Abstract
The ability to recognize anatomical landmarks, microsurgical instruments, and complex scenes and events in a surgical wound using computer vision presents new opportunities for studying microsurgery effectiveness. In this study, we aimed to develop an artificial intelligence-based solution for detecting, segmenting, and tracking microinstruments using a neurosurgical microscope. We have developed a technique to process videos from microscope camera, which involves creating a segmentation mask for the instrument and subsequently tracking it. We compared two segmentation approaches: (1) semantic segmentation using Visual Transformers (pre-trained domain-specific EndoViT model), enhanced with tracking as described by Cheng Y. et al. (our proposed approach), and (2) instance segmentation with tracking based on the YOLOv8l-seg architecture. We conducted experiments using the CholecSeg8k dataset and our proprietary set of neurosurgical videos (PSNV) from microscope. Our approach with tracking outperformed YOLOv8l-seg-based solutions and EndoViT model with no tracking on both CholecSeg8k (mean IoT = 0.8158, mean Dice = 0.8657) and PSNV (mean IoT = 0.7196, mean Dice = 0.8202) datasets. Our experiments with identifying neurosurgical instruments in a microscope's field of view showcase the high quality of these technologies and their potential for valuable applications.
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
