Machine Learning (ML) and Surgery 4.0
Recent advancements in ML for medicine have enabled physicians to leverage predictive tools for disease diagnosis (such as lung nodule detection ; and diabetic retinopathy screening ) and early intervention. However, most ML advancements have strayed away from interventional care, such as surgery, because insufficient data exists to benchmark existing and gold standards of care .
This gap is alarming, as medical error is the third leading cause of death in the US– upwards of 440,000 deaths occur annually , of which 17% are preventable (operator-induced) complications that occur during surgery .
Enter minimally invasive robotic surgery (MIRS), which promises to improve surgeon performance and patient safety — by enabling better anatomical access, instrument control and visualization through miniature incisions within the body. Robotics represents Surgery 3.0 (Figure 1) – a leapfrog technology that promises safer, higher quality-of-life post-operative outcomes compared to open surgery or laparoscopy . However, MIRS is still competency/skill dependent and complex to learn . Hence, the broader issue of surgeon-to-surgeon variability and its implications on patient outcome persist.
What if there was a way to “perfect” surgery? Deal with adverse events? Deliver care in a safe, consistent (operator-agnostic) and patient-specific manner? Enter Surgery 4.0 – a data-driven approach to operative performance and decision making (Figure 2).
Verb Surgical and the era of Digital Surgery
Verb Surgical aims to deliver surgery 4.0 to the 5 billion people  worldwide who desperately need care. To push this new mode of surgery into the operating room, Verb aims to develop a digital surgery ecosystem (DSE) around end-to-end patient care (Figure 3).
The biggest barrier to surgery 4.0 adoption is data and annotation. By developing a comprehensive data aggregation platform around DSE, Verb can start to derive contextual information on how a surgeon operates, who they’re operating on, and technologies that can predict and prevent intraoperative adversities. How can Verb collect this data en masse? How can they annotate this data to benchmark gold standards of care, so that ML models can reliably predict and sustain surgical quality metrics intraoperatively?
Industry and Clinical Partners for Data and Annotation
The answer lies in partnerships. Verb Surgical is a strategic partnership between Google and Johnson and Johnson (J&J). Google provides annotation frameworks that can be leveraged to train predictive models , while J&J provides the distribution channels within various hospital networks to acquire data at scale. However, no hospital or physician will adopt a system that does not provide clinical value today.
To address this issue, Verb Surgical is targeting procedures that enable surgeons to visualize anatomy in novel ways, and robotically operate on these structures if they so desire . In many ways, they are benchmarking to Surgery 3.0, while delivering incremental DSE benefits to drive data-driven decision making for the surgeon – before, during and after a procedure . Surgeons have been receptive to this strategy , and to Verb’s benefit, represent an early set of clinical adopters that will push data and annotation aggregation for broader surgery 4.0 adoption.
Verb’s biggest challenge to scale beyond early adopters is to overcome the heavy switching costs associated with acquiring DaVinci surgical users in the US (a robotic system manufactured by Intuitive Surgical®, with 88.8% market penetration ) and to sell their system to price-sensitive international users at a low capital cost . The latter customers are currently inaccessible to Intuitive Surgical® (each system costs upwards of $2M USD ) but are high-value data opportunities for Verb (by surgical volume, these markets are 7x larger than the US ).
A Netflix Model for Surgery 4.0
How can Verb capture price-sensitive, high volume users quickly? Netflix may provide a clue. Hospitals typically take months to perform a cost-benefit-analysis before making a large capital purchase decision. What if hospitals got the robot for free, and paid an annual subscription after the first year of use? This enables purchasers to develop an economic value proposition unique to their hospital system, with no capital downside. This is an emerging model garnering broader adoption across other device companies , and to Verb, presents a unique long-term monetization opportunity synergistic with surgery 4.0 (Data Products, Figure 3).
2030 and Beyond – Collaborative Autonomy
Fast forward a decade from now. Verb has every high-volume surgical market annotated, constantly refining predictive models that know how to execute the perfect procedure for any patient context. Surgeons in rural China, with the help of intelligent guidance systems (Figure 4), can perform nephrectomies (removal of cancerous kidney) as well as their academic counterparts abroad.
