Aug. 09, 2025
Researchers at Mayo Clinic have developed an artificial intelligence (AI) system capable of detecting surgical site infections (SSIs) from patient-submitted postoperative wound images with high accuracy. This innovation, detailed in the Annals of Surgery, represents a leap forward in remote patient monitoring and infection prevention.
A two-stage AI pipeline for SSI detection
The AI tool employs a two-stage pipeline. First, it identifies whether an image contains a surgical incision. Second, it evaluates the incision for signs of infection. The model was trained on a robust dataset of over 20,000 images from more than 6,000 patients across nine Mayo Clinic hospitals. It achieved a 94% accuracy rate in detecting incisions and an 81% area under the curve (AUC) for identifying infections.
"This process, currently done by clinicians, is time-consuming and can delay care," says Cornelius Thiels, D.O., co-senior author and surgical oncologist at Mayo Clinic in Rochester, Minnesota. "Our AI model can help triage these images automatically, improving early detection and streamlining communication between patients and their care teams."
The model is based on deep learning, using layers of artificial neurons to extract features such as edges and patterns that may not be visible to the human eye. These features help distinguish between infected and noninfected wounds. The model also categorizes image types and flags low-quality submissions, which is critical for real-world use where image clarity can vary.
Clinical relevance and implications for practice
This model offers a scalable solution for monitoring surgical wounds remotely, helping clinicians identify infections earlier and prioritize care more effectively.
"This work lays the foundation for AI-assisted postoperative wound care," says Hala Muaddi, M.D., Ph.D., first author and hepatopancreatobiliary fellow at Mayo Clinic. "It's especially relevant as outpatient operations and virtual follow-ups become more common."
Manual review of patient-submitted images can delay care, especially when submissions occur outside clinic hours. The AI system addresses this gap by triaging images in real time, potentially reducing delays in diagnosis and improving communication between patients and care teams.
SSIs remain a significant challenge, accounting for up to 20% of hospital-acquired infections and costing the U.S. healthcare system an estimated $3.3 billion annually. By identifying infections earlier, the AI tool may help reduce complications, lower healthcare costs and improve recovery outcomes.
"For patients, this could mean faster reassurance or earlier identification of a problem," says Dr. Muaddi. "For clinicians, it offers a way to focus attention where it's needed most — especially in rural or resource-limited settings."
Addressing algorithmic bias
Ensuring the model performs across patient populations was a central goal of development. To support this, the team trained the AI system using real-world images representing a broad range of surgical procedures and skin colors.
Ensuring the model performs consistently across diverse groups helps address concerns about algorithmic bias. Sensitivity analyses stratified by race showed comparable results. "We are conducting additional analysis by skin tone to ensure the model performs accurately and equitably across all patients. This validation effort includes over 100,000 real-world images," said Dr. Thiels.
Next steps: Prospective validation underway
While the results are promising, the team says that further validation is needed.
"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," says Hojjat Salehinejad, Ph.D., a senior associate consultant of healthcare delivery research within the Mayo Clinic Robert D. and Patricia E. Kern Center for the Science of Health Care Delivery at Mayo Clinic in Rochester, Minnesota, and co-senior author. "Prospective studies are underway to evaluate how well this tool integrates into day-to-day surgical care."
Dr. Thiels added that the team is working on a larger-scale validation and developing a framework for implementation at Mayo Clinic and ways to bring this to patients outside of Mayo Clinic.
For more information
Muaddi H, et al. Imaging based surgical site infection detection using artificial intelligence. Annals of Surgery. In press.
Refer a patient to Mayo Clinic.