Probability based controller gain

• US 10,478,556 B2
• Filed: 03/04/2016
• Issued: 11/19/2019
• Est. Priority Date: 03/04/2016
• Status: Active Grant

• Current Assignee: Roche Diabetes Care Inc. (Roche Holding AG)
• Original Assignee: Roche Diabetes Care Inc. (Roche Holding AG)
• Inventors: Buck, Jr., Harvey B., Duke, David L.
• Primary Examiner(s): Riggs, II, Larry D
• Application Number:
  US15/061,202
• Publication Number:
  US 20170252513A1
• Time in Patent Office: 1,355 Days
• Field of Search: None
• CPC Class Codes: A61B 5/14532 for measuring glucose, e.g. by tissue impedance measurement
• CPC Class Codes: A61B 5/7221 Determining signal validity, reliability or quality A61B5/7207 and A61B5/7217 take precedence
• CPC Class Codes: A61M 2205/3584 using modem, internet or bluetooth
• CPC Class Codes: A61M 2205/3592 using telemetric means, e.g. radio or optical transmission
• CPC Class Codes: A61M 2205/52 with memories providing a history of measured variating parameters of apparatus or patient
• CPC Class Codes: A61M 2230/005 Parameter used as control input for the apparatus
• CPC Class Codes: A61M 2230/201 Glucose concentration
• CPC Class Codes: A61M 5/1723 using feedback of body parameters, e.g. blood-sugar, pressure measurement of body parameters A61B5/00
• CPC Class Codes: G05B 13/024 in which a parameter or coefficient is automatically adjusted to optimise the performance
• CPC Class Codes: G06N 7/01 Probabilistic graphical models, e.g. probabilistic networks

Abstract

Methods and systems are disclosed for estimating a glucose level of a person having diabetes and selecting automatically open-loop and closed-loop control for a connected therapy delivery device. The method may comprise analyzing measured glucose results and corresponding impedance values received from a glucose sensor coupled to the person with a probability analysis tool implemented by a microcontroller to determine a total quality score that is based on the minimum constraint of a probability of glucose sensor accuracy determined measured glucose results and a probability of sensing quality determined from the impedance values. The microcontroller may estimate the glucose level of the person with a recursive filter based on the plurality of measured glucose results weighted with the total quality score and select automatically either open-loop control or closed-loop control for the connected therapy delivery device based on the value of the total quality score.

Claims

1. A method for estimating glucose level of a person having diabetes and selecting automatically open-loop and closed-loop control for a connected therapy delivery device, the method comprising:

receiving into a blood glucose management device having a microcontroller a plurality of measured glucose results and corresponding impedance values from a glucose sensor coupled to the person;

using the microcontroller to analyze the plurality of measured glucose results and corresponding impedance values with a probability analysis tool configured to determine a total quality score Q_total that is based on the minimum constraint of a set of quality metrics which comprises a probability of glucose sensor accuracy P_A based on the plurality of measured glucose results and a probability of sensing quality Q_ac based on the impedance values;

using the microcontroller to estimate the glucose level of the person with a recursive filter configured to estimate the glucose level based on the plurality of measured glucose results weighted with the total quality score Q_{total; and}

switching, via the microcontroller, the connected therapy delivery device automatically between open-loop control and closed-loop control based on the value of the total quality score Q_total .

2. The method of claim 1, wherein the blood glucose management device comprises a blood glucose meter, and the microcontroller is coupled to the glucose sensor, a cellular phone, a smart phone, a personal digital assistant, a personal computer, or a computer server.

3. The method of claim 1, wherein the glucose sensor comprises a continuous glucose monitoring system physically coupled to the person having diabetes and configured to automatically measure the glucose level of the person, and the connected therapy delivery device is an infusion pump, wherein the infusion pump is wired or wireless connected to the blood glucose management device.

4. The method of claim 1, wherein the plurality of measured glucose results comprises periodic glucose measurements taken in period selected from the range of every 1 to 10 minutes.

