Led feasibility, development, and production deployment of an internal airway segmentation system for treatment planning, replacing a third-party API with a customized in-house pipeline.
- Led feasibility analysis, dataset strategy, and production development of an internal airway segmentation system for treatment planning.
- Designed the annotation and QA workflow, trained a customized U-Net, and deployed TensorRT inference in production.
- Replaced a third-party API, achieved over 95% overall recall, extended predictions into deeper airway branches, and reduced segmentation-related complaints by 90%.