Preserving Patient Safety as AI Transforms Clinical Care - Curt Langlotz, Stanford University

Explore the intersection of artificial intelligence and healthcare in this 59-minute conference talk by Stanford University's Curt Langlotz. Delve into cutting-edge machine learning methods designed to enhance patient care and support healthcare professionals. Examine the unique data privacy and security challenges in the medical field. Investigate the crucial role of interpretability in improving the accuracy and safety of human-machine collaboration. Learn about clinical examples, the health AI industry, and intelligent devices. Understand the importance of data needs, narrative reports, and transformer models in healthcare AI. Discover knowledge graphs and their applications. Analyze case studies on bone age assessment and underdiagnosis bias. Explore ethical machine learning, generalizability, and algorithm performance. Gain insights into explainability techniques, including heatmaps. Investigate the impact of reading race in medical imaging and the differences between high-pass and low-pass filters in MR physics. Conclude with discussions on deep reinforcement learning, ethnicity detection, physics training, data collection methods, differential privacy, synthetic data generation, and data sharing in healthcare AI.

Intro
Clinical example
Health AI industry
Augmentation
Intelligent devices
Data needs
Narrative reports
Transformer models
How they work
Knowledge graphs
Results
Bone Age
Study
Underdiagnosis bias
Ethical machine learning
Generalizability
Reliability
Algorithm performance
Heatmaps
Explainability
Reading Race
High Pass vs Low Pass
MR Physics
Thank you
Questions
Deep reinforcement learning
Ethnicity detection
Physics training
Data collection
Differential privacy
Synthetic data
Data sharing