March 25, 2015
The brain’s connectivity matrix, or netlist if you will, is something that neuroscientists are still scratching their heads about. Some of them have already put their money where their mouth is and embarked on massive ‘connectome’ projects. The goal of these is to reconstruct a 3D model of the brain from impossibly thin slices, either pealed off or etched away with gallium ion beams, and then scanned with fancy multi-beam electron microscopes. A better way, if it can be done, is keep the brain intact and somehow scan it whole.
Researchers at the University of Florida have recently found a way to extend the resolution of noninvasive MRI down to the 10um range. This flies in the face of the unwritten Moore’s law of MRI, which throws up hard physical limits to what the technology is capable of. As the researchers have found, the first thing you have to do is change the name of your technique. Anything above a 100um^3 voxel resolution can still be called MRI. However, below that level, what you want to do is use the newer term MRM, which stands magnetic resonance microscopy.
As the researchers report in the journal Scientific Reports, the key is not just to use a super-strong magnet, but to also have equally strong and fast-switching gradients (the coils that provide for a graded field to superimpose location information) and the most sensitive RF detector coils possible.
RF coils can take any number of forms, but the design used here is a simple flat surface coil. Generally speaking, the depth of the image possible with surface coil maxes out at about one radius. The good news is that the sensitivity of the coil is inversely proportional to its size. Therefore, to image small brains, or parts of brain, small coils are better. The trick for would-be connectomists will be to get these coils right up next to where you want them, presumably either in the ventricles or the vasculature of the brain.
To demonstrate proof of principle, the researchers imaged a small fly brain scarcely half a micron wide. It may be tiny, but with over 10^5 neurons and 10^7 synapses this brain is certainly no slouch. Rather then sticking the head inside a complex solenoid or birdcage RF coil, as we would need to do in the hospital, the researchers were able to simply stick the fly brain right on top of the flat coil, essentially turning the scan geometry inside-out.
By combining several different ways to energize and read out the various coils, the researchers were able to create images on the scale of individual neurites. Instead of adding special contrast agents, or recording the so-called BOLD (Blood Oxygenation Level Dependent) signals sometimes used for functional mapping, the endogenous water in and around the cells was all that was needed for contrast. Normally known as diffusion MRI, this technique has been used to great effect in a more clinical setting for imaging the long axon tracts in the white matter of the brain.
The problem with diffusion MRI as it is practiced now is that the only thing you get are the big wires, bundles of wire really. You don’t get any actual connections, and you don’t get the directions. That is to say, you can’t unambiguously determine which direction the axon sends signals or to whom it connects. The beauty of a synapse-scale imaging technology is that since water is constrained in the small roundish synapses in much the same way that it is constrained in the longish axons, you can in theory fill in the fine-scale network topology from diffusion alone.
Imaging a live brain means that rather than just a one-shot picture of a static brain, you can potentially revisit it and record actual change. That’s all great for neuroscience, but what really sets the technique apart is that all the brain-uploaders, mind-cloners, and other would-be immortals no longer have to accept the illusory consolation prize of a vicarious immortality-by-proxy. In other words, you don’t have to turn your brain to mincemeat first in order to render a decent facsimile of its structure in a computer.
An important immediate benefit of this new tech is that it will better define the cellular basis of the actual MR signal itself. Right now the researchers are scanning brains for 40 hours just to get the first indications of subcellular detail. They liken the situation to optical imaging in poor illumination, where one needs a very long exposure time to generate a visual signal. With improvements to the hardware, they hope to bring this time down to the point where it becomes more practical to image living, breathing brains in the lab.