A new Frontiers in Science Lead Article from Granados, Khanna, Fischer, and colleagues at King’s College London — with Prokar Dasgupta as senior author — takes a swing at one of the harder questions in modern surgery. When the operating room becomes a sensorized, situationally aware ecosystem with embodied AI agents, robotic scrub assistants, and real-time causal inference engines… who, exactly, is in charge?
Their answer, delightfully, is that the question itself is the wrong frame. The OR isn’t getting replaced. It’s getting redistributed.
The scale of the problem
Roughly 30% of the global disease burden touches surgical care. There is a gap of about 143 million unmet surgical procedures every year, and a projected 52% increase in cancer surgery demand between 2018 and 2040. Surgical complications are the third most common cause of death worldwide. We are not going to scale our way out of this with more humans alone.
The redistribution thesis
Surgeons shift from instrument operators to supervisors, coordinators, and high-level decision-makers. Scrub nurses oversee robotic scrub assistants and workflow integration. Circulating nurses coordinate fleets of autonomous logistics robots. And — this is the part I love — the team gains entirely new members: clinical data scientists, AI integration engineers, robotic integration engineers. The OR becomes a cyber-physical ecosystem.
Surgical co-pilots, not just chatbots
Granados and colleagues lay out a vision built on vision-language models (VLMs) and vision-language-action (VLA) models that translate natural language instructions into coordinated robotic movements. Causal AI that can ask what if instead of just what’s correlated. Real-time anticipation of the next step rather than retrospective annotation. Federated learning that preserves privacy while building generalizable models across institutions. Their Table 2 is a generous tour of the current state of the art across surgical data science.
Ethics is not an afterthought
The authors spend serious time on liability when authority chains are diluted, bias when training data is skewed toward resource-rich nations, and the regulatory mess that comes from adaptive AI systems that change after approval. Their Table 3 alone is worth the read for anyone trying to think clearly about how these systems should be governed. Predicate creep, post-marketing surveillance, evidence generation in adaptive products — all of it spelled out in plain language.
The bit I keep coming back to
Democratization of information. When AI surfaces what is happening in real time to the entire team — not just to the surgeon at the console — the implicit hierarchy of the OR shifts. Speaking up gets easier when you have data to point at. That is a profoundly good thing for patient safety. It is also a profoundly disruptive thing for how surgeons have been trained to lead.
Why this matters for limb preservation
This isn’t abstract for those of us in the toe-and-flow world. Limb preservation already lives or dies on coordination across vascular surgery, podiatric surgery, infectious diseases, endocrinology, and rehabilitation — multiple decision-makers, multiple data streams, and time pressure that doesn’t forgive miscommunication. An AI co-pilot that anticipates the next step, surfaces causal inference across patient-specific data, and democratizes information across the team is exactly the kind of tool that could move the needle on outcomes that have stubbornly resisted incremental progress.
This is, in the best sense, weird ideas put together with weird people. Surgeons, AI engineers, ethicists, and regulators arguing in the same paper that the OR should sustain practice rather than disrupt it. Worth a read.
#AI #Robotics #Surgery #LimbPreservation #DiabeticFoot #SurgicalDataScience #FrontiersInScience #DFCon #SALSA #ALPSlimb #EmbodiedAI #VLM #CausalAI
Leave a Reply