Related Experiment Video
Updated: Apr 12, 2026
11:34
Breathing-controlled Electrical Stimulation BreEStim for Management of Neuropathic Pain and Spasticity
Published on: January 10, 2013
23.8K
An Active Learning Algorithm for Control of Epidural Electrostimulation.
Thomas A Desautels1, Jaehoon Choe2, Parag Gad2
1California Institute of Technology, Pasadena, CA 91125, USA. He is now with the Gatsby Computational Neuroscience Unit, University College London, London WC1N 3AR UK.
IEEE Transactions on Bio-Medical Engineering
|May 15, 2015
Summary
An automated algorithm, Gaussian process upper confidence bound (GP-BUCB), efficiently selects epidural electrostimulation parameters for spinal cord injury therapy. This method matches or exceeds human expert performance in eliciting muscle responses, facilitating clinical application.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Rehabilitation Technology
Background:
- Epidural electrostimulation shows therapeutic potential for spinal cord injury (SCI).
- Manual selection of effective stimulation parameters is time-consuming and inefficient for clinical use.
- An autonomous system is needed to optimize stimulus selection for SCI therapy.
Purpose of the Study:
- To develop and validate an autonomous algorithmic system for selecting epidural electrostimulation parameters.
- To compare the performance of the proposed algorithm against human expert selection.
- To assess the algorithm's ability to discover effective stimulation patterns and adapt to variability.
Main Methods:
- Implementation of a Gaussian process bandit algorithm (GP-BUCB) for stimulus selection.
- Testing the algorithm in spinally transected rats with epidural electrode arrays.
- Comparison of algorithm-selected stimuli with human expert-selected stimuli for eliciting muscle responses.
Main Results:
- GP-BUCB successfully controlled spinal electrostimulation in 37 sessions, selecting 670 stimuli.
- The algorithm's performance was comparable or superior to human expert selection.
- GP-BUCB autonomously discovered effective stimulation patterns and demonstrated adaptability to temporal variability.
Conclusions:
- Automated stimulus selection using GP-BUCB is a viable and effective approach for epidural electrostimulation in SCI therapy.
- This method can significantly streamline the therapeutic process and facilitate clinical application.
- The algorithm's ability to learn and adapt suggests potential for personalized SCI treatment.
