Patients supported by extracoporeal membrane oxygenation (ECMO) or a left ventricular assist device (LVAD) have an incredible amount of data associated with their clinical course. These are important life-saving treatments for many patients, but are associated with a great deal of morbidity and mortality. We are using modern big data analytical techniques combined with machine learning to create predictive models capable of providing insight to aid in the complex clinical decision making. We are creating databases directly from the EHR at an unprecedented scale, such as over 3 million lab values and over 500k echo/cath results to try and improve patient care.

Dr. Ghazi and his team of biomechanical engineers have made incredible progress in the development of high-fidelity surgical simulation models. These 3D printed models are built from actual patient CT scans and can be sutured, cauterized and will bleed just like real tissue. We have been working for the past year to develop a cardiac model to create a realistic surgical simulation for cardiac procedures without the need of animal models. The beating of the heart is controlled by a custom code run on an arduino with an associated touch screen user interface to let the practicing surgeons control rate and rhythm.

Lung cancer is the second most common type of cancer and the number one cause of death from cancer worldwide. The National Lung Screening Trial demonstrated evidence that annual low-dose CT scans significantly reduce mortality for high-risk individuals. However, the trial draws criticism because of its high rate of false positive findings, leading to unnecessary stress and invasive procedures for many patients. The aim of our study is to use machine learning to better classify lung nodules as benign vs malignant directly from low-dose CT scans. We were approved to receive over 15,000 CT scans from the NLST trial database and are fortunate to have access to the University of Rochester's BlueHive linux cluster with 210 teraFLOPS of computing capacity to train the algorithm.

Dr. Wakeman and the URMC pediatric surgery team have created flowsheet algorithms for determining when a pediatric trauma patient requires imaging. Implementation of these clinical resources in the ED, which describe the proper utilization of imaging for this patient population, has decreased unnecessary radiation for our pediatric patients. I am working with the team to create an easily distributed and user-friendly mobile app for these algorithms. The interactive mobile app was built on the react native framework and is currently in clinical beta testing in the ED for both iOS and Android devices.