Pedicle Screw Pairing and Registration for Screw Pose Estimation from Dual C-arm Images Using CAD Models

TL;DR

The paper tackles the challenge of accurate pedicle screw pose estimation from dual C-arm AP/LAT images, where correct screw correspondence is crucial for reliable registration. It introduces a CAD-model–assisted, dual-view pipeline that projects CAD screws, computes a Gradient Correlation Loss $\mathcal{L}_{grad}$ on AP/LAT projections, and optimizes pose via differential evolution to align the CAD with real screws. The study shows that correct screw pairing yields superior Dice scores and lower projection losses in both pre- and post-registration stages, underscoring the importance of dual-view constraints for robust 3D pose estimation. The approach promises practical improvements in intraoperative screw localization and spine surgical safety by providing reliable, image-based feedback on screw positioning.

Abstract

Accurate matching of pedicle screws in both anteroposterior (AP) and lateral (LAT) images is critical for successful spinal decompression and stabilization during surgery. However, establishing screw correspondence, especially in LAT views, remains a significant clinical challenge. This paper introduces a method to address pedicle screw correspondence and pose estimation from dual C-arm images. By comparing screw combinations, the approach demonstrates consistent accuracy in both pairing and registration tasks. The method also employs 2D-3D alignment with screw CAD 3D models to accurately pair and estimate screw pose from dual views. Our results show that the correct screw combination consistently outperforms incorrect pairings across all test cases, even prior to registration. After registration, the correct combination further enhances alignment between projections and images, significantly reducing projection error. This approach shows promise for improving surgical outcomes in spinal procedures by providing reliable feedback on screw positioning.

Figures (5)

• Figure 1: Screw classification and registration pipeline. Red and orange circles represents the screw tip and light blue and purple circles represents the center of screw body for different screws. (a) Calculate 3D coordinates from 2D coordinates in AP and LAT views and align two CAD models. (b) Rotate and translate the models. (c) Project the models into AP and LAT views. (d) Compute the loss between the projection and the real data. (e) Predict the pairs of the screws and repeat the process until the loss converges for accurate screw registration.
• Figure 2: Overlay after after alignment based on 2D coordinates and rotation on axial axis. Image on the left shows the result from the correct combination and image on the right shows the result from the wrong combination. The screw in the white color is the projection from the 3D CAD model and the screw in the gray color is the real screw.
• Figure 3: Ground truth screw matching. Left: LAT and AP images of the original screw position. Right: LAT and AP images after placing a needle adjacent to the screw. The needle serves as a spatial reference to establish precise correspondence of the screw across both views.
• Figure 4: Overlay after registration. Image on the left shows the result from the correct combination and image on the right shows the result from the wrong combination.
• Figure 5: Cases where the screw and tower are not rigidly fixed to each other. Pedicle screws in the red box show differing orientations between the screw and the tower components.