Neurostimulator devices using a machine learning method implementing a gaussian process optimization
First Claim
Patent Images
1. A neurostimulator device comprising:
- a stimulation assembly connectable to a plurality of electrodes, wherein the plurality of electrodes are configured to stimulate a spinal cord using an applied complex stimulation pattern;
one or more sensors configured to measure a response related to stimulation of the spinal cord; and
at least one processor configured to modify the applied complex stimulation pattern deliverable by the plurality of electrodes to create a modified complex stimulation pattern for subsequent stimulation of the spinal cord by integrating data from the one or more sensors and performing a machine learning method implementing a Gaussian Process Optimization (“
GPO”
) relation that describes a predicted mean and a variance of a motor performance function for a plurality of candidate complex stimulation patterns, including the applied complex stimulation pattern, based on at least on one of (i) previous data from the one or more sensors, and (ii) data derived in a previous stimulation study,wherein the GPO relation includes an upper confidence bound rule for applying a weight to modify the applied complex stimulation pattern based on a number of times data is received from the one or more sensors regarding stimulation of the spinal cord, andwherein the upper confidence bound rule modifies the applied complex stimulation pattern through convergence of the GPO relation toward an optimal candidate complex stimulation pattern.
8 Assignments
0 Petitions
Accused Products
Abstract
Neurostimulator devices are described comprising: a stimulation assembly connectable to a plurality of electrodes, wherein the plurality of electrodes are configured to stimulate a spinal cord; one or more sensors; and at least one processor configured to modify at least one complex stimulation pattern deliverable by the plurality of electrodes by integrating data from the one or more sensors and performing a machine learning method implementing a Gaussian Process Optimization on the at least one complex stimulation pattern. Methods of use are also described.
232 Citations
20 Claims
-
1. A neurostimulator device comprising:
-
a stimulation assembly connectable to a plurality of electrodes, wherein the plurality of electrodes are configured to stimulate a spinal cord using an applied complex stimulation pattern; one or more sensors configured to measure a response related to stimulation of the spinal cord; and at least one processor configured to modify the applied complex stimulation pattern deliverable by the plurality of electrodes to create a modified complex stimulation pattern for subsequent stimulation of the spinal cord by integrating data from the one or more sensors and performing a machine learning method implementing a Gaussian Process Optimization (“
GPO”
) relation that describes a predicted mean and a variance of a motor performance function for a plurality of candidate complex stimulation patterns, including the applied complex stimulation pattern, based on at least on one of (i) previous data from the one or more sensors, and (ii) data derived in a previous stimulation study,wherein the GPO relation includes an upper confidence bound rule for applying a weight to modify the applied complex stimulation pattern based on a number of times data is received from the one or more sensors regarding stimulation of the spinal cord, and wherein the upper confidence bound rule modifies the applied complex stimulation pattern through convergence of the GPO relation toward an optimal candidate complex stimulation pattern. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13)
-
-
14. A method of improving neurologically derived paralysis, the method comprising:
-
applying a first complex stimulation pattern to a spinal cord of a patient using a neurostimulator device that includes a stimulation assembly connectable to a plurality of electrodes for stimulating the spinal cord; measuring with one or more sensors a response related to stimulation of the spinal cord; and modifying, via a processor, the first complex stimulation pattern to create a second complex stimulation pattern for subsequent stimulation of the spinal cord by integrating data from the one or more sensors and performing a machine learning method implementing a Gaussian Process Optimization (“
GPO”
) relation that describes a predicted mean and a variance of a motor performance function for a plurality of candidate complex stimulation patterns, including the first complex stimulation pattern, based on at least on one of (i) previous data from the one or more sensors, and (ii) data derived in a previous stimulation study,wherein the GPO relation includes an upper confidence bound rule for applying a weight to modify the first complex stimulation pattern based on a number of times data is received from the one or more sensors regarding stimulation of the spinal cord, and wherein the upper confidence bound rule modifies the first complex stimulation pattern through convergence of the GPO relation toward an optimal candidate complex stimulation pattern. - View Dependent Claims (15, 16, 17, 18, 19, 20)
-
Specification