Johns Hopkins teaches robot to perform a gallbladder removal

The future of surgery just took a bold leap forward in a lab at Johns Hopkins. For the first time, a robot has autonomously performed a critical chunk of a gallbladder removal on a hyper-realistic surgical dummy — no human hands guiding it. It wasn't just playing back pre-programmed motions. This machine was learning, adapting, and responding to the surgical team's spoken instructions as if it were a first-year resident under the eye of a seasoned attending.

"This advancement moves us from robots that can execute specific surgical tasks to robots that truly understand surgical procedures," said Axel Krieger, the lead brain behind the tech. That's not just a small upgrade — it's the kind of breakthrough that nudges science fiction a little closer to hospital corridors.

How did it work?

At the heart of this innovation is a machine-learning system called Surgical Robot Transformer-Hierarchy, or SRT-H. Think of it as the ChatGPT of the operating room — built on the same transformer architecture — but instead of generating text, it generates surgical moves.

Here's what makes SRT-H tick:

• It watches real surgical videos (in this case, gallbladder removals on pig cadavers).
• It learns from captions and visual cues, just like a human med student might annotate notes from a recorded lecture.
• It responds to voice commands during surgery, such as "grab the gallbladder head" or "move left a bit."
• It adapts in real time to changing anatomy and unexpected variations like blood-like dyes or shifted camera angles.

Unlike earlier systems like Krieger's STAR robot — which needed precisely marked tissues and followed a rigid playbook — SRT-H is built for the unpredictable. "It's like teaching a robot to navigate any road, in any condition," Krieger explained.

Why does it matter?

Let's not beat around the bush — this is a big deal. Previous surgical bots could handle tidy, bite-sized tasks in pristine lab settings. But real surgery? It's messy, variable, and packed with last-minute surprises.

Here's why SRT-H changes the game:

• It completed a full phase of gallbladder surgery — 17 coordinated tasks — from identifying ducts to cutting with surgical scissors.
• It did so across different anatomical conditions and still nailed 100% task accuracy.
• It learned modularly, like human trainees who gradually piece together complex procedures over time.

"Just as surgical residents often master different parts of an operation at different rates," said Johns Hopkins surgeon Jeff Jopling, "this work illustrates the promise of developing autonomous robotic systems in a similarly modular and progressive manner."

Sure, the robot took longer than a human surgeon. But it performed on par with experts. And speed, as any surgeon will tell you, comes after mastery.

The context

This isn't Johns Hopkins' first robotic rodeo. Back in 2022, Krieger's STAR bot did a laparoscopic surgery on a pig. But it needed a sanitized script and carefully prepped tissues. This time, there was no hand-holding. The robot had to think, adjust, and act independently.

The broader context? Surgical robotics has long hovered between tool and partner. Systems like da Vinci still rely entirely on human control. Full autonomy? That's the brass ring.

SRT-H suggests we might just be reaching for it now. "Our work shows that AI models can be made reliable enough for surgical autonomy — something that once felt far-off but is now demonstrably viable," said Ji Woong "Brian" Kim, now at Stanford.

The team isn't done. They want to train SRT-H for more surgeries and, eventually, full autonomous operations. It's still early days, but if these machines keep learning the way med students do — one procedure, one mistake, one success at a time — then the operating room of the future may look very different, very soon.