What’s New in Surgical AI: 7/4 Edition
State of the A.I. in the Operating Room in 2023: Overview
Happy Independence Day, for all who celebrate! If you’re new to ctrl-alt-operate, we do the work of keeping up with AI, so you don’t have to. We’re grounded in our clinical-first context, so you can be a discerning consumer and developer. We’ll help you decide when you’re ready to bring A.I. into the clinic, hospital or O.R.
We’re refueled and recharged to focus on our mission of delivering you what you need to know about the state of surgical AI in 2023.
We can now announce registration is open for the Digital Neurosurgery 2023 Conference, in Palo Alto, CA from October 13-15, 2023.
www.digitalnsgy.com
Leaders in AI and surgery are joining forces to chart a course for the future of our field. Sign up today since our room blocks are going fast! If you’re a student or trainee and need assistance with registration or have other logistical questions, email info@digitalnsgy.com
Our deep dive continues from the prior week to define the landscape of AI-enabled tools in the operating room in 2023. These are actual companies shipping actual products that you might decide to use in your cases next week. We’ll give a brief overview of each market segment this week and then go into the specific players in the weeks to come.
Table of Contents
📰 The News: Wild, Wild West No More? Enter: Government
🤿 Deep Dive: State of the AI Landscape in Surgery
🪦 Best of Twitter: RIP
📰The News: Governments Getting Involved with AI
This week in AI news, we discuss a growing trend of governments and governmental agencies dipping their toes into the AI landscape. And unlike the usual oversight committee hearings (I will again, link the hilarity that was the TikTok CEO testimony before Congress), agencies are actually taking out their checkbooks and committing real resources to A.I.
The NHS announced a £21M initiative for improving healthcare efficiency using A.I. for diagnostic tooling. They aim to start with Chest XRs and stroke detection - well known AI use cases that have seen significant clinical adoption and validation.
…Is it just me or does 21M pounds seem like how much it would cost to implement any initiative at a handful of hospitals? Considering Inflection.AI just raised $1.3 BILLION, I am skeptical as to the scale this will enable technologic adoption.
Back on this side of the pond, The New York Department of Financial Services is exploring purchasing a supercomputer and hiring AI experts to run it, to better understand AI and use it to predict bad actors in the regulatory sector. Similarly, the FTC pre-announced a national rule making surrounding data security, specifically outlining how advanced computational techniques are used in highly regulated industries like healthcare.
Why does this affect you? With the clip that AI is moving, we can rest assured that the agencies that govern clinician’s lives, from the FDA to CMS, to even the AAMC / ACGME / American Academy of {insert specialty here} will soon weigh in AI, and start to regulate how and where their members can utilize it. Maybe this is okay? It’s becoming increasingly clear that chatGPT accounts are being compromised. We’re always trying to balance innovation with safety. I hope these guardrails don’t preclude the innovation we expect when it comes to these tools.
At the end of the day, we need to focus less on the mechanics of AI systems, and more about how they can actually be deployed. An example recently highlighted is Insilico, one of the many AI-powered drug development companies. They recently got FDA approval for an AI-discovered drug targeting idiopathic pulmonary fibrosis, and are in Phase II clinical trials having just dosed their first patient. They also recently announced a transformer-based platform which can assist in identifying factors associated with aging.
Many specialties, particularly surgical ones, suffer from the small data problem. Diseases are too rare to make hyper-specific inclusion/exclusion criteria for clinical trials, and resulting trial outcomes tend to have obfuscated results or come with a laundry list of caveats. The ability to generate new drugs, or even simulate trials with any sort of validity may allow us to get to larger scale trials faster, or allow us to design better trials knowing the next trial is only months, not years, away.
Finally, we’d like to remind everyone that if you’re reading this newsletter, you are so far ahead of the curve.
A new study by The Verge found that >40% of people haven’t even heard of chatGPT. Forget about the rest of the field - pretty much an afterthought. Not to mention physicians themselves are a rather tech-backward bunch. You are ahead of the curve, and you must continue to keep up with the tools to stay there.
Interested in building something in healthcare x AI? Need some help to get started? Shoot us an email with your idea / project and we’ll send you some resources to get you going.
🤿Deep Dive: State of the AI Landscape in Surgery (Part 1: Creating the Roadmap)
Last week, we gave you an overview of AI in the hospital and medical records. This week we turn our focus to our favorite place: the operating room. Rather than being forward-looking or visionary, as we often want to be, we’ll now focus on the goods that can be delivered today. This practical tour through commerically available solutions will be valuable to all of our readers who want to know more about the technologies they can deploy today.
Wait, What is AI in Surgery?
Artificial intelligence is generally defined as computational systems that have the capability to produce human-level performance across reasoning tasks. The subset of artificial intelligence we're most interested in is machine learning, which are systems that are capable of improving their performance without explicit instruction. Almost any piece of software that contains the ability to intake information and output information is rapidly becoming an artificial intelligence system. Even the cautery device goes to an electrical system with a circuit board - as soon as a company figures out how to monetize the “AI electrocautery”, it’ll be sold to us. (Side Bet: this will happen before Q1 2025) The real question will soon be: what isn’t AI in surgery?
