AI-assisted software for chronic wound detection: Development, validation, and certification strategy

Area of Science:

• Medical technology
• Artificial intelligence in healthcare
• Wound management

Background:

• Chronic wounds pose a significant challenge in healthcare, requiring timely and accurate assessment.
• Early identification and classification are crucial for effective treatment and patient outcomes.

Purpose of the Study:

• To evaluate the feasibility of an AI-assisted software tool for early chronic wound identification and classification.
• To support clinical decision-making regarding the need for professional review of chronic wounds.

Main Methods:

• Development of a pilot assistive software solution integrating wound image analysis (YOLO, U-Net) and patient-reported metadata.
• Establishment of a standardized wound imaging and annotation protocol.
• Evaluation of diagnostic performance using sensitivity, specificity, positive predictive value, and negative predictive value.

Main Results:

• A Python-based prototype and multimodal software architecture were successfully developed.
• A dataset of over 500 standardized wound images has been collected and is continuously expanded.
• A regulatory certification strategy was outlined, considering the Medical Device Regulation and the AI Act.

Conclusions:

• The study provides a technical and regulatory foundation for future AI-based wound care tools.
• The developed tool aims to support earlier triage and professional assessment of chronic wounds.
• This research paves the way for AI-driven advancements in chronic wound management.