A New Paradigm in Neural Recording
January 7, 2020
Summary: Paradromics, together with partner Caeleste, have developed a novel mixed-signal sensor technology that enables high data rate neural recordings with 60 times lower power consumption than conventional digital readouts. The compact, low-power sensors interface with recording electrodes, reduce in-body thermal dissipation, and can be tiled to support active microelectrode arrays at densities up to 10,000 channels per sq cm. This development represents an important milestone in the development of high data rate brain computer interfaces (BCIs), allowing implants to transmit more neural data without overheating the brain. This will enable a broad range of novel and advanced applications in neural prosthetics.
“It’s important to keep brain implants low-power. That’s a tough constraint, because arguably the most important figure of merit for a brain-computer interface is its data rate, and more data costs more power. Our sensor technology allows for nearly two orders of magnitude increase in data throughput without dissipating more heat, so that’s a huge win for opening up new applications for neurotechnology.”Matt Angle, PhD, CEO and Founder, Paradromics
Paradromics is bringing to market a next-generation brain-computer interface that is fully implantable, portable, and durable, and has up to two orders of magnitude higher effective neural data rate than state-of-the-art technologies. This dramatic increase in the data rate will enable a new generation of neurotech applications and devices. High data rate brain computer interfaces have transformative potential for physical and mental health, restoring speech and mobility for paralyzed patients, recovering sensory capabilities, and creating bioelectric therapies for patients suffering from neuropsychiatric disorders.
Thus far, the development of high data-rate BCIs has been limited by the outsized power requirements and related heat dissipation that comes with scaling existing BCI electronics to large neural channel counts. Massively parallel recordings generate a large volume of raw data that needs to be digitized and processed to extract all relevant information. Digitization dissipates a lot of heat, and scaling existing digitization systems to large arrays of electrodes would dissipate so much heat as to cause irreparable damage to the surrounding brain tissue. This thermal damage is a major obstacle to scaling a standard data processing system to large neural channel counts.
Data Scientists and Electrical Engineers at Paradromics and Caeleste have resolved this seemingly insurmountable problem by designing an on-chip data processor. This new circuit can perform real-time extraction of key information from massively parallel neural recordings with 60 times lower power dissipation than conventional approaches.
Why it matters
“I am very proud of the work we have done at Paradromics to create the mixed-signal sensor. Its optimal design is an elegant consequence of the mathematical process we developed, which will keep our sensor at the forefront of analog compression for many years to come.”Rob Edgington, PhD, Head of Artificial Intelligence, Paradromics
The state-of-the-art technology used in therapeutic applications has enabled initial proof-of-concept demonstrations of control of prosthetics and computers by recording neural activity using up to ~100 parallel channels. This is extremely impressive, but higher data rates are required to enable the kind of seamless brain-computer communication that lends itself to complex applications and sophisticated prostheses. This performance limitation is one of the key reasons why BCI systems have remained strictly confined to the realm of research, with no commercial application reaching the market.
Paradromics’ pixel technology compresses the raw input signal from the brain without degrading the effective neural data rate output by digitizing and reading out only the key information contained within the input signal, rather than the entire raw signal. The lower digitization load results in ~60x lower power dissipation, and allows electrodes to be implanted into the brain at higher density than previously possible without causing thermal damage. Tiling large numbers of these miniature sensors across the brain will make it possible to record from an unprecedented number of neural channels.
By making large-scale neural recording possible, Paradromics’ technology represents a fundamental paradigm shift for the field of brain computer interfaces, enabling a new generation of BCI technology and neuroprosthetic applications to come of age.
Paradromics will be presenting this news, together with other exciting updates about the company’s recent progress, at the upcoming North American Neuromodulation Society (NANS) 2020 conference in Las Vegas, Jan 23-26, 2020.