Medical Physics Seminar - James Lamb

Automation and artificial intelligence (AI) are increasingly used in radiation oncology and show great promise to promote safety and increase workflow efficiency. Conversely, maladroit implementations may increase errors and slow down work. We discuss theoretical approaches to understanding and mitigating human error, from Reason’s theory of latent errors to cognitive biases in medical decision making. Practical approaches to implementing new automation-based safety systems in the clinic are described, with a focus on analysis of risk such as over-reliance and alert fatigue. We present machine-learning and AI-based systems being developed at UCLA for the mitigation of human errors in the planning and delivery of radiotherapy and for the risks of automation and AI, such as over-reliance and alert fatigue, and approaches to understanding the interaction of technology with human factors, from Reason’s theory of latent errors to cognitive biases in medical decision making. Practical approaches to implementing new automation-based safety systems in the clinic are described, with a central focus on an AI system being developed at UCLA to automate image-guidance review in order to make radiotherapy delivery safer and more resource efficient.

•       Theoretical approaches to understanding and mitigating human errors

•       Automating the image review function of radiation therapists

•       Practical automation in the clinic: promise, perils, and pitfalls