5. The method of claim 1, wherein the probability analysis tool is configured to determine the total quality score Q_total further based on at least one of:

when the person eats a meal;

when the person exercises; and

when insulin is delivered to the person.

6. The method of claim 1, wherein the probability analysis tool comprises a hidden Markov model, wherein:

the hidden Markov model has two states;

a first state S_A indicating the glucose sensor is accurate, and

a second state S_I indicating the glucose sensor is inaccurate; and

the hidden Markov model is configured to determine the probability of glucose sensor accuracy P_A, based on a state of the hidden Markov model and the plurality of measured glucose results.

7. The method of claim 6, wherein probability of the glucose sensor being in the second state S_I is based on a most-recent measured glucose result, a most-recent change in the plurality of measured glucose results, or a combination thereof.

8. The method of claim 6, wherein probability of the hidden Markov model transitioning from the first state S_A to second the state S_I is

9. The method of claim 6, wherein probability of the hidden Markov model remaining in the second state S_I is

10. The method of claim 6, wherein the probability of glucose sensor accuracy P_A is

1−

[(P<sub>A →

I</sub>×

S_A )+(P<sub>I→

I</sub>×

S_I)], wherein

S_A=1 when the hidden Markov model is in the first state S_A and S_A=0 otherwise,

S_I =1 when the hidden Markov model is in the second state S_I and S_I=0 otherwise,

P<sub>A→

I</sub> is a probability of transitioning from the first state S_A to the second state S_I, and

P<sub>I→

I</sub> is a probability of remaining in the second state S_I when in the second state S_I.

11. The method of claim 1, wherein the recursive filter is a Kalman filter or an Extended Kalman filter.

12. The method of claim 11 further comprising using the microcontroller to predict a future glucose level of the person with the Kalman filter or the Extended Kalman filter, wherein:

the Kalman filter or the Extended Kalman filter comprises a prediction step and a measurement step; and

the prediction step is performed one or more times in order to predict the future glucose level of the person.

13. The method of claim 11, wherein the Kalman filter or the Extended Kalman filter comprises a state vector, x_k=[g ġ

{umlaut over (g)}]^T, where k represents a kth sample of the state vector, g represents the estimated glucose level of the person, ġ

represents a first derivative of g, and {umlaut over (g)} resents a second derivative of g.

14. The method of claim 13, wherein using the microcontroller to estimate the glucose level of the person comprises determining the state vector

x_k={circumflex over (x)}_k+K_k(z_k−

h({circumflex over (x)}_k))Q_total, where {circumflex over (x)}

15. The method of claim 14, wherein the Kalman gain K_k is based on a measurement uncertainty R_k such that the measurement uncertainty R_k is variable and is based on the probability of sensor accuracy.

16. The method of claim 15, wherein the measurement uncertainty R_k is:

R_k=max[min((σ

<sub>z−

x</sub>²)^α

, σ

_max²(1−

Q_total)σ

_max²]+σ

_cgm²,

where (σ

<sub>z−

x</sub>²)^α

, represents is the standard deviation of the difference between raw CGM measurements (z) and filtered CGM measurements (x) over a last sampling period, σ

_max² represents a maximum physiological variance for glucose in a person with poorly controlled diabetes, Q_total is the total quality score, and σ

_cgm² represents a minimum variance for the raw CGM measurements.

17. The method of claim 16, wherein the recursive filter is configured to estimate the glucose level of the person further based on at least one of:

when the person eats a meal;

when the person exercises; and

when insulin is delivered to the person.

18. The method of claim 1 further comprising using the microcontroller to predict a future glucose level of the person with a regression analysis tool configured to predict the future glucose level based on the estimated glucose level of the person from the recursive filter.

19. The method of claim 18, wherein the regression analysis tool comprises a Gaussian process regression analysis.

20. The method of claim 19, wherein the Gaussian process regression analysis comprises a training algorithm configured to learn one or more characteristics of the person related to the glucose level of the person.

