A team of researchers from Tokai University in Japan recently produced the first 3D model of a fruit fly's neuronal network. The team's work is an important step forward for neuroscience insofar as it has overcome the limitations of previous brain imaging techniques by providing a much higher resolution image of the brain's neural network, capable of depicting the shape and position of roughly 100,000 neurons.
As detailed in a paper published earlier this month on arxiv, the researchers were able to accomplish this by repurposing an imaging technique usually used for creating 3D models of complex molecules.
Known as x-ray crystallography, this technique works by first crystallizing a purified sample of the molecule being studied. It is then bombarded with x-rays and the x-ray diffraction patterns that result are measured. The problem is that the resulting diffraction patterns are actually only showing changes in electron density inside the molecule, not the positions of the atoms themselves—this must be inferred from the electron diffraction measurements.
Yet as the Tokai team discovered, leveraging this technique to image a neural network is slightly trickier. This is because unlike atoms in a molecule, which could be conceived as points in space, neurons in a brain are more like lines that twist and turn, making their location and structure harder to infer based on diffraction data.
So the team slightly tweaked this imaging approach by using a technique called x-ray tomography. Instead of creating a crystallized version of a fruit fly's brain, the team soaked the brain in silver dye. Afterwards they bombarded this silver brain with x-rays and measured how the silver in the neurons absorbed the radiation.The data that is produced by this technique is fed into a computer imaging program which uses the data to create a 3D model of the shape and positions of neurons within the brain.
The end result of repurposing this bio-imaging technique to map a fruit fly's neural network was impressive. This first 3D image showing neural position and connections has a resolution around 600 nanometers and depicts around 100,000 neurons. While most of these neurons were mapped by computer modeling, which was able to check the consistency of neural connections, when an anomaly was found it was up to the researchers to correct the error by hand.
By the time everything was said and done, the researchers had spent about 1700 person hours creating this neural map—nearly 71 full days. This is a huge investment to map out a neural network comprised of some 100,000 neurons, which is just a small fraction of the number of neurons in a human brain: around 100 billion. Clearly, this technique wouldn't be able to scale to humans for this reason, but nevertheless is an important step in figuring out a way to map increasingly complex brains.