<< Back to the abstract archive
Artificial Intelligence for Microsurgery: Microsurgical Instruments Detection With Convolutional Neural Networks
Berk B Ozmen, MD; Graham S Schwarz, MD, MSE, FACS
Cleveland Clinic Department of Plastic Surgery
2025-01-10
Presenter: Berk B Ozmen, MD
Affidavit:
Yes
Director Name: Graham S Schwarz, MD, MSE, FACS
Author Category: Fellow Plastic Surgery
Presentation Category: Clinical
Abstract Category: General Reconstruction
Background: Artificial Intelligence (AI) integration into microsurgery represents a promising frontier for enhancing surgical precision and outcomes. Recent advancements in computer vision and machine learning offer potential solutions to augment the capabilities of microsurgeons in complex procedures. For AI technologies to be developed, robust computer vision models are needed for integration into surgical microscopes. This study presents a convolutional neural network (CNN) for real-time detection of microsurgical instruments, representing an initial step toward AI-assisted microsurgery.
Methods: We developed a custom AI CNN model using TensorFlow and Keras, specifically tailored for microsurgical instrument detection. The dataset comprised 316 high-resolution intraoperative images obtained from lymphovenous anastomosis procedures. We applied data augmentation techniques to enhance model generalizability. The model was optimized using the Adam algorithm with a custom loss function combining binary cross-entropy and Dice coefficient. Performance was evaluated using accuracy, loss metrics, Dice coefficient, and Jaccard index.
Results: The AI CNN model demonstrated high performance across all evaluation metrics. On the test set, the model achieved an accuracy of 98.15%, a combined loss of 0.1397, a Dice coefficient of 0.9114, and a Jaccard index of 0.8373. These results indicate excellent segmentation quality and robust microsurgical instrument detection capabilities. Qualitative assessment of test set overlay images confirmed accurate instrument identification.
Conclusion: This study demonstrates the feasibility of real-time microsurgical instrument detection using AI. The model's robust performance supports the potential integration of AI into microsurgical practice for AI-based decision support and visualization tools.
