AI in Dermatology - The Pitfalls and Promises

Explore the potential and challenges of artificial intelligence in dermatology through this insightful lecture by Dr. Roxana Daneshjou from Stanford University. Delve into the promising applications of AI in diagnosing skin diseases, while critically examining the pitfalls such as biased datasets and algorithms. Learn about the importance of developing equitable AI systems to prevent exacerbating existing health disparities. Discover how AI can streamline healthcare processes in dermatology when fairness is prioritized. Gain valuable insights into diverse dataset creation, fair algorithm development, and their applications in precision medicine. Engage with topics including bias in data collection, labeling challenges, skin cancer and skin tone classification, and the need for diverse dermatology images. Understand the significance of data accuracy, consensus labeling, and finetuning in AI models. Examine the impact of dataset biases, FDA regulations, and the differences between prospective and retrospective data in medical AI development.

Introduction
Why I became interested in MedAI
Problems with MedAI
Bias Data
Datasets
Labelling
Skin cancer classification
Skin tone classification
Diverse Dermatology Images
Data Accuracy
Consensus Labeling
Finetuning
Data biases
Dataset
FDA
Prospect vs retrospective data
Bias in dermatology education
AI in dermatology
Questions