Systems, devices and methods to deliver biological factors or drugs to a subject
A method, computer implemented method and associated apparatus for the management of diabetes comprises utilizing zone model predictive control (Zone-MPC) to control delivery of an insulin or insulin analog within a zone of desired values.
1. A zone model predictive control (MPC) method of continuous monitoring and delivery of an insulin or insulin analog to a subject comprising:
utilizing a linear difference model in combination with a model predictive control (MPC) algorithm to control delivery of the insulin or insulin analog within a zone of desired values, wherein the linear difference model and MPC algorithm are implemented, using sufficiently programmed computer, according to steps:
obtaining insulin and meal data values and continuous glucose monitoring (CGM) data values for the subject;
mapping the data using transfer functions:
I
map
(
s
)
=
1
(
τ
1
s
+
1
)
(
τ
2
s
+
1
)
I
(
s
)
M
map
(
s
)
=
1
(
τ
3
s
+
1
)
(
τ
4
s
+
1
)
M
(
s
)
generating a linear difference model comprising a plurality of states:
G k+1 =α 1 G k + . . . +α p G k−p +β 11 I map,k−d 1 −1 + . . . +β 1q 1 I map,k−d 1 −q 1 + . . . +β 21 M map,k−d 2 −1 + . . . +β 2q 2 M map,k−d 2 −q 2
I map,k+1 =γ 1 I map,k +γ 2 I map,k−1 +γ 3 I k +γ 4 I k−1
M map,k+1 =δ 1 M map,k +δ 2 M map,k−1 +δ 3 M k +δ 4 M k−1 ,
obtaining a defined glycemic zone for the subject;
calculating a next administration dose and/or time of delivery of the insulin or insulin analog based on a predicted glycemic value for the subject using the linear difference model and the defined glycemic zone; and
delivering the insulin or insulin analog to the subject based upon the calculated next insulin administration, wherein:
I and M are insulin and meal data values expressed as amounts of insulin and carbohydrates, respectively,
I map and M map are new states representing mapped insulin and meal values, respectively,
τ 1-4 are time points of measurement,
s is a variable in unit of inverse time,
G is glucose blood concentration,
k is a time instant,
α 1 , β 11 , and β 21 are model coefficients,
d 1 and d 2 are insulin and meal time delays, respectively,
p, q 1 , and q 2 are orders of glucose, insulin, and meal, respectively, and
γ i and δ i represent weighting factors for insulin and meal, respectively, after being absorbed into the blood.
2. The method of claim 1 , wherein the next administration is performed by an insulin delivery pump or device.
3. The method of claim 1 , wherein the next administration is performed by a computerized implanted insulin pump.
4. The method of claim 1 , wherein the delivery of insulin varies temporally and by dose, the delivery of insulin is a continuous or discrete infusion, and the maximum glucose levels following each meal are maintained within the glycemic zone.
5. The method of claim 1 , wherein the CGM data are episotic glucose measurements or self-monitoring measurements.
6. The method of claim 1 , wherein the CGM data are sensor data, wherein the sensor is selected from an implanted glucose sensor, an optical glucose sensor, enzymatic glucose sensor and a finger stick glucose sensor.
7. The method of claim 1 wherein the Zone-MPC is based on linear difference equations initiated by an auto-regression with exogenous input (ARX) model selected by its prediction competency and then re-optimized to improve its prediction abilities.
8. The method of claim 1 wherein open loop data collection of insulin meals is obtained by feeding the subject followed by bolus injection of insulin.
9. The method of claim 1 wherein an automatic model identification formula is identified from a plurality of different autoregression exogenous (ARX) input models based upon the measured and mapped insulin meals and glucose measurements.