Sujay Nagaraj: Using machine learning to automate catheter access detection

We recently caught up with Sujay Nagaraj. He's one of 10 U of T trainees selected to present his research exploring new ways to use artificial intelligence in medicine. He will discuss his research in a public T-CAIREM Trainee Rounds presentation on April 13. Sujay is currently pursuing a dual MD/PhD degree, with his PhD in the department of computer science. His co-advisers are Drs. Anna Goldenberg and Sebastian Goodfellow. 

What inspired you to pursue this field?
As technology and innovation drive the evolution of clinical medicine, it is imperative that there are practitioners who are experts in multiple domains to be leaders in this field. I aim to be a clinician who fluently speaks the languages of medicine, mathematics and computer science to translate innovation into better outcomes for patients.

What excites you the most about the possibilities of AI in healthcare? What do you see as the biggest challenge to the field of AI in healthcare right now?
My excitement about this growing space comes from a cautious optimism. There is great potential for AI in healthcare to improve quality of care, boost efficiencies in our healthcare system, and help us better understand complex human physiology. However, there is also great potential for harm and worsened inequities. I am (cautiously) excited to be part of the broader conversation regarding how AI will fit into healthcare and how this work can be done responsibly, safely and equitably.                

What do you like to do when you aren’t working?
I love being outside – hiking or camping! I am a (very poor) violin player and a (somewhat less poor) rock climber.

+++

Sujay will be presenting this research titled: "Development and implementation of a machine learning tool to automate vascular catheter access detection in a pediatric critical care unit"

And here's some more information: In the Critical Care Unit (CCU), vascular catheters are used routinely for blood sampling, pressure measurement, and drug administration. Awareness of catheter utilization is required to reduce the associated adverse outcomes (i.e.: bloodstream infections). Current means of documenting utilization are manual and subject to error. We developed a novel machine learning tool to automate catheter access detection in real-time. Our model can help augment clinician insight about catheter utilization patterns, and we are currently in the process of validating this tool at the bedside.