For Immediate Release
September 27, 2013
This fall, about 2,000 juniors and seniors from high schools across Maryland will have the opportunity to help the Johns Hopkins University Applied Physics Laboratory (APL) in Laurel, Md., process hundreds of thousands of high-resolution brain tissue images to create a picture of how the cortex — the outer part of the brain where most cognitive processes are thought to occur — is wired at a single neuron level. The endeavor is the inaugural project for the Maryland Business Roundtable for Education’s STEM Challenge Program, created to provide high school students the opportunity to develop solutions to real-world problems posed by industry experts.
Earlier this year the Roundtable (MBRT) asked its industry partners for science, technology, engineering or math (STEM) projects that students might be able to impact. At the time, a team of scientists in APL’s Research and Exploratory Development Department (REDD) were grappling with their own challenge. As part of REDD’s Electron Microscopy (EM) Connectome project to transform detailed pictures of brain structures into a computational model, REDD scientist Dean Kleissas built a crowd-sourcing web application for proofreading automated image processing results on large datasets.
“To build a wiring diagram of the cortex, researchers must find each neuron and all of its connections to other neurons,” said REDD’s Jacob Vogelstein. APL uses computer vision algorithms (developed by Kleissas and William Gray Roncal of APL’s Force Projection Department) to analyze high-resolution pictures of the brain to find these connections between nerve cells, called synapses. “The only problem,” explained Vogelstein, “is that there are millions of synapses in every cubic millimeter of the cortex, so even an algorithm that is 99% accurate will make thousands of errors.”
APL STEM Program Manager Dwight Carr thought this would be an ideal project for Maryland high school students to take on, and MBRT agreed. “Real workplace challenges being tackled by real workplace professionals are key to opening the eyes and minds of students and giving them a sense that they could play a role in changing the world for the better,” said MBRT Executive Director June Streckfus.
In the first part of the challenge, students and teachers will become familiar with the image data and learn about what they are looking at. “Understanding the neuroanatomy and what synapses look like in electron microscopy images is vital for this research and an acquired skill,” said Vogelstein. Then, using a web application developed at APL, the students will vote on whether they think the computer has labeled synapses correctly — this is called “proofreading.”
Students and their classmates will race other schools to proofread as many synapse detections as possible. The goal for each school is to check at least 20,000 images. The school that correctly identifies the most synapses will receive a trophy.
In another optional part of the challenge, students will develop their own automated synapse detector and compete against existing state-of-the-art algorithms and other student teams. Participants will be given two sets of EM images of the brain, one with the true labels for synapses, and one that is unlabeled. Students will submit a short white paper detailing their algorithm and implementation, code and installation instructions in Matlab or Python, and a labeled volume containing the output of their synapse detection algorithms. Solutions must be fully automated with no human intervention.
“These students will be helping us tackle a critical issue: understanding how the fundamental computing circuits of the brain work,” Vogelstein said. “We hope this leads to the development of more intelligent artificial intelligence: computers and robots that make decisions more like people than like today’s systems.”
MBRT will identify the participating schools in October. Additional information about participating in the challenge is available online at http://mbrt.org/stemnet/coordinators/. The winner of the STEM Challenge will be featured on APL’s STEM website, http://www.jhuapl.edu/STEM/, and showcased at MBRT’s annual meeting.
The Applied Physics Laboratory, a not-for-profit division of The Johns Hopkins University, meets critical national challenges through the innovative application of science and technology. For more information, visit www.jhuapl.edu.