Tracking anatomy and surgical devices in pediatric SCFE using a stationary C-arm with a rotating X-ray source

Le résumé fourni par la source

Purpose. Accurate screw placement is critical in slipped capital femoral epiphysis (SCFE), where malposition can cause long-term complications. This work presents a learning-based 3D registration framework for intraoperative tracking of the femur and guidewires using sparse, low-dose 2D fluoroscopy from a stationary C-arm with a rotating x-ray source. Ablation studies investigate the impact of key simulation and training parameters on registration performance. Methods. The approach uses sparse radiographic views acquired from an nView s1 cone-beam tomosynthesis system. Training data are simulated from patient CT volumes, employing virtual guidewire models with randomized anatomy and device poses. A dual-branch network extracts anatomy and device features from radiographs to regress pose parameters. Cascaded inferences iteratively refine predictions. The method was validated on 3400 images simulated from 34 subjects. Ablation studies examined the effects of the training dataset size, projection spacing, and the number of views. Results. Preliminary studies demonstrated a median target registration error (TRE) of 1.26 mm when registering the femur anatomy, and tip/orientation errors of 0.4 mm and 0.7° when registering guidewires. Algorithm runtime was 1 s total for femur registration and 0.1 s total for device registration. Ablation experiments showed that performance improved with increasing dataset size up to 1.6 million samples. Projection spacing of 1.64 mm provided the optimal balance between accuracy and efficiency. Using four evenly distributed views achieved the lowest TRE. Conclusions. The study demonstrated accurate femur and guidewire registration via direct pose regression, and identified key simulation and training parameters. Compared to iterative model-based optimization, the solution offers a favorable trade-off between accuracy, robustness, and efficiency, supporting low-dose intraoperative guidance well-suited for guiding implant placement in pediatric orthopaedics.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.