Task-outcome error signals and their use in brain-machine interfaces
A brain machine interface (BMI) for improving a performance of a subject is provided. The BMI has two decoders that act in real-time and in parallel to each other. The first decoder is for intention execution of a subject's intention. The second decoder is for error detection in a closed-loop error fashion with the first detector and to improve the performance of the first detector. Embodiments of this invention may enable an entirely new way to substantially increase the performance and robustness, user experience, and ultimately the clinical viability of BMI systems.
1. A brain-machine interface, comprising:
at least one microelectrode array configured to be implanted into a subject's brain in the motor cortex region;
a virtual keyboard comprising a virtual cursor; and
a computing component, comprising:
an intention decoder for decoding the subject's intention from a first neural signal; and
an error signal decoder for decoding error signals from a second neural signal;
where the intention decoder is directed to:
obtain the first neural signal comprising at least one intracortical spike signal from the at least one microelectrode array describing neural activity representative of an intended action by the subject;
infer the intended action from the first neural signal; and
initiate the performance of the intended action using the virtual keyboard; and
where the error signal decoder is directed to:
obtain the second neural signal comprising at least one intracortical spike signal from the at least one microelectrode array describing neural activity in the subject's brain;
reduce the dimensionality of the second neural signal;
classify the second neural signal as an error signal based on the reduced second neural signal;
delay selection of a key in the virtual keyboard when the error signal decoder detects an error signal; and
cancel the performance of the intended action by the virtual keyboard.
2. The brain-machine interface of claim 1 , wherein the at least one microelectrode array comprises a first microelectrode array and a second microelectrode array;
wherein the first neural signal is obtained from the first microelectrode array; and
wherein the second neural signal is obtained from the second microelectrode array.
3. The brain-machine interface of claim 1 , wherein the dimensionality of the second neural signal is reduced using a principal component analysis; and the second neural signal is classified using a linear support vector machine.
4. The brain-machine interface of claim 1 , wherein the cancelled action is to select a key in the virtual keyboard.
5. The brain-machine interface of claim 1 , wherein the intention decoder is updated based on the classifications made by the error signal decoder.
6. A method for utilizing error signals to increase accuracy in a brain-machine interface system, comprising:
obtaining a first neural signal comprising at least one intracortical spike signal from at least one microelectrode array in the motor cortex of a subject's brain describing neural activity representative of an intended action by the subject using an intention decoder;
inferring the intended action from the first neural signal using the intention decoder;
initiating the performance of the intended action using a virtual keyboard using the intention decoder, the virtual keyboard comprising a virtual cursor; and
obtaining a second neural signal comprising at least one intracortical spike signal from the at least one microelectrode array describing neural activity in the subject's brain activity using an error signal decoder;
reducing the dimensionality of the second neural signal using the error signal decoder;
classifying the second neural signal as an error signal based on the reduced second neural signal using the error signal decoder;
delaying selection of a key in the virtual keyboard when the error signal decoder detects an error signal and
cancelling the performance of the intended action by the virtual keyboard using the error signal decoder.
7. The method for utilizing error signals to increase accuracy in a brain-machine interface system of claim 6 , wherein at least one microelectrode array of the at least one microelectrode array is implanted into the subject's brain at the primary motor cortex region.
8. The method for utilizing error signals to increase accuracy in a brain-machine interface system of claim 6 , wherein the at least one microelectrode array comprises a first microelectrode array and a second microelectrode array;
wherein the first neural signal is obtained from the first microelectrode array; and
wherein the second neural signal is obtained from the second microelectrode array.
9. The method for utilizing error signals to increase accuracy in a brain-machine interface system of claim 6 , wherein reducing the dimensionality of the second neural signal comprises using a principal component analysis; and classifying the second neural signal comprises using a linear support vector machine.
10. The method for utilizing error signals to increase accuracy in a brain-machine interface system of claim 6 , wherein the cancelled action is to select a key in the virtual keyboard.
11. The method for utilizing error signals to increase accuracy in a brain-machine interface system of claim 6 , further comprising updating the intention decoder based on the classifications made by the error signal decoder.