The last mile: deriving practical clinical insights from electronic medical dictations
Brown Bag Lecture by Ernest Sohn and Serge Blok, Ph.D. | 11/21/2017 11:00AM – 12PM | 7th Floor Conference Room, Bldg 38A
Electronic Health Records (EHRs) provide an immense opportunity for improving patient outcomes, reducing costs, improving research. Information in these records can improve disease models, discover new treatments, identify safety concerns, as well as craft better treatment plans grounded in the patient’s personal health history. The majority of clinical information today is recorded by medical staff in free-text form which facilitates communication between medical professionals, but are difficult for a non-expert. Approaches to information extraction have been highly domain-specific, rather than a domain-general NLP solution. A customizable pipeline approach that combines NLP with machine learning allows the information processing steps to be encapsulated in configurable modules that can be added or modified.
Ernest Sohn is a chief technologist at Booz Allen with more than 10 years of experience developing data science and analytics solutions —including predictive analytics, machine learning, natural language processing, operations research, and data visualization—across Government and Commercial health organizations.
Dr. Blok is a statistician and data scientist with experience with statistical analysis, computational modeling, machine learning and text mining. He has successfully used these methods to improve the quality of decisions made in government, nonprofits and industry. His recent clients include the Consumer Financial Protection Bureau, FDA, VA, as well as nonprofit hospitals.