Jan 22 2019

New Technique to Map Brains

One of the “holy grails” of neuroscience is the ability to scan a brain and create a complete detailed map, including all networks and connections. Scientists use several techniques, all with their own drawbacks, and the process is very slow – it can take a year to completely scan a single fly brain. A collaboration of scientists, however, report in Science that they have developed a new technique that can accomplish a detailed scan of an entire fly brain (or a section of mouse cortex) in 2-3 days.

The team has been described as an “Avengers” type collaboration, and it is impressive. Specialists provided the prepared fly brains. Two different types of microscopy were combined (that’s really the new bit), along with a third imaging technique. Finally, computer specialists had to figure out how to combine all the data like puzzle pieces into an image. The result was a complete map of a fly brain in three days, which is an impressive leap forward.

The core innovation of the new technique is to use a combination of expansion microscopy and lattice light-sheet microscopy. Expansion microscopy is pretty much what it sounds like – the brain sample is expanded, retaining the relative positions of neurons and connections, but creating more space to facilitate imaging. Expansion is done chemically, similar to injecting an expanding gel into a specimen. The researchers expanded their samples four-times to provide optimal results. The potential problem with this technique is that it may introduce artifacts giving spurious results, so anyone using it has to be careful and validate their techniques (by reproducing known outcomes, for example).

This expansion technique was combined with the lattice light-sheet microscopy. This is a complicated setup that illuminates the specimen with high energy thin sheet of light, only that part of the specimen that is in focus to the microscope, keeping all the out-of-focus parts dark. Finally, this is all combined with fluorescence microscopy, which tags specific biological structures (such as certain amino acids) with fluorescent molecules. This way only certain cell types or certain connections or structures can be imaged and mapped. Specifically they used confocal microscopy, which provides better resolution and contrast.

The result of all these techniques combined is that they were able to produce high resolution images of all the neurons with their dendrites and connections of the entire Drosophila brain and a section of mouse brain. What comes next? The researchers will want to refine their techniques, and they plan to build microscopes specifically for this task. Computer technology is only going to get better as well. It is reasonable to predict that this new technique will benefit from some incremental advance the more it is used.

This brings us a pretty large step closer to the time when we can image an entire human brain with all of its connections. There are some obvious benefits to this for research – understanding not only the connectome but the function of all the various circuits and networks in the brain, how they interact, and how variations in these networks contribute to neurological and mental illness.

Theoretically, we could also duplicate such a detailed map in a virtual model – one that will not only record a map of the brain but possibly function as a virtual brain. It is quite possible that the first “self aware” AGI (artificial general intelligence) will be such a virtual model of a biological brain. This will also create incredible possibilities for research – if we want to know what the effect is of altering a network in the brain, we just change it and see the result. With virtual models we can have thousands of copies, and run millions of simulations (or some suitably large number, I am just winging the orders of magnitude here).

If the AGI is truly self-aware this brings up some thorny ethical issues, but let’s say we can get around them somehow. Perhaps we can study limited networks that are insufficient to be “awake” but can still model function. This will get complicated, but in a very interesting and educational way.

This also brings up an interesting thought-experiment about resolution. How much resolution and fidelity will it take to create not only a virtual human brain, but a virtual brain of a specific person – John Smith’s brain? What would that even mean? At what point of fidelity is the virtual brain John Smith? Perhaps at certain levels personality traits will be preserved, overall abilities, etc., but nothing else. What about John Smith’s memories? At what level of resolution are those contained?

Beyond reproducing a specific person, there is reproducing that specific person’s current mental state. This is definitely the stuff of science-fiction (at least for the foreseeable future). The transporter in Star Trek, for example, not only scans the patterns of a person, but their instant mental state. This requires such a high level of fidelity that they invented “Heisenberg Compensators” to deal with the limits of quantum mechanics. This is not limited to Star Trek, and is now a staple of science fiction.

For now neuroscientists would be thrilled to be able to create a virtual human brain, forgetting about reproducing an individual. Although understanding better how memory works would certainly be a priority of such research. This new technique doesn’t get us there, but it does get us a lot closer. At the very least, it should prove to be a fantastic tool of neuroscience research.

No responses yet