Evaluation of One-Stage Object Detectors for Implant Detection in Intraoperative CBCT Projection Images – MT Intro Talk by Yang Yuan

Join us for the intro talk of a Master’s thesis on “Evaluation of One-Stage Object Detectors for Implant Detection in Intraoperative CBCT Projection Images.” In modern surgery, metal implants are widely used, but their high X-ray attenuation introduces severe metal artifacts in CBCT reconstructions, significantly degrading image quality. To mitigate these effects, Metal Artifact Avoidance (MAA) strategies are employed in intraoperative imaging using mobile C-arm systems. A key prerequisite for effective MAA is the accurate detection and localization of metal implants in CBCT projection images, which enables optimized scan trajectory planning and improved 3D reconstruction quality.

This thesis investigates the use of one-stage object detectors, with a focus on the YOLO family, for implant detection in MAA applications. While two-stage detectors offer high accuracy, their computational cost limits real-time clinical use. In contrast, one-stage models provide an efficient alternative with good accuracy–speed trade-offs. This thesis will evaluate these models on both simulated and clinical CBCT projection data.