ASGE Annual Postgraduate Course: Clinical Challenges and Real-Time Decision Making in GI Endoscopy (Live and Virtual) | May 2023-AI in predictive modeling: Complex EHR Data

Video Transcription

Video Summary

In this video, TAPOCHE Roy, senior manager at Kaiser Permanente and author of "Notes on Medical Image Processing with Deep Learning," discusses the role of computer vision and endoscopy in analyzing big data sets and predictive modeling. He explains that electronic health records (EHR) are viewed differently by clinicians, IT professionals, and data scientists, with AI in EHR aiming to provide insights at both the individual and population levels. Roy emphasizes the complexity of healthcare data, referencing frameworks proposed by Dr. Thomas Kannapalli and Dr. Gell-Mann. He then discusses the life cycle of predictive modeling, including data cleaning, feature engineering, model development, testing, and deployment. Roy explains three paradigms in machine learning: unsupervised learning, supervised learning, and reinforcement learning. He provides explanations and examples of classification, regression, clustering, and reinforcement learning techniques in the context of healthcare. Roy concludes by highlighting opportunities for improvement in machine learning, including fairness, bias, transparency, explainability, and trustability. He also presents an EHR solution that integrates cohort management, lineage mapping, and visualization. The video provides various sources and encourages viewers to reach out for further questions.

Asset Subtitle

Taposh Roy

Keywords

computer vision

predictive modeling

electronic health records

unsupervised learning

reinforcement learning