Deep Learning Applications in Electronic Health Records: Clinical Research Use Cases
The use of electronic health records (EHR) to identify specific clinical phenotypes has gained significant momentum over recent years. A good portion of the information within the EHR is trapped in free-text within clinical notes. A variety of natural language processing pipelines have been developed and used for information extraction. With the advent of vast computational power, significant strides have been made in deep learning approaches. Working with these models is far less daunting than it used to be, thanks to open source frameworks such as Google’s, TensorFlow. During this presentation, we will discuss the use of deep learning-based clinical text classifiers in a variety of pressing clinical scenarios, such as mental health and COVID-19, and compare the results to traditional classifiers using bag-of-words models. We will also examine the impact of using pre-trained language models and de-identification.
Dr. Obeid is the Co-director of the Biomedical Informatics Center and SmartState endowed chair in Biomedical Informatics at the Medical University of South Carolina. He is a pediatrician who was formally trained in Medical Informatics at the Division of Health Sciences and Technology, a joint Harvard-MIT fellowship program. At MUSC, he leads the effort on multiple translational research informatics infrastructure projects at MUSC, such as the electronic health records (EHR) Research Data Warehouse, REDCap, and many others. He is principal or co-investigator on several NIH funded projects. His research interests include secondary use of EHR data for research, electronic consents, and machine learning. His presentation will focus on the use of deep learning techniques to identify diseases using EHR clinical data and for predictive modeling.
Friday, October 30, 2020 at 2:30pm to 3:30pmVirtual Event