of the calibration-free SG-predictive models;
determining, by the microcontroller, whether an external blood glucose (BG) value is available;
adjusting, by the microcontroller, each respective electrochemical SG value for each of the calibration-free SG-predictive models based on the external BG value to provide respective adjusted electrochemical SG values;
fusing, by the microcontroller, the respective adjusted electrochemical SG values to obtain a single, fused electrochemical SG value to be displayed to the user;
in response to determining that the external BG value is available, comparing the external BG value to the fused electrochemical SG value to determine whether a difference between the fused electrochemical SG value and the external BG value exceeds a threshold;
in response to a determination that the difference between the fused electrochemical SG value and the external BG value exceeds the threshold, modulating the fused electrochemical SG value by scaling the respective electrochemical SG value by a modulation factor that is proportional to a ratio of the external BG value to the respective electrochemical SG value; and
filtering the fused electrochemical SG value based on a first set of measurement functions for when the external BG value is available and on a second set of measurement functions for when the external BG value is not available.
2. The method of claim 1 , further comprising:
measuring, by the physical sensor electronics, voltage values of a counter electrode (Vcntr) of the glucose sensor.
3. The method of claim 2 , further comprising:
preprocessing, by the microcontroller, the Isig signals and Ventr values prior to calculation of the respective electrochemical SG values.
4. The method of claim 3 , further comprising:
applying a low-pass filter to the Isig signals.
5. The method of claim 3 , wherein the preprocessing includes down-sampling Isig signals.
6. The method of claim 1 , wherein the plurality of calibration-free SG-predictive models are machine learning models.
7. The method of claim 6 , wherein the machine learning models include at least one of a genetic programming algorithm, a regression decision tree, or a bagged decision tree.
8. The method of claim 1 , wherein the plurality of calibration-free SG-predictive models are analytical models.
9. The method of claim 1 , further comprising calculating, by the microcontroller, a SG variance estimate for each respective electrochemical SG value,
wherein each SG variance estimate for each respective electrochemical SG value is calculated empirically from training data.
10. The method of claim 1 , wherein filtering the fused electrochemical SG value includes applying an unscented Kalman filter to the fused electrochemical SG value.
11. A glucose sensor comprising:
a working electrode;
physical sensor electronics configured to periodically measure electrode current (Isig) signals for the working electrode; and
a microcontroller configured to:
perform an Electrochemical Impedance Spectroscopy (EIS) procedure to generate EIS-related data for the working electrode;
based on the Isig signals and EIS-related data and a plurality of calibration-free sensor glucose (SG)-predictive models, calculate a respective electrochemical SG value for each of the calibration-free SG-predictive models;
determine whether an external blood glucose (BG) value is available;
adjust each respective electrochemical SG value for each of the calibration-free SG-predictive models based on the external BG value to provide respective adjusted electrochemical SG values;
fuse the respective adjusted electrochemical SG values to obtain a single, fused electrochemical SG value to be displayed to a user;
in response to a determination that the external BG value is available, compare the external BG value to the fused electrochemical SG value to determine whether a difference between the fused electrochemical SG value and the external BG value exceeds a threshold;
in response to a determination that the difference between the fused electrochemical SG value and the external BG value exceeds the threshold, modulate the fused electrochemical SG value by scaling the respective electrochemical SG value by a modulation factor that is proportional to a ratio of the external BG value to the respective electrochemical SG value; and
filter the fused electrochemical SG value based on a first set of measurement functions for when the external BG value is available and on a second set of measurement functions for when the external BG value is not available.
12. The glucose sensor of claim 11 , wherein the physical sensor electronics further measures voltage values of a counter electrode (Vontr) of the glucose sensor.
13. The glucose sensor of claim 12 , wherein the microcontroller is further configured to preprocess the Isig signals and Ventr values prior to calculation of the respective electrochemical SG values.
14. The glucose sensor of claim 13 , wherein the microcontroller is further configured to apply a low-pass filter to the Isig signals.
15. The glucose sensor of claim 11 , wherein the plurality of calibration-free SG-predictive models are machine learning models.
16. The glucose sensor of claim 15 , wherein the machine learning models include at least one of a genetic programming algorithm, a regression decision tree, or a bagged decision tree.
17. The glucose sensor of claim 11 , wherein the plurality of calibration-free SG-predictive models are analytical models.
18. The glucose sensor of claim 11 , wherein the microcontroller is further configured to calculate a SG variance estimate for each respective electrochemical SG value, and
wherein each SG variance estimate for each respective electrochemical SG value is calculated empirically from training data.
19. The glucose sensor of claim 11 , wherein the microcontroller is configured to filter the fused electrochemical SG value by applying an unscented Kalman filter to the fused electrochemical SG value.
20. A non-transitory computer-readable medium for generating recommendations for optional external calibration of a calibration-free glucose sensor for measuring a level of glucose in a body of a user, the non-transitory computer-readable medium comprising instructions which, when executed by one or more processors, cause operations comprising:
periodically measuring electrode current (Isig) signals for a working electrode of the calibration-free glucose sensor;
performing an Electrochemical Impedance Spectroscopy (EIS) procedure to generate EIS-related data for the working electrode;
based on the Isig signals and EIS-related data and a plurality of calibration-free SG-predictive models, calculating a respective electrochemical sensor glucose (SG) value for each of the calibration-free SG-predictive models;
determining whether an external blood glucose (BG) value is available;
adjusting each respective electrochemical SG value for each of the calibration-free SG-predictive models based on the external BG value to provide adjusted electrochemical SG values;
fusing the respective adjusted electrochemical SG values to obtain a single, fused electrochemical SG value to be displayed to the user;
in response to determining that the external BG value is available, comparing the external BG value to the fused electrochemical SG value to determine whether a difference between the fused electrochemical SG value and the external BG value exceeds a threshold;
in response to a determination that the difference between the fused electrochemical SG value and the external BG value exceeds the threshold, modulating the fused electrochemical SG value by scaling the respective electrochemical SG value by a modulation factor that is proportional to a ratio of the external BG value to the respective electrochemical SG value; and
filtering the fused electrochemical SG value based on a first set of measurement functions for when the external BG value is available and on a second set of measurement functions for when the external BG value is not available.