Back to 2025 Resident Posters
Artificial Intelligence Automated Segmentation of Microsurgical Instruments in Intraoperative Surgical Microscope Videos Using Convolutional Neural Networks
Berk B. Ozmen, M.D., Graham S. Schwarz, MD, MSE, FACS.
Cleveland Clinic, Cleveland, OH, USA.
PURPOSE: Precise identification of microsurgical instruments is crucial for procedural accuracy and patient safety. Integrating artificial intelligence (AI) into microsurgery offers promising avenues to enhance surgical precision and workflow. This study aims to develop and evaluate a convolutional neural network (CNN)-based model for automated, real-time detection of microsurgical instruments.
METHODS: A custom CNN model was developed using TensorFlow and Keras. A dataset of 316 high-resolution intraoperative images from lymphaticovenous anastomosis procedures was divided into training (227 images), validation (57 images), and testing (32 images) sets. Data augmentation techniques were applied to enhance generalizability. The model was optimized using the Adam algorithm with a custom loss function. Performance was evaluated using accuracy, loss metrics, Dice coefficient, and Jaccard index.
RESULTS: The model achieved 98.15% accuracy on the test set, with a loss of 0.1397, a Dice coefficient of 0.9114, and a Jaccard index of 0.8373, indicating excellent segmentation quality and robust instrument detection. Qualitative assessment confirmed accurate instrument identification across varying magnifications typical in microsurgical settings.
CONCLUSION: We present the first accurate CNN-based AI model for real-time microsurgical instrument detection. Its robust performance suggests significant potential for integration into computer-assisted microsurgery systems, enhancing surgical precision and workflow in microsurgical procedures.
Back to 2025 Resident Posters
