113. The Secret Weapon In The Fight Against Rare Diseases

What this episode covers

Hosts Daniel and Forbode discuss a Stanford Human Centered Artificial Intelligence Group study that uses an AI model called PopDX to detect rare diseases in the UK Biobank despite incomplete patient data. The hosts explain how the algorithm infers missing diagnosis codes with high precision, improving characterization of unseen diseases by 218% and rare diseases by 151% compared to existing algorithms. They also highlight the researchers' acknowledgment of demographic biases in the training data and the potential for this technology to accelerate medical research and treatment for rare conditions.