Abstract
Purpose :
This research focuses on developing a custom video annotation software for ophthalmic surgery. The objective was to create a user-friendly tool enabling graders to select and annotate surgical instrument types, surgical instrument tip location in 3D space, chapters denoting surgical maneuvers, and ocular anatomy in high-resolution from individual video frames. The software aims to address the limitations of existing tools, providing a foundation for developing artificial intelligence systems for objective data extraction from surgical videos.
Methods :
The software development process was grounded in Python, with key dependencies like PyQt5 for constructing the graphical user interface and ffmpeg-python for efficient video processing. A systematic and logical architecture was implemented to store information in a readable way. The graphical user interface design centers around a video window, surrounded by customizable annotation buttons. These annotations provide a comprehensive label-set for the surgical video. Cursor tracking functionalities were incorporated to facilitate the precise labeling of surgical instrument tip locations. Furthermore, the inclusion of scrolling bars provides the ability to track instruments’ depth.
Results :
The video annotation tool accurately displays the graphical interface featuring a central video frame encircled by annotations. The software records desired annotations of surgical tools in the XY and Z planes successfully. Beta testing demonstrated efficient surgical video annotation, with subjective feedback indicating greater facility and easier workflow than existing video annotation software.
Conclusions :
The development of this beta software for surgical video annotation has yielded a tool that is subjectively more user-friendly for surgical video graders than existing options. The software is platform-independent and adaptable to label any surgical video. Anticipated to be a valuable asset, this tool lays the foundation for the future development of artificial intelligence systems to extract objective data from surgical videos, contributing to advancements in surgical research, surgical safety tools and education.
This abstract was presented at the 2024 ARVO Annual Meeting, held in Seattle, WA, May 5-9, 2024.