By the end of 2023, The first drug The U.S. Federal Drug Administration has approved a drug with the potential to slow the progression of Alzheimer's disease. Alzheimer's is one of many debilitating neurological disorders that together affect one-eighth of the world's population, and while the new drug is a step in the right direction, it, like other similar diseases, still has a long way to go to fully understand it.
“Reconstructing the complexities of how the human brain works at the cellular level is one of the greatest challenges in neuroscience,” says Lars Gjesteby, a technical staff member and algorithm developer at MIT's Lincoln Laboratory. Human Health and Performance Systems Group“High-resolution, networked brain atlases can help improve our understanding of disorders by pinpointing differences between healthy and diseased brains. However, progress has been hampered by a lack of tools to visualize and process very large brain imaging data sets.”
A network brain atlas is, in essence, a detailed map of the brain that can help link structural information to neural function. To build these atlases, brain imaging data needs to be processed and annotated. For example, each axon, or thin fiber that connects neurons, needs to be traced, measured, and labeled with information. Current methods of processing brain imaging data, such as desktop software or textbook-oriented tools, are not yet designed to handle data sets on the scale of the human brain. As a result, researchers often spend a lot of time toiling through an ocean of raw data.
Gjesteby is leading a project to build the Neuron Tracing and Active Learning Environment (NeuroTrALE), a software that combines machine learning, supercomputing, and ease of use and access for this brain mapping challenge. NeuroTrALE automates much of the data processing and displays the result in an interactive interface that allows researchers to edit and manipulate the data to flag, filter, and look for specific patterns.
Unraveling a ball of thread
One of the defining features of NeuroTrALE is the machine learning technique it employs, called active learning. NeuroTrALE algorithms are trained to automatically label incoming data based on existing brain imaging data, but unknown data can present the potential for errors. Active 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 accurate data processing with much less burden on the user.
“Imagine taking an x-ray of a ball of yarn. You’d see all these crisscrossing and overlapping lines,” says Michael Snyder of the lab’s Homeland Decision Support Systems Group. “When two lines cross, does that mean one of the yarn strands is making a 90-degree bend, or is one going up and the other going down? With NeuroTrALE’s active learning, users can trace these yarn strands once or twice and train the algorithm to follow them correctly as it goes. Without NeuroTrALE, the user would have to trace 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 the majority of the labeling burden off the user, it enables researchers to process more data faster. Additionally, axon tracing algorithms leverage parallel computing to distribute calculations across multiple GPUs at once, resulting in even faster and more scalable processing. With NeuroTrALE, The team demonstrated a 90 percent reduction in the computing time required to process 32 gigabytes of data compared to conventional ai methods.
The team also showed that a substantial increase in data volume does not translate into an equivalent increase in processing time. For example, in a recent study They showed that a 10,000 percent increase in data set size resulted in only a 9 percent increase and a 22 percent increase in total data processing time, using two different types of central processing units.
“With the roughly 86 billion neurons forming 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 algorithm developers on the project. “This tool has the potential to automate the creation of connectomes not just for one individual, but for many. That opens the door to studying brain diseases at a population level.”
The open source path to discovery
The NeuroTrALE project was formed as an internally funded collaboration between Lincoln Laboratory and Professor Kwanghun Chung The Lincoln Laboratory team needed to build a way for researchers at the Chung Lab to analyze and extract useful information from the vast amounts of brain imaging data coming into the lab. MIT Supercloud — a supercomputer operated by Lincoln Laboratory to support MIT research. Lincoln Laboratory’s expertise in high-performance computing, image processing, and artificial intelligence made it uniquely suited to meet this challenge.
In 2020, the team uploaded NeuroTrALE to SuperCloud and by 2022 the Chung Lab was producing results. In one study, published in ScienceThey used NeuroTrALE to quantify the density of cells in the prefrontal cortex in relation to Alzheimer's disease, where brains affected by the disease had a lower cell density in certain regions than those without the disease. The same team also located where in the brain damaging neurofibres tend to tangle in Alzheimer's-affected brain tissue.
Work on NeuroTrALE has continued with funding from Lincoln Laboratory and the National Institutes of Health (NIH) to develop NeuroTrALE's capabilities. Currently, its user interface tools are integrating with Google Neuroglancer Program: An open-source, web-based neuroscience data visualization application. NeuroTrALE adds the ability for users to dynamically visualize and edit their annotated data and for multiple users to work on the same data at the same time. Users can also create and edit a variety of shapes, such as polygons, points, and lines, to facilitate annotation tasks, as well as customize the color display of each annotation to distinguish neurons in dense regions.
“NeuroTrALE offers an end-to-end, platform-agnostic solution that can be easily and rapidly deployed in standalone, virtual, cloud and high-performance computing environments via containers,” said Adam Michaleas, a high-performance computing engineer at the lab. technology-office/artificial-intelligence-technology“>artificial intelligence technology Group“It also significantly improves the end-user experience by providing real-time collaboration capabilities within the neuroscience community through data visualization and simultaneous content review.”
To align with The mission of the NIH The team's goal is for NeuroTrALE to be an open-source tool that can be used by anyone. And this kind of tool, Gjesteby says, is what's needed to reach the ultimate goal of mapping the entire human brain for research and, ultimately, drug development. “It's a community-based effort where the data and algorithms are meant to be shareable and accessible to everyone.”
The code bases for the axon tracing, Data managementand Interactive user interface NeuroTrALE's libraries are publicly available through open source licenses. Please contact Lars Gjesteby for more information on using NeuroTrALE.
