RAS-ACS Communications Committee Essay Contest: Artificial intelligence in surgery: A call to action

Editor’s note: Each year, the Communications Committee of the Resident and Associate Society of the American College of Surgeons offers an essay contest centered on a theme that the committee has selected. The theme of the 2020 essay contest was Surgeon versus Machine: Evaluating the Role of Artificial Intelligence and Innovative Technology in Surgery. Following is the winning essay.

Artificial intelligence (AI) is changing surgery several decades after similar transformations occurred in the automotive and airline industries but with a similar sense of inevitability. There are biologic limits on human information processing, decision making, and dexterity; AI offers performance advantages in each of these domains.1-6

But surgeons need not be pitted against or replaced by machines. Instead, surgeons have opportunities to integrate AI applications into clinical workflows and steer them toward optimal patient care. At present, AI is enhancing predictive analytic decision-support, surgical skill assessment and coaching, and intraoperative care. At the systems level, machine-learning models can predict trauma patients’ acuity and distribution across trauma centers and make recommendations for cancer treatments that match recommendations from multidisciplinary tumor boards.7-9

AI’s advantages

Preoperatively, machine-learning algorithms can use livestreaming electronic health record data to predict postoperative complications, achieving greater accuracy than clinicians.10-12 In surgical skill assessment and coaching, combinations of intraoperative video, virtual reality, computer vision, and machine-learning algorithms can classify surgeon skill and support perioperative and postoperative decision making.13-16

Intraoperatively, machine-learning models can use waveform data to predict hypotension and prompt anesthesiologists to act sooner, more often, and differently, resulting in fewer hypotensive episodes and less time-weighted hypotension.17 Autonomous robotic platforms can perform end-to-end sutured bowel anastomoses with leak pressures that are significantly greater than leak pressures for anastomoses sewn by surgeons.5

These technological achievements seem pedestrian relative to potential, future applications for AI in surgery. In the future, AI likely will provide clinical decision support with levels of accuracy and precision that have been unattainable with previous technologies and may change the way some operations are performed.

Reinforcement learning mimics human learning by using trial-and-error simulations to identify discrete actions that yield the greatest probability of achieving an ultimate goal. Applied to vasopressor infusions and intravenous fluid boluses for patients with sepsis, reinforcement learning can select resuscitation strategies that are associated with increased survival compared with standard care.18

Autonomous microrobots also are on the horizon. Researchers at the Massachusetts Institute of Technology, Cambridge, designed an origami-like robot that folds into an ingestible pill, unfolds in the body, and moves in response to external magnetic fields.19 In a silicone model of the human esophagus and stomach, the microrobot dislodged a battery embedded in the stomach wall and patched the stomach wall defect in approximately five minutes.

At present, no high-level evidence shows that AI improves patient outcomes compared with existing standards for performing operations or surgical decision-making tasks. Yet, history suggests that as technologies improve, AI eventually will achieve cost-effective performance advantages, and market forces will drive adoption by health care networks and hospitals, similar to what has occurred in the automotive and airline industries.

Surgeon leadership is needed

Rather than reacting to this possibility with denial, ire, or indifference, surgeons should engage in the development of AI technologies and lead the clinical implementation process. Surgeon leadership is critically important in navigating the pitfalls of machine-learning predictive analytics. If the unique pathophysiology of an individual patient is underrepresented in the data used for model training, then model predictions for that patient will err.

Surgeon leadership is critically important in navigating the pitfalls of machine-learning predictive analytics. 

Disturbingly, this phenomenon introduces the potential for model bias against underrepresented minorities. Prediction models that use biased datasets produce biased outputs, as previously demonstrated when an algorithm was used to predict crime recidivism.20 Even when large datasets are well-balanced, they often omit potentially useful information from patient interviews and physical examinations. For these reasons and others, high-performance prediction models still make errors.

But can models tell when they are wrong? There are methods for determining whether a model is confident that its predictions are accurate, but they tend to overestimate model confidence.21 Finally, decision-support tools often impose trade-offs among risks and benefits that may misalign with individual patient preferences.

Consider a postoperative patient with both pulmonary edema and prerenal azotemia. A reinforcement learning model trained to maximize 30-day survival may recommend diuresis to avoid prolonged mechanical ventilation, narrowly achieving a greater probability of short-term survival but imposing long-term requirements for renal replacement therapy and poor quality of life. What if the patient values quality of life more than a slight increase in 30-day survival? Each of these pitfalls can be remedied with human knowledge and intuition informed by rigorous training, a thorough bedside patient interview and physical examination, and careful interpretation of model outputs in a clinical context.

AI surgical platforms also face challenges in clinical implementation. Some robotic surgical platforms feature virtual constraints that are intended to protect anatomic structures from instruments.22 But what if a blood vessel that is beyond the virtual constraint is hemorrhaging from an avulsion injury generated by traction applied within the operative field? The virtual constraint could delay or prevent the surgeon from gaining control of the injured blood vessel, harming the patient and pitting human against machine in assigning liability.

Ensuring patient safety

Similar dilemmas in assigning liability may occur when microrobots err. More importantly, patients could be harmed. Therefore, establishing the safety and efficacy of AI applications in surgery is a critical first step. It seems prudent to perform prospective clinical implementation on a small scale under close surveillance and scrutiny, similar to Phase 1 and 2 clinical trials for experimental medications.23 As frontline providers of surgical care, surgeons should lead these trials, and collaborate with computer scientists, engineers, and commercial entities toward safe, effective clinical implementation of AI surgical platforms.

This essay is a call to action for surgeons to engage and lead in the clinical application of AI in surgery. AI clinical decision support, surgical skill assessment and coaching, and surgical platforms each offer performance advantages that could improve care for surgical patients. Experiences in other industries suggest that automation in surgery is inevitable. Surgeons are uniquely equipped with the knowledge, skills, and experience necessary to lead the safe, effective clinical adoption of AI in surgery, with the ultimate goal of providing the best care possible.


