真乄科技業的頂尖投資團隊

Illustration: University of Melbourne

Brain-machine interfaces have in recent years allowed paralyzed patients to control robotic arms, computer cursors, and exoskeletons simply by thinking about it. These interfaces require electrodes that are surgically implanted into or on top of the brain to read electrical signals from firing neurons.

But a novel stent-like electrode can record brain signals without the need for risky open-brain surgery. The matchstick-size “stentrode” made by Australian scientists can instead be inserted into a vein that runs beside the brain. From that spot, it can record high-quality electrical signals.

Doctors would implant the device by snaking a catheter up into the skull via a vein in the neck. The device picks up electrical signals and sends them through wires that go through the neck to a transmitter implanted on chest muscle under the skin. The wireless transmitter’s signals are read through the skin, then decoded using sophisticated software, and used to control an exoskeleton.

The research team used the stentrode to record high-frequency neural signals from a freely moving sheep for over six months. The spectral content and bandwidth of the signals from the stentrode matched those from electrode arrays that the researchers surgically implanted on the sheep’s brain. The results are published in the journal Nature Biotechnology.

Made from nitinol—an alloy of nickel and titanium—the device is a 3-mm-wide, 3-cm-long tube with a net-like surface studded with tiny disk-shaped electrodes. It is squeezed into the catheter and springs into its original form when the catheter is removed. Each electrode reads electrical activities of about 10,000 neurons.

In the first few days after implantation, the stentrode gave intermittent signals, says Thomas Oxley, a neurologist at the University of Melbourne (and currently a neurology fellow at Mount Sinai Hospital in New York City), who is the brain behind the new technology. There was a lot of interference because of noise created by blood flowing through the vein where the electrode comes to rest. But after about six days, the signals started to come in louder and clearer. X-ray imaging showed that the device was being absorbed into the vein wall. This effectively shielded the electrode from the noise, Oxley says. It also proved the device’s biocompatibility.

The stentrode was able to record signals with frequencies up to 190 hertz. “This high-fidelity activity [between 70 and 200 Hz] is the best information from the motor cortex and is the most useful for a brain-machine interface,” Oxley says.

Some brain-machine interfaces have used electrodes attached to the scalp to decipher weak electroencephalographic signals. But the scalp attenuates the brain’s electrical signals, and EEG electrodes can only record signals with frequencies up to 60 Hz.

This is why many BMIs use implanted electrodes. But those can cause chronic inflammation and tissue rejection that leads to device failure over time, Oxley says. Apart from the huge benefit of not requiring surgery, the vein-implanted stentrode has another exciting advantage, he adds. The deep folds in the brain are where are where a large portion of the brain’s neurons lie. “These are the most information-rich regions for motor decoding,” he says. “Even when surgeries are performed sometimes getting access to those areas is challenging because you can’t get into a deep fold. But veins naturally line those folds.”

The researchers haven’t done any signal decoding yet. That would have required signals from a more sophisticated animal such as a monkey, Oxley says. But monkeys’ veins were too small for the stent. So the scientists plan to do decoding and motor control directly from human brain signals.

Human trials are set to begin in late 2017, when a select group of paraplegic and quadriplegic patients from the Royal Melbourne and Austin Hospitals in Australia will be implanted with the stentrode.

Source:IEEE