Project Spotlight: Realtime Evaluation for Adverse Events using Intraoperative Neurophysiological Monitoring (READE-IONM)

Stroke during surgery is a devastating complication resulting in post-operative disability, or even death, and can present a financial and social burden to patients and their families. Stroke rates are estimated at 6% for cardiovascular and neurological surgery and about 1% for general surgery, resulting in an estimated 550,000 perioperative strokes in the US per year. Lifesaving interventions are available if the stroke is detected early and this team is developing tools to enhance and partially automate that process, making it more widely available and ultimately improving patient outcomes.

Please share a little about your background and your research experiences.

Dr. Parthasarathy D. Thirumala, MD, is an associate professor at the University of Pittsburgh and Director of the Center for Clinical Neurophysiology, where he specializes in real-time intraoperative neuromonitoring. The primary responsibility in this job is to evaluate the status of the brain, spinal cord, and nerves to diagnose neurological injury with intraoperative neurophysiological monitoring (IONM). IONM is performed continuously in real-time during neurological, orthopedic, and cardiovascular surgery. This is similar to how an electrocardiogram is used to diagnose problems with the heart during surgery. The Center for Clinical Neurophysiology (CCN) research is focused on improving the diagnostic accuracy of neurophysiological monitoring to detect neurological disorders like stroke, spinal cord, and peripheral nerve injury.

Walk us through your project.

Our team is building a machine learning (ML) algorithm, which can continuously review electroencephalography (EEG) data in real-time during surgery to detect stroke. Diagnosis of stroke in real-time will help physicians offer approved lifesaving therapies for patients. Currently, we perform real-time visual analysis of the EEG data, which limits quantitative evaluation required for detection of ischemia and stroke. Lack of consensus on the quantitative EEG criteria for detection of stroke makes visual analysis even more challenging and highly variable in clinical practice. Finally, the current software platforms lack the capability to perform real-time advanced quantitative analysis. Furthermore, trained personnel for visual monitoring are not available in all hospitals in the US, with even fewer trained personnel around the world. Thus, the key challenge is the lack of automated methods for real-time intraoperative monitoring to detect brain ischemia. Automated ML analysis, with alerts for brain ischemia, can help the neurologist offer patients lifesaving therapy like mechanical clot removal. It can also exponentially increase their ability to reliably monitor multiple surgeries by focusing on a specific time of the EEG.

How did you and your project partners come together? If this is your first time working together, how did you learn about one another and come up with this project?

Dr. Shyam Visweswaran, MD, PhD, was an advisor to the medical student Amir Mina, who wanted to spend his pre-doctoral rotation in my lab at CCN. Interestingly, Dr. Visweswaran is a neurologist by training but spends most of his time on research in Clinical Decision Support and Machine Learning in Medicine. Once the medical student expressed interest in our work, we decided to assemble a team and work on this clinical problem. It felt like an emotional homecoming for Dr. Visweswaran, who was excited about discussing ML solutions for neurological conditions. First, we engaged the service of machine and deep learning expert Dr. Kayhan Batmanghelich, PhD. Second, we engaged Dr. Jeremy Espino, expert in agile software development and machine learning. Finally, we added research analyst Louisa Zhang and Stephanie Paras to help with scaling the ML models and labeling the data.

What got you interested in doing research aimed toward commercialization?

Adoption of approved clinical research is sometimes very slow in medical practice. Thus, we felt commercialization of our research will accelerate the development of solutions. Further, it will provide the necessary tools to increase the adoption of our technologies across the world to benefit patients. For example, cardiac surgery is a common procedure performed all over the world, and our algorithm to detect stroke during surgery can be deployed to enhance care for thousands of those patients using available EEG machines.

What has been the most rewarding part of your project so far? 

The most rewarding aspect of the project so far has been the interactions with the team in tackling our challenges for training and testing the algorithms.