To induce a more realistic response, it may be possible to deliver biomimetic spatiotemporal patterns using multi-channel microstimulation. With this aim, researchers from the Polytechnic Institute of NYU, SUNY Downstate and the University of Florida have developed a method for eliciting naturalistic responses in the somatosensory cortex using optimized microstimulation patterns (J. Neural Eng. 13 056007).

"Algorithms for encoding touch stimuli artificially through electrode arrays have not yet been designed explicitly to evoke natural responses," explained first author John Choi. "This is what we wanted to accomplish in our study."

Mimicking natural responses

Choi and colleagues developed their scheme using two microelectrode arrays implanted in the brains of nine anesthetized rats: one array in the forelimb representation of the VPL thalamus (which deals with sensory information) to deliver intra-thalamic microstimulation (ITMS); and the other in the corresponding projection area in the primary somatosensory cortex, to measure neural activity during stimuli.

First, the researchers recorded downstream responses to physically touching a rat's forepaw with varying pressure, duration and location. These responses – multi-electrode recordings of local field potentials (LFP) – served as templates. Next, they delivered single-pulse probing ITMS, and used the neural responses to train a linear state-space model of the cortical LFP response to VPL microstimulation.

The team then optimized a set of input ITMS pulse patterns to reproduce as closely as possible the naturally occurring responses to each touch type. Finally, they applied these optimized patterns to the VPL, and assessed the animals' responses to natural touch and microstimulation.

"The responses that our procedure generated for different touch locations and temporal patterns were very similar to the natural responses that they were optimized to mimic," said Choi. "The responses also contained the same amount of information about touch parameters [pressure and location], with the same latency of information transfer."

Across all conditions and rats, the correlation coefficient between natural and virtual responses was 0.78±0.05. For time periods within 100 ms of touch onset, this coefficient was 0.90±0.03.

Comparing spatial responses from two sites on a rat's paw revealed that touching different digits activated overlapping but clearly distinct zones in the recording electrode array. Stronger touches exhibited higher spatial reproduction accuracy than medium or light touches, and shorter-duration touch patterns had higher accuracy than longer patterns. Over all patterns in all animals, the spatial reproduction accuracy between virtual and natural touches was 0.72±0.22.

Two types of recording arrays were used in the animals: a 32-channel Utah array in three rats, and a 32-channel Michigan array in the other six. Results showed that spatial reproduction accuracy was 24.3% higher with the Utah array, attributed to the fact that the Utah array spreads its channels horizontally across the cortical surface, while contacts in the Michigan array span one horizontal and one laminar axis.

Clinical translation

The researchers concluded that this stimulus optimization approach holds great promise for restoring naturalistic levels of sensation; they are presently applying for funds to test it in humans. "One of the difficulties in working with animals is determining exactly how something feels to them, and therefore, we might not be able to answer the question of realism until we translate this work into humans," Choi told medicalphysicsweb.

One obstacle in translating this scheme to clinical application is that the method requires template responses to natural touch for the controller to mimic. "In our study, we utilized animals with intact somatosensory systems and just measured their natural touch responses," Choi explained. "These of course would not be available in the case of somatosensory prostheses." Instead, the researchers envision that a microstimulation encoder could be initially optimized using templates generalized from another hemisphere, subject or from a non-human primate study, with simpler fine-tuning used to individualize the encoder to the patient.

"Natural levels of proficiency in making dexterous movements require somatosensory feedback. Our general goal is to incorporate such feedback into brain machine interface systems for the restoration of movement," said Choi.

Related stories

• Implant restores hand motion after paralysis
• Optimizing brain stimulation therapy
• FES helps stroke patients move again
• Patient's thoughts control robotic arm