Surface Implant Uses Machine Learning to Capture Deep Brain Images | Research & Technology | Jan 2024

SAN DIEGO, Jan. 23, 2024 — A neural implant developed on the University of California San Diego may assist advance the trail to minimally invasive brain-computer interface (BCI) know-how. The implant supplies high-resolution knowledge about deep neural exercise by recording on the mind’s floor.
Built by a staff led by professor Duygu Kuzum, the implant consists of a skinny, clear, versatile polymer strip that conforms to the mind’s floor. The strip is embedded with high-density arrays of graphene microelectrodes that allow up to 256 channels.
The graphene microelectrodes have ultrasmall openings and huge, clear recording areas. The diameter of the microelectrodes is scaled down to 20 μm. Each electrode within the implant is linked to a circuit board by a micrometer-thin graphene wire.

When positioned on the floor of the mind, this skinny, versatile implant permits researchers to seize high-resolution details about neural exercise deep contained in the mind with out damaging its delicate tissue. Courtesy of David Baillot/UC San Diego Jacobs School of Engineering.
Researcher Mehrdad Ramezani stated that fabricating a single layer of graphene as a skinny, lengthy wire introduced a problem, as a result of any defect within the wire would make it nonfunctional. To stop open-circuit failures, the researchers used an interlayer-doped, double-layer graphene. Instead of fabricating the wires as a single layer of graphene, they constructed the graphene wires as a double layer doped with nitric acid within the center.
“By having two layers of graphene on prime of each other, there’s a superb likelihood that defects in a single layer can be masked by the opposite layer, guaranteeing the creation of totally useful, skinny, and lengthy graphene wires with improved conductivity,” Ramezani stated.
The use of graphene wires as an alternative of conventional steel wires to join the electrodes to the circuit board ensures transparency and supplies a transparent subject of view for a microscope throughout imaging experiments.
To obtain excessive density arrays, the researchers used a microfabrication method that concerned depositing platinum nanoparticles onto the graphene electrodes. This strategy improved electron circulation with out affecting the transparency or extraordinarily small measurement of the electrodes. To one of the best of the researchers’ data, they’ve developed essentially the most densely packed, clear electrode array for any surface-sitting neural implant to date. The high-density arrays allow neural exercise to be recorded with excessive spatial decision throughout giant areas.
“This new technology of clear graphene electrodes embedded at excessive density permits us to pattern neural exercise with increased spatial decision,” Kuzum stated. “As a consequence, the standard of indicators improves considerably.”
The researchers additionally built-in machine studying strategies into the know-how. This made it attainable to predict deep neural exercise from the implant’s floor indicators.

In exams on transgenic mice, the implant enabled the researchers to seize high-resolution details about two sorts of neural exercise concurrently. When the researchers positioned the implant on the floor of the mind, it recorded electrical indicators from neurons within the outer layers. When on the identical time the researchers used a two-photon microscope to shine laser mild by means of the implant, it imaged calcium spikes from neurons situated as deep as 250 μm beneath the mind’s floor.
The staff found a correlation between floor electrical indicators and calcium spikes in deeper layers. Based on this discovering, the researchers used floor electrical indicators to practice neural networks to predict calcium exercise for particular person neurons, in addition to giant neuron populations, at varied depths.
“The neural community mannequin is educated to study the connection between the floor electrical recordings and the calcium ion exercise of the neurons at depth,” Kuzum stated. “Once it learns that relationship, we are able to use the mannequin to predict the depth exercise from the floor.”

Closeup of the clear graphene electrode array developed by a staff led by Duygo Kuzum on the University of California San Diego. Courtesy of David Baillot/UC San Diego Jacobs School of Engineering.
Typically, a topic’s head is mounted below a microscope when imaging calcium spikes, and the experiment can solely final for an hour or two. The skill to predict calcium exercise from electrical indicators will get rid of these restrictions.
“Since electrical recordings would not have these limitations, our know-how makes it attainable to conduct longer period experiments wherein the topic is free to transfer round and carry out advanced behavioral duties,” Ramezani stated. “This can present a extra complete understanding of neural exercise in dynamic, real-world eventualities.”
Next, the staff plans to give attention to testing the know-how in several animal fashions with the long run objective of translating the know-how to people. To advance neuroscience analysis, the staff is sharing the know-how with labs throughout the U.S. and Europe, the place it’s being utilized in various research. To make the know-how extra broadly out there, the staff has utilized for a National Institutes of Health (NIH) grant to fund efforts in scaling up manufacturing and facilitating its adoption by researchers worldwide.
“This know-how can be utilized for therefore many various elementary neuroscience investigations, and we’re keen to do our half to speed up progress in higher understanding the human mind,” Kuzum stated.
“We are increasing the spatial attain of neural recordings with this know-how,” she stated. “Even although our implant resides on the mind’s floor, its design goes past the bounds of bodily sensing in that it may infer neural exercise from deeper layers.”
The analysis was revealed in Nature Nanotechnology (https://www.nature.com/articles/s41565-023-01576-z).

https://www.photonics.com/Articles/Surface_Implant_Uses_Machine_Learning_to_Capture/p5/a69653

Recommended For You