A Computer Vision Approach for Verifying the Count of Surgical Material and Instruments to Prevent Retention of Surgical Items

Retained surgical items (RSIs) pose serious risks to patient safety, often causing postoperative complications and mortality. Manual instrument counting remains error-prone due to time pressure and cognitive overload in operating rooms. To address this, we present SmartSurgiCount (SSC), a computer vision-based system integrating YOLOv11m object detection and hybrid OCR to automate surgical instrument verification. Accessible via smartphones, SSC processes pre- and postoperative images and handwritten count sheets, cross-referencing detected instruments with documented records in real time. Trained on 8,969 annotated instances across 18 classes, the model achieved 89.6% precision, 70.6% recall, and 79.3% mAP@50, while the OCR module reached 7.4% CER and 9.8% WER. The system delivers verification within 5–8 s, demonstrating the feasibility of intelligent vision solutions for reducing RSIs and enhancing surgical safety.