Computer Engineering Trainee Spotlight: Pradeep Raj Krishnappa Babu, PhD

After passing the GATE 2012, a rigorous exam in his native India, Pradeep Raj Krishnappa Babu, PhD, jumped at the opportunity to pursue a master’s degree in engineering, where his passion for virtual reality and machine learning began. Later, while participating in an international collaboration experience as a postdoctoral fellow, he visited Duke and met James B. Duke Distinguished Professor of Electrical and Computer Engineering Guillermo Sapiro, PhD, who, along with Geraldine Dawson, PhD, leads research using computer vision analysis to create and validate a digital phenotyping tool to screen for autism. Babu had a “lightbulb moment.”

“When I realized that computer vision analysis could be used to help screen for autism or developmental challenges for kids, this awakened a new interest for me,” says Babu. “With it, we can capture data — even miniscule changes in facial movements — that are way beyond the ability for human observation.”

In the past, research on facial expressions usually required manual coding of observations from recorded videos that are often complex and both labor and time intensive for clinicians. These methods are difficult to deploy at scale and universally. As a member of Sapiro’s research team, Babu has helped develop advanced computer vision algorithms to capture the distinct, tiny changes in facial movements that are evoked while children watch carefully designed, developmentally appropriate, short, fun movies.

“We are exploring whether these subtle differences in behaviors can help us identify infants as young as six months who are showing early signs of autism. Using these, along with provider and caregiver questionnaires, we have the potential to help more children get earlier diagnosis,” says Babu. “Research has shown that interventions canpromote improved language and social skills for life. Our work could help kids get started on these earlier.”

Working with Sapiro, Babu has helped to develop an algorithm that relies on capturing the complexity of facial expressions and facial landmark movements, not on “reading” emotions, which are sometimes challenging for autistic children to express. Currently, they are refining the computer vision analysis, focusing on analysis of even more subtle differences in head movements, blink patterns, and attention in autistic toddlers. Babu has published research in the Journal of Child Clinical Psychology and Psychiatry and in IEEE Transactions on Affective Computing.

“Our hope is also that these tools will help clinicians monitor behavioral changes so they can more quickly adjust and modify intervention strategies for children needing support,” says Babu.