Vocal cord localization improved through local point tracking method
Beyond Region Memory: Local Point Tracking for Vocal Cord Localization in Video Laryngoscopy
Computer Vision and Pattern Recognition
Summary
Finding the vocal cords accurately during emergency video laryngoscopy is difficult because the tissues move quickly and instruments can block the view. The authors propose a new way to track points in the video frames to keep detecting the vocal cords even when visibility gets tricky. Their method first finds key points on the vocal cords, follows these points through the video, and then adjusts the detected area to match the vocal cords more tightly. Tests on multiple datasets show this approach works better than previous methods, especially in keeping up with fast movements and improving boundary accuracy.
What this means in practice
- •For medical device developers: Improve emergency video laryngoscopy devices by integrating local point tracking to enhance vocal cord detection accuracy under challenging conditions.$Commercial implications: Enables development of more reliable emergency intubation tools with better vocal cord visualization, improving patient outcomes.
- •For medical imaging software teams: Enhance video laryngoscopy analysis software by adding local point tracking to maintain vocal cord localization during occlusions and rapid tissue movement.
Authors
Huayu Wang, Gyungmin Ko, Jung Heon Kim, Jia-Xian Jian, Dae Kon Kim, Dong Keon Lee, Hsiang-Wei Huang, Pau-Choo Chung, Linda Shapiro, Richard B. Utarnachitt, Jenq-Neng Hwang, Nathan J. White
Abstract
Reliable vocal-cord localization in emergency video laryngoscopy remains visually challenging due to rapid tissue deformation, specular highlights, and instrument occlusions. We propose Local Point Tracking (LPT), a robust framework that combines a single-frame detector, local point propagation, and stateless mask refinement. LPT follows a localize first, snap second pipeline. First, a single-frame semantic detector generates sparse anchors, within which LPT initializes local query points (SAM-supported by default, with a regular grid as fallback). A point-propagation module tracks these points through detector-missed frames and recovers temporal bounding boxes from their geometry. Second, because point-derived boxes can be loose around the visible boundary, a box-snap step refines the selected box to the target boundary. This decomposition keeps temporal state in local point coordinates rather than a stateful mask or whole-region representation, while the spatial refinement module handles single-frame boundary precision without feeding mask errors back into temporal memory. Extensive evaluations on two distinct datasets demonstrate that our proposed LPT method achieves strong overall performance against existing baselines, with the best mAP_50, success AUC, and missing rate on both datasets. Furthermore, ablation studies confirm that point tracking primarily enhances temporal tracking robustness, while box-snap refinement effectively improves boundary tightness.