Is this paradigm of care attainable? Can sensors be placed within the body to characterize instrument and anatomic interactions with sufficient accuracy to warrant robust predictive power from a machine? (800 Words)
 Nam, Ju Gang, Sunggyun Park, Eui Jin Hwang, Jong Hyuk Lee, Kwang-Nam Jin, Kun Young Lim, and Thienkai Huy Vu et al. “Development And Validation Of Deep Learning–Based Automatic Detection Algorithm For Malignant Pulmonary Nodules On Chest Radiographs”. Radiology, 2018, 180237. doi:10.1148/radiol.2018180237.
 Gulshan, Varun et al. 2016. “Development and Validation of a Deep Learning Algorithm for Detection of Diabetic Retinopathy in Retinal Fundus Photographs”. JAMA 316 (22): 2402. American Medical Association (AMA). doi:10.1001/jama.2016.17216.
 Fecso, Andras B. et al. 2017. “The Effect of Technical Performance on Patient Outcomes in Surgery”. Annals of Surgery 265 (3): 492-501. Ovid Technologies (Wolters Kluwer Health). doi:10.1097/sla.0000000000001959.
 Makary, Martin A & Daniel, Michael. 2016. “Medical error—the third leading cause of death in the US”. BMJ: i2139. BMJ. doi:10.1136/bmj.i2139.
 La Pietra, L. 2005. “Medical errors and clinical risk management: state of the art.”. Acta Otorhinolaryngol Ital. 25 (6): 339-346. PubMed. https://www.ncbi.nlm.nih.gov/pubmed/16749601.
 Roh, Hyunsuk Frank; Nam, Seung Hyuk & Kim, Jung Mogg. 2018. “Robot-assisted laparoscopic surgery versus conventional laparoscopic surgery in randomized controlled trials: A systematic review and meta-analysis”. PLOS ONE 13 (1): e0191628. Public Library of Science (PLoS). doi:10.1371/journal.pone.0191628.
 Mazzon, Giorgio et al. 2017. “Learning Curves for Robotic Surgery: a Review of the Recent Literature”. Current Urology Reports 18 (11). Springer Nature. doi:10.1007/s11934-017-0738-z.
 Huennekens, Scott. 2018. “Verb Surgical”. Twitter.com. https://twitter.com/verbsurgical/status/1008674056546242561.
 Vedula, S. Swaroop & Hager, Gregory D. 2017. “Surgical data science: the new knowledge domain”. Innovative Surgical Sciences 2 (3): 109-121. Walter de Gruyter GmbH. doi:10.1515/iss-2017-0004.
 Webb, Marionn. 2018. “Market Intel: Rivals Catching Up To Intuitive Surgical In Fast-Growing, Fast-Innovating Robotic-Assisted Devices Market”. MedTech Insight | Pharma Intelligence. https://medtech-pharmaintelligence-informa-com.ezp-prod1.hul.harvard.edu/MT121751/Market-Intel-Rivals-Catching-Up-To-Intuitive-Surgical-In-FastGrowing-FastInnovating-RoboticAssisted.
 Udwadia, TehemtonE. 2015. “Robotic surgery is ready for prime time in India: Against the motion”. Journal of Minimal Access Surgery 11 (1): 5. Medknow. doi:10.4103/0972-9941.147655.
 Ho, Chuong Thanh. 2011. Robot-assisted surgery compared with open surgery and laparoscopic surgery. Ottawa, Ont.: Canadian Agency for Drugs and Technologies in Health = Agence canadienne des médicaments et des technologies de la santé.
 Rose, John et al. 2015. “Estimated need for surgery worldwide based on prevalence of diseases: a modelling strategy for the WHO Global Health Estimate”. The Lancet Global Health 3: S13-S20. Elsevier BV. doi:10.1016/s2214-109x(15)70087-2.
 Chandrasekaran, Brahadeesh. 2018. “Transformation Of The Medical Device Industry”. Frost And Sullivan. https://ww2.frost.com/frost-perspectives/transformation-medical-device-industry/https://ww2.frost.com/frost-perspectives/transformation-medical-device-industry/.
 Alkire, Blake C, Nakul P Raykar, Mark G Shrime, Thomas G Weiser, Stephen W Bickler, John A Rose, and Cameron T Nutt et al. 2015. “Global Access To Surgical Care: A Modelling Study”. The Lancet Global Health 3 (6): e316-e323. doi:10.1016/s2214-109x(15)70115-4.