OVERVIEW: In collaboration with Clinical Research Informatics (CRI) group within SC CTSI, data scientists from the USC Information Sciences Institute are building innovative data models to characterize COVID-19 disease progression. Deep convolutional neural networks will be applied to x-ray and CT image analysis, as well as prediction of mortality and ventilator needs, with the goal of building diagnostic tools for the disease. ISI’s model mixing strategies have shown success in the past, particularly in environments with heterogeneous data and local capabilities. A recent analysis of blood serum linked to two viral infections identified a set of latent factors that are correlated with different outcomes. 

GOALS: The models aim to predict clinical outcomes of COVID-19 patients using machine learning models that learn in a locally distributed manner and can handle longitudinal multivariate data with missing values. 

CURRENT STATUS: The machine learning teams have tailored these methods to relevant COVID-19 datasets, starting with USC's Keck Medical Center with plans to expand to collaborating institutions. DARPA has expressed an interest in using these techniques for their own COVID-related efforts.

Need informatics help?

Submit a request

Questions about the initiative

Daniella Garofalo
Program Manager, Clinical Research Informatics

NIH Funding Acknowledgment: Important - All publications resulting from the utilization of SC CTSI resources are required to credit the SC CTSI grant by including the NIH funding acknowledgment and must comply with the NIH Public Access Policy.