Creating the most complex map in Human History 🧠

“The brain is the organ of destiny. It holds within its humming mechanism secrets that will determine the future of the human race” — Wilder Penfield

Images of functional connectivity between regions of interest in the brain based on the dataset I downloaded and processed

Intro to Connectomics

Imagine you are driving someplace new. A place you have not been in a while. Most people, including myself, would use Google Maps, or WaZe, to find their way to their destination. Consider, what would you do if you did not have a digital map available? You might ask someone else. Well… slight problem with that. What if no one else whom you had asked had ever been where you wanted to go? How would you find your way to your destination then? Probably with A LOT of frustration, trial and error, and rigorous testing. These three key elements are intrinsic to the study of the brain. An organ so complex and so mysterious, that even with all the technological advancements in human history; fire, cars, computers, self-driving cars, and space exploration being just a few, we still haven’t been able to map the brain out. However, 86 billion neurons and trillions of synapses to map is quite a feat. The field of science dedicated to completing the task of mapping the brain is called connectomics.

Image of a 500-year old map that helped guide Chris Columbus

“Sounds Unnecessary”.

That’s what I said when I first started hearing about connectomics. Why invest all this manpower, money, time, and energy into mapping something which we already understand well enough to cut into? After all, a surgeon can navigate the brain well enough to clip an aneurysm (dilated blood vessel at risk of hemorrhaging) or dissect a tumor from a patients brain. We can even sever the connection between the two hemispheres (a corpus callosotomy).

When I voiced my skepticism to someone who understands the field much better than I, they explained it like this:

Connectomics is about creating a detailed mapping system like Google Maps. Right now we have a ancient understanding. Imagine trying to find your way from the Empire State Building, NY to the Hollywood Hills, LA with a map that was used by Christopher Columbus. This is what we’re doing right now. We’re trying to create a new map; one which is more efficient, effective, realistic, and representative of the thousands of illnesses which affect the brain.

This new map of the human brain has not yet been created, and no one knows what it’ll look like when we get there. It will take time, effort, and determination from teams across the world to accomplish this, but in doing so it will allow us to understand the brain — one of the most fundamental things in our universe — much better. The difference in our ability to understand mental illnesses and treat patients will be night and day. Just as the difference between our travel today and that of ancient European explorers.

Thankfully, human innovation has found new frontiers over the past 50 years which make this study not just possible, but within our very grasp.

MRI and fMRI — Functional Connectivity

Image of an MRI machine from Radiology Affiliates Imaging

One of the most fundamental discoveries for connectomics was the invention of Magnetic Resonance Imaging (MRI). MRI machines are used across the world today to help diagnose patients with different injuries and illnesses, but they are also used to identify patterns of activity in the brain.

MRI machines work because our bodies are composed of 60% water. Water is composed of 2 hydrogen atoms and 1 oxygen atom. The hydrogen atoms as tiny magnets, and thus are sensitive to magnetic fields. Since MRI machines themselves are composed of a magnet, radio waves, a gradient, and a computer, this works perfectly. The movement of these hydrogen atoms become in sync with the magnetic stimulation, and those which aren’t (“Low-energy water molecules”) take provided energy from the radio waves and begin to move. The movement of low-energy water molecules, specifically, from a higher energy state to a lower one, is captured by a computer that is used to create an image.

Diagram from ResearchGate illustrates the relationship between neural activity and BOLD signals

Functional-Magnetic Resonance Imaging (fMRI) techniques work similarly to MRI machines. However, fMRI machines have the ability to detect differences in the magnetic properties of oxygenated and deoxygenated blood → Identify changes in oxygenated blood in different regions of the brain through Blood-Oxygen-Level-Dependence (“BOLD” contrast). This type of imaging allows for the identification of areas of the brain which are more active during certain tasks which is crucial in identifying regions of functional connectivity.

Functional Connectivity

Connectomics is defined by slightly different individual neuroscientists and physicians, but they overlap to come to a definition similar to the following:

“Connectomics is the study of the brain’s structural and functional connections between cells, which is visualized as a connectome” — News Medical

The difference between structural and functional connections is that the former describes the physical connections between different parts of the brain while the latter describes areas of the brain which work together for functional tasks. fMRI data allows users to determine functional connectivity with BOLD signals and outline connections between different regions-of-interest (ROIs) in the brain.

Image of different nodes identified during default preprocessing with CONN toolbox

A connectome ring can be drawn with the connections or lack of connections between certain nodes.

An example of a connectome ring. Different colors represent the strength of connections: blue represents the lack of functional connections and red represent strong functional connections

However, barriers to discovering the human connectome also lie in the discovery of structural connections down to each individual synapse and neuron.

Making a Movie — Structural Connectivity

Mapping the structural connections is different than mapping the functional ones. Imagine identifying the tastes of a food vs the ingredients that make up that food (if there were 86 billion different ingredients). This task requires numerous steps that all need precision and time. The video below gives a 2 minute summary of how images of different brain slices are taken:

Once these images of the different slices are taken, they are layered on top of one another:

Image from ResearchGate shows the process of analyzing brain slices

Individual neurons that are identified are colored and their form can be seen as the slices are iterated through.

Below is an image of a neuron identified on the initial slice

Slide from Sebastian Seung: I am my connectome

After slices are taken away, the shape and connections of these neurons become apparent.

Slide from Sebastian Seung: I am my connectome

Now, this obviously presents an issue. To color in every single neuron in every slice for every part of the brain, would be like mowing a farm by using kindergarten scissors. As in this case, it is much more efficient and productive to have a machine do it. Specifically, using deep learning for identification can speed this process up tremendously.

Deep Learning

Deep Learning is a part of machine learning, which is a part of artificial intelligence. It focuses on using the human brain as inspiration to develop various models called “Neural Networks”. There are a variety of types of neural networks; some are used for image classification by Tesla, while others can be used to diagnose patients.

While machine learning seems intimidating at first, simple models can be created within minutes.

This part of the code I used to create a Multi-Layer Perceptron (a type of feed-forward neural network) that is used for image classification. It predicts the image passed in is the number “5” (trained on numbers 0–9) with ~98% accuracy.

Methods used for connectomics will be much more sophisticated than this, but the idea of image classification and application of neural nets remains the same.

Wrapping it Up 🔑

Connectomics is revolutionizing the way that we approach biology, and it will forever change the way we treat and diagnose patients.

  1. Connectomics can be used to treat various diseases including schizophrenia, dementia, bipolar, and many others. By mapping the individual connections in the brain we can better understand these illnesses and personalize treatment to the patients that suffer from them.
  2. Connectomics can potentially also be used to “download” or store memories. This is because it is believed our memories are stored in our synapses and the mapping of someone’s synapses would allow their memories to be recognized and thus, downloaded.
  3. Connectomics can also forever change artificial intelligence. Neuroscience and machine learning give back to each other and as our understanding of the brain develops so will the development of machine learning.

Above are just some of the applications of this truly phenomenal science.

Unlocking the brain’s potential is the key to advancing humans forward in a way that we cannot even imagine. It’s impact immeasurable and profound.

TEDx Speaker | HYRS Alum (Neurosurgical RA) | TKS Student | SHAD Alum | 2021 Calgary Brain Bee Winner