Brain Analysis

Building Better Brain Imaging Analysis Tools

Erin Mazerolle uses brain imaging to understand, for example, what’s happening in the brain of someone with Alzheimer’s disease.

“A lot of the work that I have done is to try to do a better job of understanding and diagnosing neurological diseases, and to take really advanced cutting-edge research imaging techniques — new ways to image the brain — and apply them for the first time in patient populations,” says Mazerolle, assistant professor of psychology and computer science at St. Francis Xavier University.

Over the course of this research, Mazerolle noticed that she and her peers don’t always do a good job at making sure their results are reproducible in other studies.

“If I do a study and publish it, and then another researcher reads it, can they get the same result?” she asks. “This is the part of my work where ACENET really comes in. I have a study going on right now in which we’re trying to figure out how the different ways that we analyze data are impacting the reproducibility of our results.”

She says that when analyzing brain imaging data, there are dozens of steps researchers take, and each one involves deciding which parameter to use at that juncture. And ultimately, all of those decisions impact the final result.

“There’s an analogy that we use called the ‘garden of forking paths.’ I make a decision and go down one garden path and I never know what my results would have looked like over on the other path,” she says.

“So we’re developing software that researchers can use to simultaneously analyze the imaging data using a whole bunch of different paths. That’s a really major computational challenge because these datasets are very large and it takes a lot of computational time and power to get to the end of each of these paths.”

She says the clusters at ACENET reduce the computational challenge to a reasonable time frame, but even with those high-performance computing resources, it can still take a month to do the analysis for a single study.

“It would be completely unfeasible for me to do it without ACENET,” Mazerolle says.

“We hope that by looking at all of the outcomes of the different paths, we get a better sense of whether our results are real, or if the ones that we like only occur down one or two paths,” she says. “Or, do the 200 paths actually look totally different and we never would have known that if we hadn’t tried this multiverse analysis?”

For her multiverse research, she’s using data from the Alzheimer’s Disease Neuroimaging Initiative, an international data-sharing database that looks at Alzheimer’s Disease, cognitive impairment and healthy normal aging. In particular, she’s comparing super-agers (who are in their 80s or 90s and have the memories of 20- and 30-year-olds) and comparing them to people who are aging normally. This is the study that will pilot the tool she hopes will help further brain imaging research across Canada.

Creating a Non-linear Model of our Non-linear Brains

James Hughes likes to use applied machine learning to solve real-world problems for which answers aren’t otherwise readily available. An assistant professor in St. Francis Xavier’s computer science department, his research can apply to kinematics, geology, music, finance and the human brain. He was recently, for example, doing some machine-learning modelling on people with Parkinson’s, studying them while they were using treadmills. To do his work, he uses the high-powered computing resources of Compute Canada and ACENET. “We were using Compute Canada resources to come up with mathematical functions on how people walk,” he says. “I was also working towards coming up with a predictive model for traumatic brain injury patients that will tell you what a subject’s intercranial pressure is. The best way to test that is a very dangerous neural surgery. I was working to see if we could find another way that’s safer.” He calls the human brain a non-linear computational system. But the types of algorithms and math currently used when trying to understand the human brain — those typically used in functional magnetic resonance imaging studies, for example — are done with strictly linear tools. “This means the models we end up with at best can be linear approximations of a non-linear system,” Hughes says. “This is not to say that all of this stuff done up to date has to be thrown out. We’ve been able to make a lot of big contributions. However, one can’t help but wonder what we could do and learn if we had a non-linear model of the brain that more accurately represents the underlying system.” To do his non-linear modelling, he uses “heavy-duty” machine learning, artificial intelligence and what he calls “evolutionary computation.” Think of it as designing a car, he says. “What happens if I come up with 100 random car designs?” he asks. “They will all be pretty bad, but what if I take two of the less bad ones? I’ll combine them and hopefully make a better car. If I keep doing this millions and millions of times, I will typically end up with a pretty good car design. It’s the same idea. You pick the ones that are relatively better.” He was once studying how a brain interpreted or dealt with a specific language task. He asked his subjects to figure out how words related to one another. “If you use the traditional modelling technique, it would say ‘OK, so the brain regions X, Y, Z are related to each other,’” he says. “But with non-linear models, we found that region X, Y, Z and A, B, and C are related because during the study, they started to react in a non-linear way.” “But the number of possible explanations for all of these studies are just astronomically large,” he says. And that’s the reason, he requires intensive computer resources. In 2018 alone, he used “hundreds and hundreds” of core years. If he’d done the same work on a typical desktop computer, he would have had to hit “run” on the computer and return in 200 years for any findings, he jokes.

