In image-guided surgery (IGS) practice, combining intraoperative 2D X-ray images with preoperative 3D X-ray images from computed tomography (CT) enables the rapid and accurate localization of lesions, which allows for a more minimally invasive and efficient surgery, and also reduces the risk of secondary injuries to nerves and vessels. Conventional optimization-based methods for 2D X-ray and 3D CT matching are limited in speed and precision due to non-convex optimization spaces and a constrained searching range. Recently, deep learning (DL) approaches have demonstrated remarkable proficiency in solving complex nonlinear 2D–3D registration. In this paper, a fast and robust DL-based registration method is proposed that takes an intraoperative 2D X-ray image as input, compares it with the preoperative 3D CT, and outputs their relative pose in x, y, z and pitch, yaw, roll. The method employs a dual-channel Swin transformer feature extractor equipped with attention mechanisms and feature pyramid to facilitate the correlation between features of the 2D X-ray and anatomical pose of CT. Tests on three different regions of interest acquired from open-source datasets show that our method can achieve high pose estimation accuracy (mean rotation and translation error of 0.142° and 0.362 mm, respectively) in a short time (0.02 s). Robustness tests indicate that our proposed method can maintain zero registration failures across varying levels of noise. This generalizable learning-based 2D (X-ray) and 3D (CT) registration algorithm owns promising applications in surgical navigation, targeted radiotherapy, and other clinical operations, with substantial potential for enhancing the accuracy and efficiency of image-guided surgery.

This work reports an information-based landmarks assisted simultaneous localization and mapping (InfoLa-SLAM) in large-scale scenes using single-line lidar. The solution employed two novel designs. The first design was a keyframe selection method based on Fisher information, which reduced the computational cost of the nonlinear optimization for the back-end of SLAM by selecting a relatively small number of keyframes while ensuring the accuracy of mapping. The Fisher information was acquired from the point cloud registration between the current frame and the previous keyframe. The second design was an efficient global descriptor for place recognition, which was achieved by designing a unique graphical feature ID to effectively match the local map with the global one. The results showed that compared with traditional keyframe selection strategies (e.g., based on time, angle, or distance), the proposed method allowed for a 35.16% reduction in the number of keyframes in a warehouse with an area of about 10,000 m². The relocalization module demonstrated a high probability (96%) of correction even under high levels of measurement noise (0.05 m), while the time consumption for relocalization was below 28 ms. The proposed InfoLa-SLAM was also compared with Cartographer under the same dataset. The results showed that InfoLa-SLAM achieved very similar mapping accuracy to Cartographer but excelled in lightweight performance, achieving a 9.11% reduction in the CPU load and a significant 56.67% decrease in the memory consumption.