Mayo Clinic’s new AI tool could help catch post-surgery infections

Let's face it: surgical site infections (SSIs) are sneaky — and they can turn recovery into a nightmare if not caught early. But a team at the Mayo Clinic may have just handed the healthcare world a powerful new lens. Using artificial intelligence, they've created a tool that can spot infections from photos patients snap and send in themselves. Yes, smartphone pics.

"This process, currently done by clinicians, is time-consuming and can delay care," said Dr. Cornelius Thiels, a surgical oncologist and co-senior author of the study. "Our AI model can help triage these images automatically, improving early detection and streamlining communication."

In a world shifting toward virtual checkups and outpatient care, this AI-driven approach could be a game-changer for how we keep post-op patients safe — and give them peace of mind without making them trek back to the hospital.

How does it work?

The AI system isn't just a single clever algorithm — it's a two-part pipeline built to:

• Spot the incision: First, it figures out whether a photo even shows a surgical wound. This sounds simple, but with thousands of images submitted from patients' phones, that's no small feat.
• Detect infection signs: Then, it combs through the wound area to flag any signs of infection.

The researchers trained this model — called Vision Transformer — on a goldmine of data: over 20,000 images from 6,000+ patients at nine different Mayo Clinic hospitals. The results? A jaw-dropping 94% accuracy at detecting incisions, and an 81% AUC score when spotting infections. Not perfect, but for an AI looking at selfies of stitches, that's mighty impressive.

The AI doesn't make the final call, but it acts like a digital scout — alerting care teams to problems early and letting them focus on what truly needs attention.

Why does it matter?

Think about it:

• Patients healing at home can now get faster reassurance — or a quicker heads-up if something looks off.
• Care teams, especially in busy or rural settings, don't have to sift through thousands of patient-uploaded images.
• It's a win for efficiency, equity, and plain ol' common sense.

"For patients, this could mean faster reassurance or earlier identification of a problem," said Dr. Hala Muaddi, the study's lead author. "For clinicians, it offers a way to prioritize attention to cases that need it most."

What's more, the model held up across different patient groups — a crucial step in battling algorithmic bias, which has plagued some medical AI tools in the past.

Even better? It could eventually see signs of infection before a human eye can. That's like having a guard dog that smells trouble before it walks in the door.

The context

The study, published in Annals of Surgery, is part of a bigger picture. Healthcare is leaning hard into remote care and digital tools. And surgical recovery, long stuck in the old-school model of in-person visits and delayed diagnostics, is ripe for disruption.

• Post-op complications affect up to 5% of patients — and catching infections late often means hospital readmissions or worse.
• With virtual visits on the rise, the burden is shifting onto patients to monitor wounds — and that's risky without support.

Dr. Hojjat Salehinejad, another co-senior author, summed up the ambition neatly: "Our hope is that the AI models we developed — and the large dataset they were trained on — have the potential to fundamentally reshape how surgical follow-up is delivered."

Funded by Dalio Philanthropies and the Simons Family, the research is ongoing. Prospective studies are in the pipeline, testing how the AI integrates into real-world care. But the path is clear: a future where a smartphone and smart software work hand in hand to keep recovery on track.

In short, the Mayo team didn't just build a tool — they opened a door.