The Stanford Program for AI Health Care

Where is the intersection of man and machine—health care and artificial intelligence? Artificial intelligence (AI), machine learning, and data-driven technologies are becoming part of medicine. Yet we believe that medicine must remain fundamentally an endeavor of humans caring for other humans. Stanford Health Care established the Program for AI Health Care to look at these issues and ways to improve people’s lives with AI. Our mission is to bring machine learning and AI to the clinic safely and responsibly.

It is fitting that major advances in the applications of artificial intelligence to precision health care are taking place at Stanford Health Care. We were one of the original sites where the use of AI in medicine began in the 1980s.

Today, four core activities help us fulfill our mission:

TRAINING

To train people to plan and implement AI health care projects, we run the “Boot Camp for AI in Health Care” in partnership with the Department of Computer Science. Students take a deep dive into cutting-edge research in radiology, pathology, electronic health records, mental health, and public health. We also train clinicians, particularly Clinical Informatics fellows, on our Clinical Informatics Consult Service.

DEPLOYMENT

We set up the necessary infrastructure and procedures, so we can act on predictions and gain actionable insights on data. As an example, we have a petabyte-scale searchable repository of annotated, de-identified medical images, linked to genomic and electronic medical record information for use in the creation of AI-related systems. We’ve also built a HIPAA-compliant platform to support research applications built on smartphones.

COLLABORATION

We work with leading technology companies to help expedite our discoveries and their deployment. For example, we are using the power, security, and scale of Google Cloud to support precision health and more efficient patient care. We also hold symposia surrounding humans and machines in medicine, and an annual research meeting on machine learning that connects computer scientists who have expertise in AI and machine learning with clinicians and medical researchers.

STUDY

Before writing the first line of code for an AI algorithm, we study the system and ask several questions, such as: If you had a prediction, what action would you take, how often, on how many people, and what would it cost? Based on the answers, we figure out the minimum performance measures a model must have to be useful. This design analysis upfront is done before building any potential solution. Currently, AI project teams are looking at improving palliative care with deep learning, making predictions for length of stay, and identifying patients who may have undiagnosed genetic disorders.