Detangling the human brain with open-source AI

Welcome to the 10 new deep divers who joined us since last Wednesday. If you haven’t already, subscribe and join our community in receiving weekly AI insights, updates, and interviews with industry experts straight to your feed.

DeepDive

Your weekly immersion in AI. 

The human brain is more complex than any other known structure in the universe. 

So neurological conditions, from Alzheimer’s disease to epilepsy, are difficult for neuroscientists and neurologists to truly understand – and conventional treatments can’t target the specific areas of the brain that need support. 

People living with neurological disorders often have to also live with medications that have a blanket effect across the brain, rather than targeting the problem; and they come with the potential for significant side effects. 

In a recent MIT blog post, Lars Gjesteby (Technical Staff Member and Algorithm Developer at MIT Lincoln Laboratory’s Human Health and Performance Systems Group) said: 

“Reconstructing the intricacies of how the human brain functions on a cellular level is one of the biggest challenges in neuroscience. High-resolution, networked brain atlases can help improve our understanding of disorders by pinpointing differences between healthy and diseased brains. 

“However, progress has been hindered by insufficient tools to visualise and process very large brain imaging datasets.” 

What is a networked brain atlas? 

It’s a detailed map of the brain that can enable neuroscientists to connect structural information with neurological function. 

A networked brain atlas requires a vast amount of brain imaging data to be processed and annotated before it can be effectively analysed. Every tiny fibre between neurons must be traced, measured, and labelled – and existing methods of processing brain imaging data don’t yet have the capacity to handle datasets on the scale of the human brain. 

This means that researchers have to spend many hours sifting through an overwhelming expanse of raw data in order to identify any relationships between brain structure and function. 

NeuroTrALE aims to solve the brain mapping challenge 

Gjesteby is the lead researcher of a project that aims to build the Neuron Tracing and Active Learning Environment (NeuroTrALE). It combines machine learning and supercomputing with technologies to drive ease of use and access to brain mapping processing. 

Essentially, NeuroTrALE automates a lot (but not all) of the data processing, and then displays its outputs via an interactive interface. Researchers can use this interface to edit and manipulate the data in order to search for specific patterns and filter results. 

The project uses active learning to navigate unfamiliar data 

Key to NeuroTrALE’s usability is a machine learning technique called active learning. This means that while incoming data is automatically labelled based on existing data from brain imaging, users can manually correct errors that occur when unfamiliar data is input into the system – and these manual corrections teach the algorithm what to do next time it encounters similar information. 

So NeuroTrALE isn’t fully automated – and this is a good thing. The inclusion of manual corrections in the labelling process means that accuracy can be protected. 

But because the system does most of the labelling automatically, researchers are able to process data faster.

“With the estimated 86 billion neurons making 100 trillion connections in the human brain, manually labelling all the axons in a single brain would take lifetimes,” said Benjamin Roop (an algorithm developer for NeuroTrALE) on the MIT blog. “This tool has the potential to automate the creation of connectomes for not just one individual, but many. That opens the door for studying brain disease at the population level.”

Going fully open-source 

Research and development of NeuroTrALE has been funded by Lincoln Laboratory and the National Institutes of Health (NIH), with the aim of increasing the system’s capabilities and usability. 

At time of writing, the project’s user interface tools are being integrated with Google’s Neuroglancer – an open-source web-based application for viewing neuroscience data. 

Going forward, the goal is to make NeuroTrALE completely open-source for anyone to use. Because the team behind it believe that this kind of tool is what we need in order to reach the end goal: mapping the entire human brain for research and drug development. 

The best of open-source AI? 

We want to know about the open-source projects you think will have the biggest impact on human life. Open this newsletter on LinkedIn and tell us in the comment section. We’ll see you there.