The product Steve Archer started work on 14 years ago is just about to hit the market -- he hopes.
The NeuroPace RNS is the first implant to listen to brain waves and autonomously decide when to apply a therapy to prevent an epileptic seizure. It was developed by a company with a staff of less than 90 people, only about 30 of whom work on the core electronic, mechanical, and software engineering teams.
Maintaining a lean staff was a conscious decision of managers who have raised $215 million to date and knew the project would take a long time to pass regulatory approvals. "The joke is we have been two years away from an FDA clearance for about a decade," said Archer.
The FDA now says the company is very close to getting a green light, though it is not giving any specific dates. Indeed, an FDA representative was calling the company despite the recent government shutdown to work out details.
"There are a lot of really sexy engineering fields out there that let you make a lot of money, but the medical sector is very regulated, so what takes a few weeks to design in a hard drive takes years in medical," Archer said.
On the other hand, "there's a certain feeling when something you personally designed is walking around inside a great number of people -- you feel like it sure better work, but it's more than that," Archer said, recalling some of the 256 epilepsy patients who have been able to lead more normal lives due to receiving the implant in clinical trials.
The work holds promise of advancing brain science, too. For instance, NeuroPace already has discovered patients can have hundreds to thousands of pre-seizure events every day.
"No one else is recording ambulatory brain wave activities from epileptic patients to the extent we are," Archer said. "We have the world's largest library of such recordings and that's a tremendous resource for all sorts of research -- the opportunity is huge."
Indeed, some say today's deep brain implants are in a stage of development similar to cardiac devices in the early 1980s. They may spawn families of systems addressing a wide variety of neurological disorders.
NeuroPace is generally seen as the furthest along of a handful of neural implant efforts around the world. "The field is very much in the beginning stages of sensing and algorithm development, and the devices are all in investigation stages doing pilot clinical trials," said Tim Denison, a technical fellow at Medtronic.
Denison has worked for several years developing the chips for Medtronic's Activa PC+S device, which was implanted in its first trial patient this year. About 20 teams around the world will eventually implant a couple hundred of the devices over the next two to four years as part of its trials.
NeuroPace is unique in being a closed-loop system, automatically applying therapy based on sensed brain wave patterns, Denison said. Medtronic's device is unique in continuing to sense brain wave activity even while it sends out therapeutic signals thanks to a custom amplifer design and other techniques.
"We're taking advantage of radio principles to tune into brain waves just like you would a radio station, rejecting bands where stimulation sits," Denison said. "We maintain microvolt resolution of neural activity even in presence of stimulation signals on the order of a volt."
Not everyone is convinced deep brain stimulation will become a huge new field.
"I have a jaundiced view, primarily because of the brain damage that is caused not only by the placement of the stimulation and sensing electrodes but also from the milliampere levels of stimulation," said Mir Imran, a serial entrepreneur and investor in implants who developed one of the first defibrillators. "Depending on the location of the electrode leads/stimulation, the brain damage can result in cognitive impairment, memory loss, and other deficits."
However, "a few patients with incessant seizures who are refractory to drug therapy might get some benefit from this therapy, and it may be a better option than brain resection surgery for these patients."
The medical advances that determined the NeuroPace device could be implanted in the skull -- rather than in the chest with wires snaking up to the brain -- were the most impressive parts of the design, Archer said. However, the electronics design is significant, too.
The RNS captures brain waves on eight electrodes amplified and digitized in four channels at 250 or 500 Hz. Several algorithms monitor activity, waking up a custom processor when they detect pre-seizure patterns to trigger therapy. The stimulation is very configurable, but a typical burst may be 100 ms in duration with a pulse rate of 50 Hz, with each pulse having a bi-phasic amplitude of 3 or 4 mA and a width of 160 microseconds per phase.
NeuroPace takes low power operation to rare levels, because the device must measure brain waves continuously for the life of its custom lithium battery. "When fully running, it consumes 10 microamps, which is the leakage spec of many other devices," Archer said.
Heavy use of clock gating, state machines, and custom signal processors help keep consumption low. The system's processor duty cycle is just under 1 percent.
Most of the electronics are integrated in two mixed-signal ASICs surrounded by a few medical-grade off-the-shelf discretes. The 7x10 mm chips are made in ancient third- or half-micron processes for the lowest leakage current. One chip mainly handles signal processing and conversion. The other is primarily digital and uses an obscure, licensed eight-bit RISC processor core with SRAM for storing brain-wave recordings.
Archer estimates he has worked on about five completely different ASICs in his 14 years at NeuroPace and several generations of each one.
Looking ahead, "we are evaluating the business cases and needs in a number of indications," he said. "One could imagine an implant could download different code and be a different device with a flexible therapy output that could be applied for any number of indications."
In anticipation of a thumbs up from the FDA, NeuroPace is building up manufacturing and other capabilities, he said.
This blog originally appeared on EETimes.