Machine Learning Approaches for the Image-Based Identification of Surgical Wound Infections: Scoping Review

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Image Analysis for Diagnostics

Background:

• Surgical site infections (SSIs) are a common complication impacting patient outcomes and healthcare costs.
• Current remote surveillance of surgical wounds relies on manual clinician assessment, limiting scalability.
• Machine learning (ML) offers potential for cost-effective, scalable remote assessment of postoperative wound healing.

Purpose of the Study:

• To review machine learning (ML) methods used for identifying surgical wound infections from images.
• To provide an overview of the current landscape of ML applications in SSI detection.

Main Methods:

• Scoping review of ML approaches for visual SSI detection, adhering to JBI methodology.
• Searched multiple databases (MEDLINE, Embase, etc.) for relevant studies up to November 2022.
• Assessed study eligibility, extracted data, and evaluated reporting using TRIPOD and PROBAST tools.

Main Results:

• 10 studies met eligibility criteria from 715 unique records, utilizing diverse clinical contexts.
• Both traditional ML and deep learning methods were employed for SSI identification from color images.
• Most studies exhibited high risk of bias (RoB) according to PROBAST, with no external validation performed.

Conclusions:

• Image-based identification of SSIs using ML is an emerging research area requiring standardized reporting.
• Future research must address limitations in image capture, model development, and data sources.
• Improvements in methodology are needed to enhance the reliability and applicability of ML for SSI detection.