BifDet: First 3D Airway Bifurcation Detection Dataset Released
Researchers have introduced BifDet, the first publicly available dataset specialized for 3D airway bifurcation detection. The dataset comprises annotated CT scans from the ATM22 open-access cohort, with bounding boxes covering parent and daughter branches at bifurcation points. Airway bifurcations are critical landmarks for understanding lung physiology, disease mechanisms, and lesion localization, yet no dedicated dataset existed for automated detection. The team fine-tuned RetinaNet on BifDet to demonstrate its utility. This fills a critical gap in resources for developing automated tools for respiratory disease analysis.
Key facts
- BifDet is the first publicly available dataset for 3D airway bifurcation detection.
- Dataset uses CT scans from the ATM22 open-access cohort.
- Annotations include bounding boxes for parent and daughter branches.
- Airway bifurcations are crucial for lung physiology and disease analysis.
- RetinaNet was fine-tuned and evaluated as a use-case.
- The dataset addresses a lack of annotated resources for bifurcation detection.
- Thoracic CT scans provide detailed insights into the airway tree.
- The work aims to advance automated detection and segmentation tools.
Entities
Institutions
- ATM22