A practical overview of privacy preserving deep learning in healthcare

Most healthcare data is scattered across many institutions, making it hard for researchers to access and use—especially given issues like incomplete, delayed, and insufficient demographic information. Instead of trying to aggregate all this data into a single, centralized source—which can be both impractical and risky—new approaches like split learning offer a secure way to analyze data where it already resides. Health Information Exchanges (HIEs), which hold comprehensive, localized patient data, become valuable resources when used with split learning. This method allows collaborative model training without sharing raw, sensitive data. As a result, researchers can study emerging health challenges—such as COVID-19—more effectively, tapping into rich, diverse data sources while protecting patient privacy.

In this talk, Sravan will discuss privacy preserving deep learning techniques, their uses and practical implications of using one. Sravan will lead the audience through understanding the current landscape, and delve deep into understanding split learning for AI development, its architecture and deployment scenarios in regards to the healthcare and HITRUST requirements.

Our guest speaker, Mr. Sravan Elineni (https://www.linkedin.com/in/esravan07kumar/), is a seasoned technologist with over 13 years of expertise in Healthcare Data Systems, Pharmaceutical GxP, Machine Learning, Robotics, Computer Vision, and Natural Language Processing (NLP). Throughout his career, he has consistently exceeded key performance indicators (KPIs) by innovatively designing, developing, and maintaining scalable software and hardware solutions.