Decoding the Human Brain

Jacob Levman is trying to figure out what parts of the brain develop abnormally in children with autism, so it’s surprising to learn he’s a professor in the department of mathematics, statistics and computer science at St. Francis Xavier University in Antigonish, Nova Scotia. But it turns out it makes sense. His bachelor’s degree was in computer engineering and his master’s in electrical and computer engineering. “My research has involved various contributions in the statistics domain, so I guess I was a good fit when they sought a Canada Research Chair in bioinformatics, which is a combination of medical stuff and computer technology.” Indeed it was. Levman’s PhD from the University of Toronto was in medical biophysics. “My background is really interdisciplinary,” he says. “Medical biophysicists are the people who build better MRI machines and come up with therapies such as radiation treatments. My research combined my background in computers with MRI data and MRI technology.” Levman did post-doctoral studies at Oxford’s department of biomedical engineering, looking at how to apply advanced computer technologies to help predict or identify the extent of tissue damage due to stroke from MRI examinations. His post-doctoral studies at Harvard looked at both healthy and autistic patients in pediatrics at Boston Children’s Hospital Today, doctors diagnose autism by symptoms and behaviours, but Levman thinks there are probably underlying anatomical or physical issues associated with autism.” “We look at the differences in the way the brains of children with autism present,” he says. “Potentially, our work could help inspire new therapies by helping us better understand which regions of the brain are developing abnormally and in what way.” “We’re also looking at the genetic profiles of children with autism and comparing those with the presentation of the brain from an MRI examination,” he says. “If we can say an abnormal mutation in autism seems to be highly related to this area of the brain forming abnormally, that might help decode the mysteries behind autism.” To that end, he’s completed the world’s largest autism-MRI study, reviewing the volumes of substructures in the brain. But his work also has a technical side, through which he develops general-purpose machine-learning algorithms or pattern-recognition technologies that help extract measurements from examinations using existing formal computer science methods and new ones he’s trying to create. To do his work, he uses ACENET and Compute Canada computer clusters. Just to do a basic analysis of an MRI from one child with autism, extracting thousands of measurements from across the brain, would take nine hours on one computer. A large study involves about 2000 examinations, which would take about 18,000 hours, or more than two years. “If you want to get the job done in a reasonable amount of time, you need to do parallel processing — breaking apart a huge job into a bunch of smaller ones that run simultaneously on supercomputers ,” Levman says. “ACENET and Compute Canada’s resources are very helpful for this. Without them, it would be impossible to get these things done in a reasonable time. In addition to using their computer resources, Levman has used ACENET’s training programs for every one of his lab students. “It’s been extremely useful — essential, actually,” he says.

High Powered Computing Helps in Brain Analysis

Christine Lackner studies child and adolescent brain development as it relates to the ability to control thoughts and feelings. Her research may one day help to find interventions that will positively affect brain function. By fitting an EEG cap — think of a tight-fitting shower cap — on the heads of her participants, she studies brain activity using between 64 and 128 embedded electrodes. She looks at the brains of “typically developing” children and teens because no two brains develop the same way, but she also looks at those with self-regulation problems, such as ADHD, as well as learning disabilities. Even more recently, she’s also studied adolescents and young adults with cerebral palsy and is examining the effects of stressors on their developing brains. “The skillset I’m most interested in is self-regulation and executive function,” says Lackner, an assistant professor at Mount Saint Vincent University in Halifax. “I’m looking at how they control their emotions, make plans, resist temptation. In some cases, we give them computer tasks that require them to act fast and force them to make a lot of errors. We then associate their brain responses with how they are functioning in the real world.” She has worked in conjunction with researchers at Ryerson University and with a unique program called Integra, which trains children with comorbid self-regulation (children with ADHD, for example) and learning disabilities in mindfulness meditation and martial arts. “We’re studying these children’s brain-wave patterns before, during and after they go through this program,” she says. The researchers add input from parents to their analysis and look at the changes over time. To do this work, she produces massive data files. Every millisecond, the EEG takes a reading on the (up to) 128 electrodes on the participant’s head. “I record for upwards of two hours at a time,” she says. “So, as you can imagine, those files get very big.” Just storing those files is one challenge she addresses with the help of ACENET and Compute Canada, but she also runs a lot of her analysis through the same resources. “We’re on the cutting-edge of analyzing these data files,” she says. “What we do is called independent components analysis. We put all of the data into this big matrix and ask the computer to pull out meaningful patterns. It’ll tell us how a brain network is acting at a particular time. It more accurately isolates the activity being generated without interference from other signals.” Interference can be as simple as neck tension, which has nothing to do with a participant’s brain activity, but which, before this new system emerged, would be somewhat difficult to remove from the data. Lackner says without ACENET and Compute Canada, her analyses would take months instead of days. One goal of her research is to demonstrate that interventions such as Integra, mindfulness and martial arts are altering brain function in a positive way, so that policy-makers direct funding to such programs.