September 16, 2020
3 min read
SooHoo NF, et al. Can artificial intelligence provide value in orthopedic surgery? Presented at: American Orthopaedic Foot & Ankle Society Annual Meeting; Sept. 10-12, 2020 (virtual meeting).
SooHoo reports he is on the medical/orthopedic publications editorial or governing board for SLACK Incorporated and Wolters Kluwer Health – Lippincott Williams & Wilkins; and is board or committee member for the American Orthopaedic Foot & Ankle Society and the American Academy of Orthopaedic Surgeons.
Artificial intelligence and machine learning could aid in the advancement and application of knowledge in foot and ankle surgery, according to a presenter at the at the American Orthopaedic Foot & Ankle Society Annual Meeting.
“The term artificial intelligence, I think it has caught a lot of currency in the last few years,” Nelson F. SooHoo, MD, professor and residency program director for the department of orthopedic surgery at the University of California, Los Angeles, said in his virtual presentation.
“When we talk about AI, I am not talking about a sentient doctor. I think most of you realize that,” SooHoo said. “We are talking about the more immediate and practical application of what we call ‘neural networks’ to better answer questions about how we figure out how our patients are going to do, how we predict outcomes, [and] how we improve and drive those improvements in quality and value of care.”
Nelson F. SooHoo
SooHoo explained that even though neural networks are “not anything new” and have “been around for decades,” recent interest in AI can be attributed to increased data collection initiatives and computing power.
“We have graphical processing units that are ideally suited to the implementation of neural networks and other types of machine-learning algorithms,” he said. “We have access to this on the cloud, and we can rent computer time and compute power in order to look at our problems.”
While SooHoo said diligent and dedicated analysis of patient outcomes is “intuitively valuable” and may be the foundation of the advancement of knowledge in foot and ankle surgery, AI has the potential to “move us forward.”
“Of course, there has always been the issue of levels of evidence, and we understand the weakness of clinical research, but it is difficult particularly in foot and ankle where we have a lot of heterogeneity in the things we are doing and the patients we are seeing,” SooHoo said. “It is difficult to do anything other than the careful, meticulous reporting on our patients. I spent a long portion of my career trying to get to this question from another way, and a lot of people have as well looking at administrative databases, looking at registries, trying to get bigger samples to remove some of the bias that we see in individual studies,” he said.
“This is where the power of artificial intelligence, I think, is going to drive our research forward in the next decade or 2,” SooHoo said. “Now, we have this big data. We have the individual charts that we put in there. We have our radiographic readings that we look at, and now we can do something meaningful with all the rich clinical information that we usually do on individual studies, but at a scale of a registry.”
SooHoo said physicians can now use natural language processing to read notes, identify clinical severity from radiographic data, and evaluate patient symptoms and interactions.
“I think all of us are used to studies where, of course, we have classification systems. We measure angles on the radiographs. We use those as prognostic factors. [It is] labor intensive and, again, limiting the scope and the scale of what we can do,” he said. “Well, now we can do some of those things automatically. We can use image-processing algorithms based on recurrent neural networks and pull the information out and find new information that we never even considered in radiographs,” SooHoo said.
Actionable intelligence is what physicians ultimately want, according to SooHoo.
“How is my patient going to do? What can I do to decrease my patient’s risks? How is it applied to this patient that I am facing in front of me?” he said. “We are seeing predictions at levels beyond anything that we have ever seen before in the published literature. So, we are seeing that the machine learning is identifying variables and relationships that allow us to predict outcomes better than we ever have before, and this is using the problematic datasets of just the administrative database.”
“I think that the power of this AI will be the combined detail and the rigor and the care of clinical research that we have always given as surgeons with the scale of the registry,” SooHoo concluded. “And I am looking forward to that,” he said.