Systems and methods for efficient prediction and design of neural stimulation

Unmet Need

Bioelectronic therapy is the ability to deliver various electrical impulses to patients through implanted medical devices. These devices can treat heart failure, epilepsy, and chronic pain among other ailments. Depending on the disease, the implanted device will stimulate one or more target nerves throughout the body. However, this approach is limited in efficacy because the stimulation signals can either fail to stimulate sufficiently the target nerve or can cause off-target stimulatory effects on nerves within the targeted area. Mitigating these factors is made difficult by the substantial computing power necessary to calculate the ideal stimulation signal for a specific set of nerves given a condition to treat. There is a need for a machine learning model that can predict and implement a targeted neurostimulation treatment.

Technology

Duke inventors developed a machine learning model that very efficiently predicts and designs neurostimulation treatments based on training data. This is intended to be used by physicians and researchers to design bioelectronic therapy treatments to be administered via implanted stimulation devices. Specifically, this algorithm uses data on which nerves require stimulation by the user and calculates the ideal duration, waveform, and intensity of the treatment. This significantly reduces the required computational time to design the treatment and results in more precise and effective treatment. This has been demonstrated in a state-of-the-art neuronal model used to assay neural stimulation. This software has been tested compared with NEURON, another such modeling software.

Other Applications

This technology could also be used to determine the best courses of bioelectric therapy and can also be expanded to be used in other species to study neuronal signaling and development.

Advantages

• Dramatic reductions in compute time compared to other state of the art prediction models
• Inverted model allows focus on human knowledge of exact nerves to stimulate rather than on the electrical signal impulses to use, which is much easier to determine computationally
• Streamlined and efficient algorithm design allows for up to a 10,000x increase in throughput compared to existing models, increasing the number of predictions that can be made