Using Artificial Intelligence Deep Learning to Detect and Prevent Retained Surgical Sponges

Area of Science:

• Medical Technology
• Artificial Intelligence
• Patient Safety

Background:

• Retained surgical items (RSIs), particularly sponges (RSS), are a persistent patient safety concern.
• Traditional prevention methods like manual counting and RFID have limitations, leading to ongoing surgical errors.
• Artificial intelligence (AI) presents a novel approach to enhance the detection of retained surgical items.

Purpose of the Study:

• To review the application of AI, specifically deep learning models, in preventing retained surgical sponges.
• To evaluate the effectiveness of AI techniques in improving the detection of RSS in surgical settings.
• To identify challenges and future directions for AI in RSI prevention.

Main Methods:

• Review of studies applying AI, including Convolutional Neural Networks (CNNs) and Artificial Neural Networks (ANNs), for RSS detection.
• Analysis of AI models' performance in interpreting medical images (X-rays) and video feeds (laparoscopy).
• Examination of object detection models (e.g., YOLO) and computer-aided detection (CAD) systems.

Main Results:

• CNN-based models demonstrate significant improvements in RSS detection from X-rays and laparoscopic videos, often surpassing human accuracy.
• Object detection models show promise for real-time RSS tracking during surgery.
• ANN-based CAD systems, especially with radiopaque markers, enhance the accuracy of identifying retained sponges.

Conclusions:

• AI, particularly deep learning, holds significant potential to enhance patient safety by reducing retained surgical sponges.
• Challenges such as data limitations and false positives require further research for effective AI integration.
• Widespread adoption of AI detection systems can improve patient outcomes, reduce healthcare costs, and prevent surgical never events.