Optimizing embedded formulations for drug delivery
Disclosed herein is a method for execution by a drug delivery device for determining an optimal dose of a liquid drug for current cycle of a medication delivery algorithm, the method utilizing a stepwise evaluation of a model and a cost function across a coarse search space consisting of coarse discrete quantities of the drug and a refined search space consisting of refined discrete quantities of the drug.
1. A method comprising:
determining a search space of possible drug delivery amounts for a current cycle of medication delivery in an automated drug delivery system;
separating the search space into ranges of solutions;
narrowing the search space by determining a selected range in the range of solutions that minimizes a cost function;
separating the selected range into a list of refined values for a first delivery of the drug, the list of refined values being separated by a first fine increment;
evaluating a set of subsequent deliveries of the drug following the first deliveries of the drug in the list of refined values, the set of subsequent deliveries being separated by a second coarse increment; and
selecting a solution based on the set of subsequent deliveries that minimizes the cost function.
2. The method of claim 1 , wherein the first fine increment is smaller than the second coarse increment.
3. The method of claim 1 , wherein separating the search space into ranges of solutions is performed using a third coarse increment.
4. The method of claim 1 , wherein the search space of possible drug delivery amounts ranges from zero to a maximum quantity according to one or more safety constraints.
5. The method of claim 1 , wherein the drug is insulin and the automated drug delivery system is a part of a wearable drug delivery device.
6. The method of claim 1 , wherein determining the selected range comprises evaluating a recursive glucose model configured to predict glucose levels of a user of the automated drug delivery system in a predetermined number of future cycles based on delivery of a particular quantity of insulin.
7. The method of claim 6 , wherein the cost function calculates the cost for each possible glucose and insulin trajectory and determines the trajectory with a lowest cost.
8. A non-transitory computer-readable medium storing instructions configured to cause a processor to:
determine a search space of possible drug delivery amounts for a current cycle of medication delivery in an automated drug delivery system;
separate the search space into ranges of solutions;
narrow the search space by determining a selected range in the range of solutions that minimizes a cost function;
separate the selected range into a list of refined values for a first delivery of the drug, the list of refined values being separated by a first fine increment;
evaluate a set of subsequent deliveries of the drug following the first deliveries of the drug in the list of refined values, the set of subsequent deliveries being separated by a second coarse increment; and
select a solution based on the set of subsequent deliveries that minimizes the cost function.
9. The medium of claim 8 , wherein the first fine increment is smaller than the second coarse increment.
10. The medium of claim 8 , wherein separating the search space into ranges of solutions is performed using a third coarse increment.
11. The medium of claim 8 , wherein the search space of possible drug delivery amounts ranges from zero to a maximum quantity according to one or more safety constraints.
12. The medium of claim 8 , wherein the drug is insulin and the automated drug delivery system is a part of a wearable drug delivery device.
13. The medium of claim 8 , wherein determining the selected range comprises evaluating a recursive glucose model configured to predict glucose levels of a user of the automated drug delivery system in a predetermined number of future cycles based on delivery of a particular quantity of insulin.
14. The medium of claim 8 , wherein the cost function calculates the cost for each possible glucose and insulin trajectory and determines the trajectory with a lowest cost.
15. An automated drug delivery system comprising:
a reservoir configured to store a drug;
a drug delivery device configured to dispense the drug to a user;
a processor; and
a non-transitory computer-readable medium storing instructions configured to cause the processor to:
determine a search space of possible drug delivery amounts for a current cycle of medication delivery;
separate the search space into ranges of solutions;
narrow the search space by determining a selected range in the range of solutions that minimizes a cost function;
separate the selected range into a list of refined values for a first delivery of the drug, the list of refined values being separated by a first fine increment;
evaluate a set of subsequent deliveries of the drug following the first deliveries of the drug in the list of refined values, the set of subsequent deliveries being separated by a second coarse increment; and
select a solution based on the set of subsequent deliveries that minimizes the cost function.
16. The system of claim 15 , wherein the first fine increment is smaller than the second coarse increment.
17. The system of claim 15 , wherein separating the search space into ranges of solutions is performed using a third coarse increment.
18. The system of claim 15 , wherein the search space of possible drug delivery amounts ranges from zero to a maximum quantity according to one or more safety constraints.
19. The system of claim 15 , wherein the drug is insulin and the automated drug delivery system is a part of a wearable drug delivery device.
20. The system of claim 15 , wherein determining the selected range comprises evaluating a recursive glucose model configured to predict glucose levels of the user in a predetermined number of future cycles based on delivery of a particular quantity of insulin.