From improving patient outcomes, reducing complications and decreasing healthcare spending, the potential of artificial intelligence and machine learning in spine surgery is vast.
Ten spine surgeons discuss how the technology will affect the future of the specialty:
Ask Spine Surgeons is a weekly series of questions posed to spine surgeons around the country about clinical, business and policy issues affecting spine care. Becker's invites all spine surgeon and specialist responses.
Next week's question: What’s one improvement you would like to see for the next generation of spine robots?
Please send responses to Alan Condon at firstname.lastname@example.org by 5 p.m. CDT Wednesday, Nov. 30.
Editor's note: Responses were lightly edited for clarity and length.
Question: What will artificial intelligence and machine learning look like in spine surgery in 10 years?
James Dowdell, MD. Hospital for Special Surgery (New York City): AI and machine learning will drive better outcomes in spine surgery within the next 10 years. The most difficult aspect of spine surgery is deciding on the correct operation for the correct patient. There are many variables at play that determine a successful surgical outcome for a patient. AI/machine learning will use these variables to help support our surgical decision making. Reducing the variability in decision making will help drive better patient outcomes, reduce complications, and reduce healthcare spending.
Jun Kim, MD. Mount Sinai (New York City): We are still in the very nascent stages of applying deep learning to medicine and surgery. In 10 years, I believe that patient care from every phase of care will be improved by machine learning. Spine surgery will see multiple advances in the application of machine learning and deep learning for the purposes of preoperative planning, intraoperative execution and post-surgical care and prognostication. The availability and use of patient data will lead to algorithms that can make spine surgery more predictable, efficient and ultimately safer.
Philip Schneider, MD. The Centers for Advanced Orthopaedics (Bethesda, Md.): AI and machine learning should ideally improve spine surgery practices and options to provide safer surgeries and yield better results. The hope is that AI and machine learning will continue to develop in the next decade to predict surgical outcomes and success rates, provide insight into potential complications and readmission, and create clinical decision support tools. In the next 10 years, we will be able to predict and provide the best surgery fit for a specific patient. Currently, there are several surgery options available for a patient. In 10 years, the technology will continue to evolve, so we can offer the best course of action for each patient.
Nick Jain, MD. DISC Sports & Spine Center (Newport Beach, Calif.): I believe the true value of machine learning and AI in spine surgery will be with improving surgical indications to optimize patient outcomes. Using large-scale databases, machine learning can help determine specific MRI traits and values in combination with patient-specific factors to determine specific and detailed cohorts of patients. We can then isolate cohorts that will have the best outcomes based on thousands of individual data points per patient.
Vik Mehta, MD. Hoag Hospital (Newport Beach, Calif.): The large amount of data being collected from navigation and robotic platforms will allow us to develop predictive models which describe how varying treatments compare from a risk and benefit perspective. In 10 years, we will be able to integrate data from a variety of sources including wearable technologies, imaging and functional assessments to develop highly predictive models which will allow surgeons to counsel patients with better quality data.
Brian Gantwerker, MD. The Craniospinal Center of Los Angeles: The inception of AI and machine learning in spine surgery is already here. There are large databases being used as sources for data mining. Although many of those in this space insist it will be better for patients and allow us to make predictions on outcome (both great things), it will be only a matter of time until those databases are sold — at a very healthy profit — to insurers. Then they will in turn cherry pick the guidance from the algorithms generated to deter surgeons from operating and practice medicine with no accountability. We will need to empower our patients to advocate for themselves and to let surgeons use the tools to help patients, and not to keep money in the insurers' pockets.
Alexander Butler, MD. Lenox Hill Hospital (New York City): Predictive analytics has until now been mostly applied within the spine deformity space. With regards to pre-operative clinical decision making, large databases have been created and leveraged to allow for machine learning algorithms to assist with risk stratification and forecasting results. The "who, what, where, when and how" of planning interventions can be informed by these models to maximize outcomes for each patient and for the system. Less has been done to explore how predictive analytics can help with technical decision making in the MIS arena. As much as most spine surgeons enjoy debating the subtle nuance of indications and surgical methods for routine cases, having more objective guidance regarding predicted structural and biologic outcomes would certainly be helpful for all.
Christian Zimmerman, MD. St. Alphonsus Medical Group and SAHS Neuroscience Institute (Boise, Idaho): Current and future applications of AI and machine learning to the complexities of spinal surgery include data analytics and clinical decision support systems. The implications of AI for surgeons and the role of practitioners in advancing this technology (to optimize clinical effectiveness) will be forthcoming, ongoing and eventually standardized. Machine learning is particularly useful for identifying subtle patterns in large datasets and its global application will grow in its utilization.
Machine learning allows a computer to utilize partial labeling of data sets detected (within itself) to explain and assist surgeons in predictions about the data outcomes without explicit programming. Supervised learning is useful for training a machine learning algorithm to predict a known result or outcome while unsupervised learning is useful in searching for patterns of efficiencies. ML has currently outperformed in the clinical setting certain statistical analysis like logistic regression for prediction of surgical site infections. Its utility will only continue to assist surgeons, especially in their care of high acuity patients.
Brian Fiani, DO. Weill Cornell Medicine/NewYork-Presbyterian Hospital (New York City): AI has tremendous potential to revolutionize spine care. In the next 10 years, predictive analytics using algorithmic decision support tools will be used for the preoperative patient selection process. Automated radiographic detection software is increasingly prevalent and will be on the rise in the next several years. Meanwhile, intraoperative benefits from artificial intelligence and machine learning include robotic-assisted spine surgery. Robotic assistance, much like surgical augmented reality glasses will be more common in operating rooms in 10 years.
Harel Deutsch, MD. Midwest Orthopaedics at Rush (Chicago): AI may be involved in future medical decision making, especially on the corporate (insurance) side. I also think that areas such as radiology will increasingly have AI involvement.