AANS Beyond 2021: Scientific Papers Collection-Deep Neural Networks Can Perform Automated Instrument Detection In Endoscopic Skill Base Surgery

Deep Neural Networks Can Perform Automated Instrument Detection In Endoscopic Skill Base Surgery

Video Transcription

Video Summary

In the video, Guillaume Couganer discusses his team's project on using deep neural networks for automated instrument detection in endoscopic skull-based surgery. This project was awarded the Mizzouho Minimally Invasive Brain Tumor Surgery Award by the AANS. Couganer highlights the importance of computer vision in surgical research and describes the development of an intraoperative video dataset and analytical pipeline. The objective is to demonstrate that intraoperative video analysis is valuable and can improve surgical outcomes. The team developed a high-fidelity simulator to train surgeons in managing complications like internal carotid artery injury. They collected a large dataset of 150 trials and used deep learning techniques to automatically identify surgical tools in video footage. The results showed promising potential for using video analysis to understand surgical patterns and outcomes. Challenges for future research include improving accuracy and precision, creating interpretable metrics, and accessing diverse datasets from multiple institutions. The team also plans to apply their approach to other surgical settings within neurosurgery. The dataset used in this project is publicly available for download on Figshare.

Keywords

deep neural networks

automated instrument detection

endoscopic skull-based surgery

computer vision

intraoperative video analysis