AI System Shows Promise in Early Detection of Post-Surgical Infections, Could Revolutionize Wound Care
ROCHESTER, MN – A new artificial intelligence (AI) system developed by Mayo Clinic researchers demonstrates a high degree of accuracy in identifying surgical site infections (SSIs) from patient-submitted images, potentially enabling earlier intervention and improved patient outcomes. The findings, recently accepted for publication in Annals of Surgery, represent a critically important step towards AI-assisted postoperative care, especially as telehealth and outpatient surgeries become increasingly prevalent.
Surgical site infections are a major concern in healthcare, representing up to 20% of all hospital-acquired infections and incurring an estimated $3.3 billion in annual costs to the U.S. healthcare system.Early detection is critical to preventing complications and reducing thes costs, but traditional methods relying on manual image review can be slow, especially outside of regular clinic hours. This new AI system aims to bridge that gap by providing real-time triage of patient-submitted wound images.
How the AI Works & Why It matters
The AI system was trained on a massive dataset of over 100,000 real-world images representing a diverse range of surgical procedures and skin tones. This extensive training allows the system to analyze images submitted by patients remotely and flag potential infections for clinician review.
“This work lays the foundation for AI-assisted postoperative wound care,” explains dr. Hala Muaddi, hepatopancreatobiliary fellow at Mayo Clinic and first author of the study. “It’s especially relevant as outpatient operations and virtual follow-ups become more common.”
The system’s ability to quickly analyze images offers several key benefits:
Faster Diagnosis: Reduces delays in identifying infections, leading to quicker treatment.
Improved Patient Interaction: Facilitates more timely communication between patients and their care teams.
Prioritized Care: Allows clinicians to focus their attention on patients who require immediate intervention.
Resource Optimization: Offers a valuable tool in rural or resource-limited settings where access to specialized wound care may be limited.
Addressing Concerns About Bias & Ensuring Equity
A core focus of the advancement process was ensuring the AI system performs equitably across all patient populations. Researchers conducted sensitivity analyses stratified by race, demonstrating comparable results across groups. Ongoing analysis is specifically examining performance across different skin tones to further validate accuracy and fairness.
“We are conducting additional analysis by skin tone to ensure the model performs accurately and equitably across all patients,” stated Dr. gregory Thiels, a co-author on the study. “This validation effort includes over 100,000 real-world images.”
future Directions & Implementation
While the initial results are promising, the mayo Clinic team is currently conducting prospective studies to evaluate the AI tool’s integration into routine surgical care. They are also working on a framework for broader implementation, both within Mayo Clinic and potentially extending access to patients beyond the institution.
“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 Dr. Hojjat Salehinejad, senior associate consultant of healthcare delivery research at Mayo Clinic and co-senior author.
Crucial Details Not Previously Highlighted:
Dataset Size: The AI was trained on a dataset exceeding 100,000 real-world images. Specific Focus of Validation: Ongoing validation specifically targets performance across different skin tones.
Co-Senior Author: Dr. Hojjat Salehinejad, a key figure in healthcare delivery research at Mayo Clinic, is a co-senior author on the study.
Link to publication:
