AI healthcare

Artificial intelligence in Orthopedics – What’s it going to take?

by Danny Goel, MD Orthopedic Surgeon, Clinical Professor and CEO – PrecisionOS

There is much interest and excitement about #artificialintelliegence, but I see several hurdles in orthopedics we must overcome. I’ll be interested in the comments from others and how we can implement AI where it’s needed the most.

What is AI

Artificial intelligence (AI) is a very complex and exciting data driven initiative. In the simplest form, it is a sophisticated output to an input, that gets smarter over time with training and testing. A basic example suggests that if we all agree what a dog is (and looks like),exposing a trained AI algorithm a picture of a cat, will reveal that this is in fact, not a dog. This, you would agree is a very basic example. It should, however, in its most advanced form, be able to provide a human level of engagement from less data points without any further intervention, ie showing a dog’s tail and be able to identify that It is a dog.

As with anything, there are pre-requisites to make an AI algorithm successful. It requires meaningful but accurate data.  Those data points, at its core, have to be agreed upon, and verified, by experts to be the true likeness of X.  Surgery, however, may be somewhat challenging as it relies on a combination of both art and science. There is a process, in identifying and treating a particular problem, especially in surgery with one caveat, the human intervention during the procedure.  Further challenging this model is the level of healthy disagreement we have in certain areas. For example, not only is their lack of consensus on how we do (or interpret) a physical exam, but correlating that with the imaging (xray, CT or MRI etc), define the treatment, carry out the procedure and then evaluate the outcomes based on that treatment. This confirmation requires a full circle of agreement to validate, train and test the algorithm where importantly, we need standardization. But what happens when the treatment can also vary for the same problem across geographies? We have all been exposed to debates in every area in orthopedics. This means that neither the data inputs nor the treatments are black and white. There isn’t just disagreement, but fierce debates on how best to manage the same problem in the same patient by two surgeons. This impacts the output, which may or may not have meaning to particular population of patients or surgeons.

The Challenges

The lack of consistency on the “how” and “what” in surgery may pose a intermediate challenge.  Also, how will this work when there is a considerable human element that is also essential to surgical outcomes?

The other major barrier to artificial intelligence is data collection. In surgery, data, both on initial visit and in follow up are necessary prerequisites.  Currently we have spectrum of electronic medical health records and patient related outcome scores which provide a lot of  information. This is inconsistent across institutions, nationally and also internationally. What and how do we collect it, ensure we have enough and how best to gather this information is a major hurdle.

My Observations

In my simple observation, successful AI in today’s world will have to live in the background. It has to collect data that we all agree to be true, real and verified. It supports our world while providing meaningful information as an output that we trust, and can accept with confidence. We have much work to do with respect to data collection in healthcare before we can receive meaningful answers and create a language model that will align with our expectations, and those of our patients.

Advances in AI are not insignificant, and it is being used by some in industries with great success. But, a consideration in surgery is how do we connect the dots in the near term with such variability. We need alignment across a changing ecosystem for a particular disease, agree on the treatment and verify the expected outcomes all while considering the variation and experience of the provider. In the absence of that, if we show a poorly constructed algorithm a lion, it may very well think that it’s a zebra or something worse.

Dr. Danny Goel is a practicing Orthopedic Surgeon, Clinical Professor and CEO of PrecisionOS Technology

About PrecisionOS     

PrecisionOS is a leading provider of virtual reality surgical training for the medical industry. Healthcare professionals across numerous societies, universities, and medical device companies prefer PrecisionOS modules because they improve the transfer of knowledge and skills. Multiple, independent published trials confirm that participants using the PrecisionOS platform become better, more confident surgeons. PrecisionOS has collaborative affiliations with more than 65 major global residency programs where the platform is being used in more than 55 countries globally. Learn more at     

Media Contact 

Roberto Oliveira