43. The method of claim 1 further comprises determining the probability of sensing quality Q_ac based on the impedance values via utilizing the equation:

Q_AC=0.5*(1−

tan h(a*log(b*E)),

where E is a deviation sum from predetermined values, and parameters a and b are preselected to choose a location of a 0.5 value (1/b) and a rate of change at a 0.5 value (a).

21. An apparatus for estimating a glucose level of a person having diabetes and selecting automatically open-loop and closed-loop control for a connected therapy delivery device, the apparatus comprising a microcontroller and a display, wherein:

the microcontroller is configured to;

receive a plurality of measured glucose results and corresponding impedance values from a glucose sensor coupled to the person,

analyze the plurality of measured glucose results and corresponding impedance values with a probability analysis tool configured to determine a total quality score Q_total that is based on the minimum constraint of a set of quality metrics which comprises a probability of glucose sensor accuracy P_A based on the plurality of measured glucose results and a probability of sensing quality Q_ac based on the impedance values,

estimate the glucose level of the person with a recursive filter configured to estimate the glucose level based on the plurality of measured glucose results weighted with the total quality score Q_total,

switch the connected therapy delivery device automatically between open-loop control and closed-loop control based on the value of the total quality score Q_total, and

transmit to the display information related to the estimate of the glucose level of the person and the control switch of the therapy delivery device.

22. The apparatus of claim 21, wherein the microcontroller is coupled to the glucose sensor, a cellular phone, a smart phone, a personal digital assistant, a personal computer, or a computer server.

23. The apparatus of claim 21, wherein the glucose sensor comprises a continuous glucose monitoring system physically coupled to the person having diabetes and configured to automatically measure the glucose level of the person, and the connected therapy delivery device is a infusion pump, wherein the infusion pump is wired or wireless connected to the blood glucose management device.

24. The apparatus of claim 21, wherein the plurality of measured glucose results comprises periodic glucose measurements taken in period selected from the range of every 1, 5 or 10 minutes.

25. The apparatus of claim 21, wherein the probability analysis tool is configured to determine the total quality score Q_total further based on at least one of:

when the person eats a meal;

when the person exercises; and

when insulin is delivered to the person.

26. The apparatus of claim 21, wherein the probability analysis tool comprises a hidden Markov model, wherein:

the hidden Markov model has two states;

a first state S_A indicating the glucose sensor is accurate, and

a second state S_I indicating the glucose sensor is inaccurate; and

the hidden Markov model is configured to determine the probability of glucose sensor accuracy P_A , based on a state of the hidden Markov model and the plurality of measured glucose results.

27. The apparatus of claim 26, wherein probability of the glucose sensor being in the second state S_I is based on a most-recent measured glucose result, a most-recent change in the plurality of measured glucose results, or a combination thereof.

28. The apparatus of claim 26, wherein probability of the hidden Markov model transitioning from the first state S_A to second the state S_I is

29. The apparatus of claim 26, wherein probability of the hidden Markov model remaining in the second state S_I is

30. The apparatus of claim 26, wherein the probability of glucose sensor accuracy P_A is

1−

[(P<sub>A→

I</sub>×

S_A)+(P<sub>I→

I</sub>×

S_I)], wherein

S_A=1 when the hidden Markov model is in the first state S_A and S_A=0 otherwise,

S_I=1 when the hidden Markov model is in the second state S_I and S_I=0 otherwise,

P<sub>A→

I</sub> is a probability of transitioning from the first state S_A to the second state S_I, and

P<sub>I→

I</sub> is a probability of remaining in the second state S_I when in the second state S_I.

31. The apparatus of claim 21, wherein the recursive filter is a Kalman filter or an Extended Kalman filter.

32. The apparatus of claim 31, wherein the microcontroller predicts a future glucose level of the person with the Kalman filter or the Extended Kalman filter, wherein:

the Kalman filter or the Extended Kalman filter comprises a prediction step and a measurement step; and

the prediction step is performed one or more times in order to predict the future glucose level of the person.