  1. Furuya S, Tominaga K, Miyazaki F, Altenmuller E. Losing dexterity: Patterns of impaired coordination of finger movements in musician’s dystonia. Sci Rep. 2015;5:13360.
  2. Carmeli E, Patish H, Coleman R. The aging hand. J Gerontol A Biol Sci Med Sci. 2003;58(2):146-152.
  3. Shanafelt TD, Balch CM, Bechamps G, et al. Burnout and medical errors among American surgeons. Ann Surg. 2010;251(6):995-1000.
  4. Wolf FM, Gruppen LD, Billi JE. Differential diagnosis and the competing-hypotheses heuristic. A practical approach to judgment under uncertainty and Bayesian probability. JAMA. 1985;253(19):2858-2862.
  5. Shademan A, Decker RS, Opfermann JD, Leonard S, Krieger A, Kim PC. Supervised autonomous robotic soft tissue surgery. Sci Transl Med. 2016;8(337):337ra364.
  6. Silver D, Huang A, Maddison CJ, et al. Mastering the game of Go with deep neural networks and tree search. Nature. 2016;529(7587):484-489.
  7. Christie SA, Hubbard AE, Callcut RA, et al. Machine learning without borders? An adaptable tool to optimize mortality prediction in diverse clinical settings. J Trauma Acute Care Surg. 2018;85(5):921-927.
  8. Somashekhar SP, Sepulveda MJ, Puglielli S, et al. Watson for Oncology and breast cancer treatment recommendations: Agreement with an expert multidisciplinary tumor board. Ann Oncol. 2018;29(2):418-423.
  9. Dennis BM, Stonko DP, Callcut RA, et al. Artificial neural networks can predict trauma volume and acuity regardless of center size and geography: A multicenter study. J Trauma Acute Care Surg. 2019;87(1):181-187.
  10. Bihorac A, Ozrazgat-Baslanti T, Ebadi A, et al. MySurgeryRisk: Development and validation of a machine-learning risk algorithm for major complications and death after surgery. Ann Surg. 2018;269(4):652-662.
  11. Brennan M, Puri S, Ozrazgat-Baslanti T, et al. Comparing clinical judgment with the MySurgeryRisk algorithm for preoperative risk assessment: A pilot usability study. Surgery. 2019;165(5):1035-1045.
  12. Bertsimas D, Dunn J, Velmahos GC, Kaafarani HMA. Surgical risk is not linear: Derivation and validation of a novel, user-friendly, and machine-learning-based Predictive OpTimal Trees in Emergency Surgery Risk (POTTER) Calculator. Ann Surg. 2018;268(4):574-583.
  13. Azari DP, Frasier LL, Quamme SRP, et al. Modeling surgical technical skill using expert assessment for automated computer rating. Ann Surg. 2019;269(3):574-581.
  14. Hashimoto DA, Rosman G, Witkowski ER, et al. Computer vision analysis of intraoperative video: Automated recognition of operative steps in laparoscopic sleeve gastrectomy. Ann Surg. 2019;270(3):414-421.
  15. Winkler-Schwartz A, Yilmaz R, Mirchi N, et al. Machine learning identification of surgical and operative factors associated with surgical expertise in virtual reality simulation. JAMA Netw Open. 2019;2(8):e198363.
  16. Hung AJ, Chen J, Gill IS. Automated performance metrics and machine learning algorithms to measure surgeon performance and anticipate clinical outcomes in robotic surgery. JAMA Surg. 2018;153(8):770-771.
  17. Wijnberge M, Geerts BF, Hol L, et al. Effect of a machine learning-derived early warning system for intraoperative hypotension vs standard care on depth and duration of intraoperative hypotension during elective noncardiac surgery: The HYPE randomized clinical trial. JAMA. February 17, 2020. Available at: https://jamanetwork.com/journals/jama/fullarticle/2761469. Accessed November 6, 2020.
  18. Komorowski M, Celi LA, Badawi O, Gordon AC, Faisal AA. The artificial intelligence clinician learns optimal treatment strategies for sepsis in intensive care. Nat Med. 2018;24(11):1716-1720.
  19. Miyashita S, Guitron S, Yoshida K, Shuguang L, Damian DD, Rus D. Ingestible, controllable, and degradable origami robot for patching stomach wounds. 2016 IEEE International Conference on Robotics and Automation (ICRA). May 16-21, 2016. Available at: https://ieeexplore.ieee.org/document/7487222. Accessed November 6, 2020. [Subscription required for viewing.]
  20. Angwin J, Larson J, Mattu S, Kirchner L. Machine bias. ProPublica. May 23, 2016. Available at: www.propublica.org/article/machine-bias-risk-assessments-in-criminal-sentencing. Accessed November 6, 2020.
  21. Gal Y. Uncertainty in deep learning. University of Cambridge. 2016. Available at: http://mlg.eng.cam.ac.uk/yarin/thesis/thesis.pdf. Accessed November 6, 2020.
  22. Cha J, Broch A, Mudge S, et al. Real-time, label-free, intraoperative visualization of peripheral nerves and micro-vasculatures using multimodal optical imaging techniques. Biomed Opt Express. 2018;9(3):1097-1110.
  23. Shortliffe EH, Sepulveda MJ. Clinical decision support in the era of artificial intelligence. JAMA. 2018;320(21):2199-220.

Tagged as: , , ,

Notify of
Inline Feedbacks
View all comments


Bulletin of the American College of Surgeons
633 N. Saint Clair St.
Chicago, IL 60611


Download the Bulletin App

Apple Store
Get it on Google Play
Amazon store