December 17, 2019
Affiliate Profile: Seong Jae Hwang
Featured Articles
Affiliate Profile: Seong Jae Hwang
Seong Jae Hwang, assistant professor, School of Computing and Information, University of Pittsburgh

“Complex models can help understand complex things,” said Seong Jae Hwang, assistant professor in the Department of Computer Science at the University of Pittsburgh.

That idea is at the heart of his work, which seeks to model complex things, such as the progression of diseases, in ways that provide insights regular data may not.

Hwang’s work with medical imaging is a perfect example. During his PhD studies at the University of Wisconsin – Madison, Hwang worked with an advisor doing interdisciplinary work with the Wisconsin Alzheimer’s Disease Research Center (WADRC).

“These medical images of diseases like Alzheimer’s often have subtle progressions,” Hwang said, “patterns and trends that can’t be detected just by looking at the images.”

Sophisticated modeling, by contrast, can provide a new perspective on those patterns by analyzing progressions previously thought too subtle to be detected. Hwang views modeling as offering new ways of looking at the same data doctors and researchers have, and trying to extract the underlying information.

He cited a project he worked on at Wisconsin that sought to understand what was happening in the brains of patients who were at risk of developing Alzheimer’s but were still cognitively healthy. The question is important because early detection and treatment can be critical in Alzheimer’s cases. It was a challenging project, Hwang said, because the abnormalities were extremely subtle.

Modeling allowed Hwang and fellow researchers to analyze the brain connectivity of healthy individuals and those at greater risk of developing Alzheimer’s. They found evidence that a number of structural connections within the brain were weaker among those at higher risk of developing the disease—an insight that was not readily available to neuroscientists and others analyzing raw data or medical images.

Of course, insights like this are just the beginning of a long scientific journey: the neuroscience and neuroimaging experts Hwang worked with must form and test new hypotheses, in addition to pursuing additional funding.

But the opportunity to help better understand a disease like Alzheimer’s, bringing scientists closer to a possible cure, is at the heart of what excites Hwang about his work.

“One of the main reasons I wanted to work in this area,” Hwang said, “is that I wanted my research to have a more direct impact in terms of science.”

Hwang came to modeling, and to the Modeling and Managing Complicated Systems (momacs) Institute, through his interest in the still-evolving field of computer vision. A subfield in artificial intelligence, computer vision seeks to train computers to understand higher-level information about images, such as the location or context of an object in a photograph and its relation to the objects surrounding it.

That makes computer vision a natural fit, Hwang said, for modeling the kinds of complex patterns found in medical images.

“Prediction is one thing, but what I want to do is represent data in new ways that show an underlying progression that couldn’t be observed in the data directly,” Hwang said.

Like a neuroscientist, he is trying to present a “master perspective” of the disease’s progression. “This is just from a different angle,” he said.

That sense of working alongside doctors, neuroscientists, and neuroimaging experts informs Hwang’s work flow. He talks regularly with experts at the WADRC and has begun creating relationships with Pittsburgh Alzheimer’s researchers since arriving at Pitt in fall 2019. Hwang asks what they are seeking to do and what challenges he can help with, seeking to determine whether there are models he has been building that could be modified or improved to better fit the researchers’ goals.

In ongoing work, Hwang goes back to the neuroscientists and neuropsychologists after running a model to talk about the results. “Does this have any meaning?” Hwang might ask. “Does this look promising?”

“What matters is if it’s meaningful to them,” he said. “Are there associations here they couldn’t make with just their data?”

Hwang’s research includes more than just medical imaging, extending to more technical questions concerning computer vision. But his work investigating diseases like Alzheimer’s has sparked an enduring sense of the possibilities of modeling.

“Computer vision is very fun, and it can have a lot of interesting applications, but I also wanted to make contributions that could impact the daily lives of people,” Hwang said. “With modeling it’s very clear what my research could lead to if I can answer this hypothesis.”