Background: Artificial intelligence (AI) integration in microsurgery represents a frontier with significant potential to enhance surgical precision. Despite advancements in computer vision and machine learning, their application in microsurgical settings remains limited. This study aims to develop a Convolutional Neural Network (CNN) for automated detection of microsurgical instruments in microsurgical fields.

 

Methods: A custom CNN model was developed using TensorFlow and Keras frameworks. The dataset comprised 316 high-resolution intraoperative images from lymphovenous anastomosis procedures, allocated as 227 for training, 57 for validation, and 32 for testing. A U-Net architecture with VGG16 encoder was implemented, optimized using the Adam algorithm with a hybrid loss function combining binary cross-entropy and Dice coefficient. Performance was evaluated using accuracy, loss metrics, Dice coefficient, and Jaccard index.

 

Results: The CNN model demonstrated exceptional performance metrics across all evaluation criteria. Test set results showed 98.15% accuracy, 0.1397 combined loss, 0.9114 Dice coefficient, and 0.8373 Jaccard index, indicating high-quality segmentation and robust instrument detection capabilities. The model maintained consistent performance under varying magnifications typical in microsurgical settings.

 

Conclusions: This study presents a highly accurate CNN-based AI model for microsurgical instrument detection. The robust performance suggests significant potential for integration into computer-assisted microsurgery systems and workflow potentially enhancing microsurgical outcomes. These findings provide a foundation for future development of autonomous or semi-autonomous surgical systems in microsurgery.