“A center of excellence is, by definition, a place where second class people may perform first class work.”

Michael Faraday, English scientist

Since its emergence in healthcare about a decade ago, artificial intelligence and its adoption in clinical medicine and healthcare has been accelerating in the past few years. Areas such as deep learning in the form of convolutional neural networks in medical image interpretation and machine learning in decision support, as well as natural language processing in healthcare, have matured and proven to be useful in both improving quality of care and increasing accuracy of diagnosis. Too often, however, projects actually deploying AI in clinical medicine and healthcare are relatively sequestered in a few areas, or with a few investigators.

In the past few years, there has been ongoing discussion amongst clinician and hospital leaders focused on the center of excellence concept, with its people, processes, and tools for the domain of artificial intelligence. The purpose of such a center is to convene people who have expertise or interest in this area to facilitate academic discussions and enable collaborative projects. I was the privileged convener of such a program in the form of an artificial intelligence institute called the Medical Intelligence and Innovation Institute, or MI3, at the Children’s Hospital of Orange County. Based on this near decade experience, I can share the advantages and disadvantages of such a center of artificial intelligence in healthcare.

A few of the advantages of such a center of excellence of artificial intelligence in clinical medicine and healthcare include the following:

  • Concentration of expertise

It is relatively difficult to recruit and maintain expertise of machine and deep learning, as well as other areas of AI such as natural language processing, so one single resource for this nascent domain in clinical medicine is very useful

  • Efficiency of manpower

As the pool of data scientists is relatively limited, especially those with experience in healthcare, efficiency for this human resource is of paramount importance, so a single data scientist can take on projects from multiple sectors in the health system

  • Wisdom of the crowd

Both clinicians and data scientists can gather and share, not only ideas for projects deploying artificial intelligence, but also results of these projects to minimize costly errors and maximize chance of impact on clinical care using artificial intelligence


There are concomitantly potential disadvantages or problem areas for such a center that focuses on artificial intelligence:

  • Lack of resources:

A center such as this, focusing on artificial intelligence, sometimes will have a lack of resources as there may be confusion about where resources can come from (research sector vs. information technology vs. separate source all together from the former two areas)

  • Conflict of interest:

A standalone center of artificial intelligence in the health system may directly or indirectly conflict with a few other domains in the hospital, such as research, information technology, or even biomedical informatics – depending on the ecosystem

  • Isolation of expertise

By having such a separate resource with experts who are not clinicians, the perception from clinicians can be that this knowledge is too abstruse for the average clinician and that learning these methodologies is beyond the reach of the clinicians.