In the early 1900s, Korbinian Brodmann decided that the best way to tell apart the different parts of the brain would be by looking at cytoarchitecture or what types of cells and neurons are located there. By performing Nissl cell staining. Brodmann created maps of the cortex listing 52 different areas that have become a standard way of describing different brain regions. Although these areas have been updated a little bit over time, for the most part scientists have used these brain maps for over 100 years. So isn’t it time that we had an update?
An important part of neuroscience is trying to figure out what parts of the brain are responsible for our thoughts, behaviors, senses, and more basic functions. There is an idea out there that structure implies function, meaning you can guess what part of your body does based on how it looks. The brain is difficult to decipher because you can easily get lost in an endless sea of lumpy folds without ever having any idea what a particular part does!
Many areas of the brain have been identified by chance, looking at activity, or study of how many of a particular cell inhabit an area, but creating a map of the brain in this way means that you are only examining one way of drawing the map. In truth, examining the brain from multiple perspectives can show more information about what makes different regions different. Researcher’s from the Human Connectome Project used several different types of information mixed with a little bit of machine-learning to generate a multi-modal brain atlas. They looked at structural information about the thickness of the cortex and the concentration of myelin (an important part of white matter) as well as how the brain activated in 7 different tasks or how the brain activated when resting.
They built this multi-modal map using brainscans from 210 different people and compared it to another map using 210 different people. Instead of using how the brain folds to mark these different regions, researchers used “areal-features”, meaning that they flattened out the 3D maps to create the 2D maps of features used. They did this because even though the folds in the brain are very visible, they don’t always mark regions that are distinct, and they can vary from person to person. Some regions like the sensory and motor areas are known for having a strong relationship between how the brain folds and what the area does, but in different parts of the brain this is less clear and less reproducible. When they compared the averaged maps between both groups, they found that the two flattened maps were very similar for all the different modalities they measured with (all with correlations r>0.944). Having two similar sets of maps is evidence that the map wasn’t accidentally generated because of unusual subjects or outliers. This is more evidence that it should be able to represent the average person’s brain reliably.
In order to come up with the 180 brain regions that are part of the new map, the lines had to meet several requirements. First, the border need to mark a difference between two areas in at least two different feature maps (spatially overlapping gradient ‘ridges’). Second, there had to be some similar trend on the other side of the brain in about the same place (opposite hemisphere). Third, this border couldn’t be caused by artifacts or errors in the measurements. And Fourth, the differences had to be statistically significant. They also considered whether or not the region had been studied before, but this didn’t stop them from marking new regions that had not been studied. Of the 180 regions in the new brain map, over 100 regions are completely new and have never been studied on their own before!
It’s often times helpful to be able to understand where an area is in the brain when you are studying it, if only to help make such a huge and complicated organ simpler. When studies use whole images of the brain they sometimes have thousands of voxels to examine, and when you analyze so many different bits of information, sometimes you get false positives, when something looks really important, but is actually just random. Reducing the amount of data (dimension reduction) you look at in a smart way actually makes for better science. For example, in the original data for this brain map, if you only look myelin thickness, there are over 30,000 different points per hemisphere (only at 1 of the features!). Using the new ‘parcellated’ brain regions, scientists can focus on only 180 places, making the data much easier to manage.
The last thing the authors did was create an automatic classifier to apply on brainscans from new people. Scientists trying to do their own research can automatically identify which parts of their subject’s brains belong to which of these areas with 96.6% accuracy. This makes it easier for researchers to use the investigate how these new areas are activated in cognitive tasks and how they are affected in different diseases. It’s been over 100 years since the first modern maps of the brain were created and making new maps may completely change the way we view the brain. But this is only version 1.0 and new versions will continue to help us learn and understand more.
Here is the new brain atlas, color coded according to five specialized areas, each responsible for different functions. Red indicated areas associated with audio and hearing, green indicates areas associated with senses and muscle movements, Blue shows areas related to vision. Together these are fundamental three types of input into our brain. Areas linked to different tasks were marked in white and areas that are normally part of resting state but turn off during tasks are shown as black. Feel free to explore!
Glasser, M. F., Coalson, T. S., Robinson, E. C., Hacker, C. D., Harwell, J., Yacoub, E., … Van Essen, D. C. (2016). A multi-modal parcellation of human cerebral cortex. Nature, 61(16), 5985–91. http://doi.org/10.1038/nature18933