Skip to Content

Molinaroli College of Engineering and Computing

  • Ph.D. candidate Deepa Tilwani

Deepa Tilwani: Pioneering AI to transform autism diagnosis

In 2019, the Mayo Clinic used artificial intelligence to predict heart failure up to 10 years in advance—technology that has the potential to save millions of lives. Valued at $19 billion in 2023, the AI healthcare market continues to increase. Computer science and engineering Ph.D. candidate Deepa Tilwani is building on this momentum, pioneering research into how this technology can be used for early autism diagnosis.

“Current approaches are not very flexible to understand the whole brain effectively. There is a lot more research needed in our understanding of autism,” Tilwani says. “The brain is complex and understanding it requires a lot of patience.”

Tilwani’s research has taken her from earning bachelor’s and master’s degrees in computer science in India to the Artificial Intelligence Institute at the University of South Carolina (AIISC). Her eventual journey to Columbia began in 2019 when Professor Amit Sheth, founding director of the AIISC, delivered a keynote speech on technology’s uses in healthcare. Tilwani contacted Sheth, who offered her a remote research internship. Today, Sheth is one of Tilwani’s co-advisors.

After her remote internship, Tilwani moved to South Carolina in 2022 to begin her Ph.D. studies in computer science. As a winner of the LINZ-BR4IN.IO Hackathon and a recipient of the AIISC’s Nirmala and Jashwantlal Clerk Memorial Scholarship for Women from India, she was already a promising student. Despite moving to another country, she did not feel afraid and was ready to work.

“I didn’t have any fears; I was excited,” Tilwani says. “I felt a little homesick, but I talked to my parents every day and still talk to them daily. That has become a tradition for us.”

Tilwani’s interest and research in AI is focused on neuroscience, due in part to her family’s history of autism. She works with the Carolina Center for Autism and Neurodevelopmental Disorders to use AI to analyze brain waves, paving the way for cost-effective, non-invasive early interventions. In a recent publication in the Journal of Bioengineering, Tilwani explored the potential of an electrocardiogram (EKG) as a predictor for early autism.

Autism is famously hard to diagnose, especially in young children. While early interventions such as intensive therapy can be beneficial, symptoms usually do not begin until at least two years old. Tilwani is working to create a tool to help doctors diagnose and begin interventions in patients as young as three-to-six months old.

Her research uses deep learning, a type of AI made of several layers of analytic processes, to analyze brain scans. The model analyzes EKG scans in autistic and non-autistic children to search for differences that can be used in diagnoses. This research can also be used to study language disorders in stroke survivors.

I want to build an innovative approach that can provide insights into these disorders. If I can contribute solutions, it will touch many lives.

- Deepa Tilwani

No current technology can fully comprehend these dynamic brain changes, and it is something Tilwani hopes to change. “We will get more insights on what is happening inside the brain, which will help professionals understand autism better,” Tilwani says.

Computer Science and Engineering Assistant Professor Christian O’Reilly has extensive research experience addressing autism and other neurodevelopmental disorders. As Tilwani’s co-advisor, he recognizes her growth as a researcher since arriving at USC.

“When you do research there’s a type of precision that you need to develop,” O’Reilly says. “To develop this kind of systematic way to approach a problem is a very fundamental research skill. From the first day I knew her, Deepa was always very driven.”

Tilwani also reflects on her growth since moving to the U.S. Her research has been well-received, winning Best Research Presentation at the South Carolina Autism and Neurodevelopmental Disorders Consortium Symposium in 2023. Tilwani says presenting her findings has boosted her confidence.

Her work is also guided by her advisors. In addition to Sheth and O’Reilly, she also works with professors Rutvik Desai, Svetlana Shinkereva and Jessica Bradshaw from the USC Department of Psychology.

“I truly admire my advisors, particularly because of the collaborations I formed in a short time at USC,” Tilwani says. “People here are curious, and I often brainstorm with my advisors. When I reach out to them, they will help me out.”

Tilwani is currently working on her dissertation, researching how AI can be used to understand the connections between brain regions. She is also working as a teaching assistant for O’Reilly’s Python coding class and hopes to continue her research and teaching in the future.

“I want to build an innovative approach that can provide insights into these disorders,” Tilwani says. “If I can contribute solutions, it will touch many lives.”

Challenge the conventional. Create the exceptional. No Limits.