At the end of 2023 first medicine with the potential to sluggish the progression of Alzheimer’s disease has been approved by the U.S. Federal Drug Administration. Alzheimer’s disease is one of many devastating neurological diseases that collectively affect one-eighth of the world’s population, and while the up-to-date drug is a step in the right direction, there is still a long way to go to fully understand the disease and others like it.
“Reproducing the intricacies of human brain function at the cellular level is one of the greatest challenges in neuroscience,” says Lars Gjesteby, a research scientist and algorithm developer at MIT Lincoln Laboratory. Health Systems and Human Performance Group“High-resolution, networked brain atlases have the potential to help us better understand disorders by showing differences between healthy and diseased brains. However, progress has been hampered by inadequate tools for visualizing and processing very large brain imaging datasets.”
A network brain atlas is essentially a detailed map of the brain that can lend a hand connect structural information to neural function. To build such atlases, brain imaging data must be processed and annotated. For example, each axon, the slim fiber connecting neurons, must be traced, measured, and labeled with information. Current methods for processing brain imaging data, such as desktop software or hands-on tools, are not yet designed to handle datasets on the scale of the human brain. As a result, researchers often spend a lot of time wading through oceans of raw data.
Gjesteby leads the project to build the Neuron Tracing and Vigorous Learning Environment (NeuroTrALE), a software pipeline that combines machine learning, supercomputing, and ease of apply and access to this brain-mapping challenge. NeuroTrALE automates most of the data processing and displays the output in an interactive interface that allows researchers to edit and manipulate the data to tag, filter, and search for specific patterns.
Untangling a ball of yarn
One of NeuroTrALE’s defining features is the machine learning technique it uses, called vigorous learning. NeuroTrALE’s algorithms are trained to automatically label incoming data based on existing brain imaging data, but unfamiliar data can create potential errors. Vigorous learning allows users to manually correct errors, teaching the algorithm to improve the next time it encounters similar data. This combination of automation and manual labeling ensures correct data processing with significantly less burden on the user.
“Imagine taking an X-ray of a ball of yarn. You see all these crisscrossing, overlapping lines,” says Michael Snyder of the lab’s Homeland Decision Support Systems Group. “When two lines cross, does that mean one of the pieces of yarn is making a 90-degree arc, or is one going straight up and the other straight down? With NeuroTrALE’s active learning, users can track those strands of yarn once or twice and teach the algorithm to follow them correctly as they go along. Without NeuroTrALE, the user would have to track the ball of yarn, or in this case, the axons of the human brain, every time.” Snyder is a software developer on the NeuroTrALE team, along with staff member David Chavez.
Because NeuroTrALE takes most of the labeling burden off the user, it allows researchers to process more data faster. In addition, axon tracking algorithms apply parallel computing to distribute computations across multiple GPUs simultaneously, leading to even faster, scalable processing. Using NeuroTrALE, the team demonstrated A 90 percent reduction in the computational time required to process 32 gigabytes of data compared to conventional AI methods.
The team also showed that a significant boost in data volume does not translate into an equivalent boost in processing time. For example, in last examination showed that a 10,000 percent boost in data set size resulted in only a 9 percent and 22 percent boost in overall data processing time using two different types of CPUs.
“With an estimated 86 billion neurons and 100 trillion connections in the human brain, manually labeling all the axons in a single brain would take a lifetime,” adds Benjamin Roop, one of the project’s algorithm developers. “This tool has the potential to automate the creation of connectomes not just for one person, but for many. This opens the door to studying brain disease at a population level.”
The Path to Discovery with Open Source
The NeuroTrALE project is an internally funded collaboration between Lincoln Laboratory and Professor Kwanghun Chung lab on the MIT campus. The Lincoln Lab team had to create a way for Chung Lab scientists to analyze and extract useful information from the enormous amounts of brain imaging data that flowed into MIT SuperCloud —a supercomputer run by Lincoln Laboratory to support MIT research. Lincoln Lab’s expertise in high-performance computing, image processing, and artificial intelligence made it uniquely suited to take on the challenge.
In 2020, the team uploaded NeuroTrALE to SuperCloud, and by 2022, Chung Lab was producing results. In one study published in used NeuroTrALE to quantify prefrontal cortex cell density in relation to Alzheimer’s disease, with affected brains having lower cell density in some regions than brains without the disease. The same team also mapped where in the brain harmful nerve fibers tend to get tangled up in Alzheimer’s-affected brain tissue.
NeuroTrALE continues to be developed with funding from Lincoln Laboratory and the National Institutes of Health (NIH) to expand NeuroTrALE’s capabilities. Its current user interface tools are integrated with Google services Neuroglancarz program — An open-source web browser application for neuroscience data. NeuroTrALE adds the ability for users to dynamically visualize and edit their annotated data and allows multiple users to work with the same data at the same time. Users can also create and edit multiple shapes, such as polygons, points, and lines, to facilitate annotation tasks, and customize the color display for each annotation to distinguish neurons in dense regions.
“NeuroTrALE provides a platform-independent end-to-end solution that can be easily and quickly deployed in standalone, virtual, cloud, and high-performance environments via containers,” says Adam Michaleas, High-Performance Computing Engineer at the lab Artificial Intelligence Technology Group“Furthermore, it significantly improves the end-user experience by providing real-time collaboration opportunities within the neuroscience community through data visualization and simultaneous content review.”
To adapt to NIH Mission research products, the team’s goal is to make NeuroTrALE a fully open tool that anyone can apply. And that kind of tool, Gjesteby says, is what’s needed to achieve the end goal of mapping the entire human brain for research and, ultimately, drug development. “It’s a grassroots community effort where the data and algorithms are meant to be shared and accessible to everyone.”