Despite the isolated success of deep brain stimulation for Parkinson’s disease, and a few other impressive parlor tricks with BCIs, running massive networks of wire electrodes through the brain is probably not going to be a long-term solution to any brain problem. Recognizing this truth, many researchers have turned to nanoparticle-based approaches for remote, non-invasive stimulation of the brain. A single 10 ug bolus of injectable smartdust delivering hundreds of 30nm motes to each cell can in theory provide the kind of mass neural coverage needed for complete brain control.
Fantastic as that might sound, reliable stimulation of neurons can now be routinely achieved by optically or magnetically activating special-purpose nanoparticles. For optical activation, typically some kind of structured gold is used while for deep brain magnetothermal stimulation, iron oxide is the material of choice.
There are several major problems with these technologies as they are now envisioned. For one, they generally rely on poorly characterized heating effects for the stimulation. Although heat-responsive protein channels have been added to neurons in an attempt to reduce the amount of heat required, there are simply too many unknowns still involved in the process. The other way to remotely activate neurons is to use high-intensity magnetic fields (as in TMS) to create local stimulation currents.
The problem with TMS is not just that there is no specificity, but that you need to blast away with a magnetic field (H field) of 10,000 Oersteds (Oe) just to get an effect. A way to use the energy intensification capability of nanoparticles for reducing this huge magnetic field requirement, without resorting to heating effects for stimulation, has recently been developed by Sakhrat Khizroev and his colleges at Florida International University. Their trick is to use a new kind of nanoparticle that exploits a phenomenon known as the magnetoelectric effect (ME).
Sakhrat’s group has previously shown that ME nanoparticles (MENs) could be used as nonthermal nanoscale actuators for controlled release of anti-HIV drugs within the brain. They have now shown that these same particles can also be used to stimulate mice brains using field of just 100 Oe. The MENs were made using a core of CoFe2O4 ferromagnetic spinel, which enhances the magnetic moment. A shell made from BaTiO3 peroskivite nanostructure induces the ME effects. By substituting in other transition metals from within the same series, the ME coefficient can be tuned over a large range. The nanoparticles are then made biocompatible by coating them with glycerol mono-oleate.
In order to pull the MENs into the brain from the bloodstream, the researchers applied a 3000 Oe/cm magnetic field gradient with a permanent neodymium magnet for 30 minutes. That was enough to get at least 10% of the MENs (around 20 billion particles) into the brain. They then applied an alternating magnetic field using a one inch coil energized at frequencies from 1-20 Hz. This resulted in significant modulation of the EEG signals obtained from two electrodes implanted in the mice. The researchers estimated that local ME coupling in neurons which had taken up or bonded the nanoparticles would generate electric fields up to 1000 volts/m.
Without recording single unit spike activity of neurons directly, it is probably not possible to unequivocally demonstrate direct stimulation by the presumed excitatory neural mechanisms. The researchers were able to image sections of the brain using atomic force and electron microscopes to prove the particles were at least in the right places. They were also able to show uniformity of the particle sizes and measure both the magnetization saturation and magnetic coercivity with a vibrating sample magnetometer.
Since the MENs are not yet commercially available, anyone wanting to experiment will either have to ask the researchers for some of them, or else have access to a fairly sophisticated materials characterization lab. The external coil probably wouldn’t be too difficult to make, however, to get real spatial selectivity — something more precise will probably be needed. Although a modified MRI machine could probably do the trick, it wouldn’t be something you could wear around. The researchers suggest that it eventually may be possible to use multiple near-field coherent antennas operated in a holographic mode to get the required selectivity.
For a full blown BCI, the other piece of the puzzle is the ability to record from the neurons. While the MRI, again, could in theory create a real-time electric field map of the brain by using the MENs as contrast agents, something better would be needed. We won’t delve too deep here, but the ideal solution to the ‘readout’ problem may already be in hand. Ed Boyden and his colleagues just posted a proposal on the Arxiv server to use optical reflectometry to read multiple neurons from a single fiber. Tiny capacitive effects generated by the local firings of neurons would create “intensified” electric fields along the fiber.
This intensification is analogous, but reciprocal, to the the MEN approach we have described. The reflectometer would work by sending light pulses down the fiber and recording the time and magnitude of the reflections. The voltage at points along the fiber would modulate the local index of refraction and generate a signal that could be decoded to give the activity of the neurons at each position. The beauty of the fiber approach is that they could be threaded through the vasculature to reach any neuron.
While combining the best ‘read out’ techniques with the best ‘write’ techniques is still a task for the future, we are already getting the flavor of what the ultimate brain interface might look like. Please note that Boyden’s work, although submitted for publication to the Journal of Biomedical Optics, has yet to condoned for general consumption by the proper peer review channels. Sakhrat’s research, on the other hand, has already been appropriately vaulted away the under the protective shield of a $60.00 pay-per-view rental fee that self-obliterates after 24 hours.