33. The apparatus of claim 31, wherein the Kalman filter or the Extended Kalman filter comprises a state vector, x_k=[g ġ

{umlaut over (g)}]^T, where k represents a kth sample of the state vector, g represents the estimated glucose level of the person, ġ

represents a first derivative of g, and {umlaut over (g)} represents a second derivative of g.

34. The apparatus of claim 33, wherein the microcontroller estimates the glucose level of the person via state vector

x_k={circumflex over (x)}_k+K_k(z_k−

h({circumflex over (x)}_k))Q_total,

where

35. The apparatus of claim 34, wherein the Kalman gain K_k is based on a measurement uncertainty R_k such that the measurement uncertainty R_k is variable and is based on the probability of sensor accuracy.

36. The apparatus of claim 35, wherein the measurement uncertainty R_k is:

R_k=max[min((σ

<sub>z−

x</sub>²)^α

, σ

_max²), (1−

Q_total)σ

_max²]+σ

_cgm²,

where σ

<sub>z−

x</sub>²)^α

, represents is the standard deviation of the difference between raw CGM measurements (z) and filtered CGM measurements (x) over a last sampling period, σ

_max² represents a maximum physiological variance for glucose in a person with poorly controlled diabetes, Q_total is the total quality score, and σ

_cgm² represents a minimum variance for the raw CGM measurements.

37. The apparatus of claim 36, wherein the recursive filter is configured to estimate the glucose level of the person further based on at least one of:

when the person eats a meal;

when the person exercises; and

when insulin is delivered to the person.

38. The apparatus of claim 21, wherein the microcontroller predicts a future glucose level of the person with a regression analysis tool configured to predict the future glucose level based on the estimated glucose level of the person from the recursive filter.

39. The apparatus of claim 38, wherein the regression analysis tool comprises a Gaussian process regression analysis.

40. The apparatus of claim 39, wherein the Gaussian process regression analysis comprises a training algorithm configured to learn one or more characteristics of the person related to the glucose level of the person.

44. The apparatus of claim 21, wherein the microcontroller is configured to determine the probability of sensing quality Q_ac based on the impedance values via the equation:

Q_AC=0.5*(1−

tan h(a*log(b*E)),

where E is a deviation sum from predetermined values, and parameters a and b are preselected to choose a location of a 0.5 value (1/b) and a rate of change at a 0.5 value (a).

41. A method for estimating glucose level of a person having diabetes, the method comprising:

collecting, via a blood glucose management device having a microcontroller, a plurality of measured glucose results and corresponding impedance values from a glucose sensor coupled to the person;

using the microcontroller to analyze the plurality of measured glucose results and corresponding impedance values with a probability analysis tool configured to determine a total quality score Q_total that is based on the minimum constraint of a set of quality metrics which comprises a probability of glucose sensor accuracy P_A based on the plurality of measured glucose results and a probability of sensing quality Q_ac based on the impedance values;

using the microcontroller to estimate the glucose level of the person with a recursive filter configured to estimate the glucose level based on the plurality of measured glucose results weighted with the total quality score Q_total; and

alerting via the microcontroller updating a display of the management device with a warning upon the estimated glucose level of the person being one of hypoglycemic and hyperglycemic.

42. An apparatus for estimating a glucose level of a person having diabetes, the apparatus comprising a microcontroller and a display, wherein:

the microcontroller is configured to;

collect a plurality of measured glucose results and corresponding impedance values from a glucose sensor coupled to the person,

analyze the plurality of measured glucose results and corresponding impedance values with a probability analysis tool configured to determine a total quality score Q_total that is based on the minimum constraint of a set of quality metrics which comprises a probability of glucose sensor accuracy P_A based on the plurality of measured glucose results and a probability of sensing quality Q_ac based on the impedance values,

estimate the glucose level of the person with a recursive filter configured to estimate the glucose level based on the plurality of measured glucose results weighted with the total quality score Q_total,

transmit to the display information related to the estimate of the glucose level of the person, and

alert via the display with a warning upon the estimated glucose level of the person being one of hypoglycemic and hyperglycemic.