Thus our mission: to define where we are today, looking at common threads of use cases and scenarios you might find yourself in. You’ll find that most of these companies rely on sensors and output devices, commerical platforms (hololens, Meta Quest, common robotic or navigation systems) and other similar elements. A major difference is the software and specific optimizations for defined use cases. Let’s dive in!
Here is the roadmap for our survey of the AI landscape in Surgery for July, 2023.
Each week, we will turn to one of the following areas of AI in surgery and review the current state of the art affecting the operating room today. Many of these examples come from our experience with neurosurgery and orthopedics, and we welcome insights from our audience about other commercially available AI tools. Initially note, we aren’t going to discuss AI-radiology or AI-pathology, even though those are two of the most common use cases in the Software as a medical device (SaMD) marketplace today because of our focus on perioperative care.
Training and Education (Examples: OssoVR, PrecisionOS, Figure1). Although most platforms focus on virtual environments and data sharing, we believe these organizations will be the focus of AI-generated assessments in the months and years to come. If they have anything to say about it, “See one, do 100 in the virtual” might become the new training paradigm. From an education standpoint, image and data-sharing platforms are poised to become important AI model training grounds even as they continue to train human surgeons.
Preoperative planning and decision support (Examples: Nuvasive iGA, Medtronic UNID / Medicrea, Atlas Pathfinder). Surgeons are looking for preoperative plan development for their most challenging cases. Difficult problems can have multiple, potentially correct solutions. Surgeons consider different approaches, such as the use of a larger or smaller incision, the specific trajectory to navigate to a target area, or the specific maneuvers to perform once the target is reached. Creating a plan before the first incision is made and predicting its downstream effects could reduce the likelihood of poor outcomes or treatment failures.
Pre-/Intraoperative Visualization Assistance (Examples: Augmedix, Novarad, ImmersiveTouch, SurgicalTheater, Medivis, OpticSurg, SentiAR, BrainLab, Medtronic Stealth, Zeta Surgical). During surgery, surgeons mentally integrate a variety of imaging modalities including radiographic images and the actual visual data collected during surgery to create an minds-eye view of the patient. Performing the equivalent of multiple 3-D mental rotation task while manipulating delicate and vital structures can lead to predictable human error. These organizations create patient-specific representations that can be displayed before or during surgery on novel or existing displays.
Intraoperative imaging (Examples: ActivSurgical, Proprio, Advanced Scanners, InveNIO, Medtronic GI Genius, Olympus x Odin Vision). Although preoperative imaging can be quite useful, what about all of the new data seen at surgery? These tools use the data from existing sensors such as endoscopes and laparoscopes or provide new sensors allowing the perception of the previously imperceptible. But we just said that surgeons might be troubled by the myriad of information already available to them, how could they handle more? You guessed it - these systems deploy computer vision and artificial intelligence methods to streamline and integrate novel sensor types into surgical workflows and displays.
Intraoperative Effectors aka Robots (Examples: Intuitive, Stryker Mako, Medtronic Hugo + Mazor, CMR Versius, Globus ExcelsiusGPS, Vicarious, ROSA, Moon Surgical, Asensus, Corindus(?spinning down) and many more). The most classic use case for surgical technology in the operating room are robotic effectors. Now, these aren’t robots in the classical sense, like the roomba you have cleaning your house. They don’t (generally) move without specific user direction, but they translate the surgeon’s movements into the movement of the arms. That statement is about to change with the introduction of semi-autonomous and supporting robotic tools, including the Moon Surgical assistant as well as the orthopedic surgery robots that plan and execute cuts during arthroplasty (joint replacement surgery). Many new technologies are coming to the market in this high-cost, high-growth area
Postoperative decision support: (Examples: none). This is an area ripe for exploitation by surgeons and surgical data scientists. Carrying forward intraoperative data into the immediate postoperative period to minimize adverse effects, reduce costs, and improve long-term patient outcomes represents a major opportunity. There’s a lot of exciting research being done in this area, and yet…
Postoperative analysis of performance (Theator, C-SATS, Hutom, Proximie, Caresyntax, Medtronic x Touch Surgical). This is an area near and dear to our hearts. Suffice it to say that this will be a major focus of our series, so we don’t want to spoil things too much :)
So there you have it, that’s our summer newsletter roadmap for surgical AI. Of course, new events will come up and certain segments may have to be put on hold, but we are very excited for what’s to come.
🪦 Best of Twitter
Any suggestions for what you’d like to see here instead?
Feeling inspired? Drop us a line and let us know what you liked.
Like all surgeons, we are always looking to get better. Send us your M&M style roastings or favorable Press-Gainey ratings by email at ctrl.alt.operate@gmail.com