5. Through magnifying glass: twenty micron resolution explained.
“The resolution in the BigBrain model is 20 micrometres. At HIBALL, you now want to create a brain model with an accuracy of one thousandth of a millimetre. Why?
At 20 micrometres, you might see most of the cells. However, there are also cells that are only ten micrometres in size, and these have previously remained “blurred”. Therefore, within each tissue section, we want to go to the level of one micrometre. The goal will be a spatial resolution at 1 x 1 x 1 micrometre. Only then can the cells with their different forms and extensions be seen and it can be understood how they are arranged in the brain. On this basis, brain functions and cognitive performance can be associated.”
As incredible as it sounds, it seems to me this answer is not just a simple error. To illustrate this issue, I post below two carefully selected fragments of the same section from BigBrain data set. To select the same region as precisely as possible I used the landmark – a vessel at the top of the fragment, and section coordinates. Left fragments is from image file at 20 micrometers resolution, middle fragment is from image file at 1 micrometer resolution, and left image is the same as right, with outlines. But you should look very carefully!
Fig. 5: 20 micrometer resolution compared to 1 micrometer resolution.
My Question 1: Do you see any cells in the left image?
My answer: Actually, this question, or rather – Prof. Amunts’ answer, triggered my decision to write this article and to get to the bottom of this issue. I could not tolerate this anymore: I am not blind yet (thank God!). Of course no cells are visible in 20-micron resolution image. What you can see are two blurry shadows. The shadow on top corresponds to the vessel in the middle image (outlined in red on the right), and blurry shadow at the bottom is a cluster of neurons, composed from three large pyramidal neurons surrounding 8 to 10 smaller neurons, outlined in blue in the right image.
By MRI standard, 20 microns would be a high resolution. But by microscopic standards, this is very low resolution image. For example, even using plan achromat objective lens 1x with numerical aperture (N.A.) 0.04 and a mid-spectrum wavelength of 550 we would have resolution of 6.875 micrometers. With objective 4x/N.A. 0.1 we would have resolution 2.75 [see explanation here].
Moreover, the resolution of human eye without any optical device is approximately 30 microns (read explanation here). It should be clear: resolution of 20 microns is not a microscopic resolution by any standard. Practically, the image at 20 microns resolution can be created by a magnifying glass with factor between 2x and 3x. Now it becomes quite obvious, that such declarations as “We created an ultrahigh-resolution three-dimensional (3D) model of a human brain at nearly cellular resolution of 20 micrometers” [1, Summary] or “The three-dimensional model of the brain “BigBrain” makes it possible to understand the complex structure of the human brain at a microscopic level in all three directions (C. and O. Vogt Institute of Brain Research)”are purely speculative.
My Question 2: What is the reason for such a long delay with 1 micrometer-resolution data?
My answer: I do not know. Hopefully, now it should be clear enough: if no cells could be found in an image – then 20μm model cannot show any cells, and cannot be honestly called “cellular ” or even “nearly cellular”. Papers, published from the same group, clearly show that “light microscopic images on a scale of 1μm allow detecting details of single cell morphology”  . However, surprisingly, as reported in 2016, the second BigBrain data were again collected with the same 20 micron resolution, and again the flat-bed scanner was used. So, even though the reported workflow was new, the raw data resolution remained the same as in the old “BigBrain number 1”. However, it becomes a complete mystery, why despite projected files size for total raw data at 1 micron resolution was limited to a “modest” 180TB, they estimate that space needed for “1 mk isotropic will require several PB”. Where did this value come from is another mystery of BigBrain project.
I would not be surprised, if authors of the project have in mind very inefficient data format, and never considered recent development in the area of big data storage, like OpenVDB. This data format, created for efficient manipulation with sparse volumetric data, can save at least 70% of raw isotropic data volume, so petabytes might easily be reduced to the same ~200 TB space, as in “raw data” situation.
The availability of 3D data at one micron resolution or better might become a unique source for the analysis of functional significance of different areas. But it is not available despite 17+ years anniversary of this project. My guess is: today, eight years after BigBrain #1, and 5 years after BigBrain #2, high-resolution data might be available in abundance, but they are not shared yet. Why? As I tried to show, even if large amount of it and “petabyte’ish” size of “isotropic” data files looks like such insurmountable barrier, sharing raw data might already solve many problems with “single-neuron level of resolution”. Still, they continue to claim, to the contrary of all evidences, that single neuron resolution is available at 20 microns, but do not share 1 micron resolution data.