Vision Transformers (ViTs) leverage the attention mechanism for feature extraction but often suffer from high computational costs. To address this issue, prior works have introduced token reduction methods involving fixed-window local merging and global Bipartite Matching. However, these methods face significant challenges, such as insufficient merging due to fixed-size local windows and incorrect merging of informative tokens in global merging. To overcome these limitations, we propose Efficient Semantic Segmentation via Token Joint Merging (ESTJ) for ViT-based semantic segmentation networks. Specifically, ESTJ merges tokens using two strategies: Hierarchical Condition Pooling (HCP), which employs hierarchical local windows to effectively select sufficient tokens, and Protected Bipartite Matching (PBM), designed to preserve informative tokens using average similarity between a token and all other tokens. Experimental results demonstrate that ESTJ improves throughput by 75%, reduces GFLOPs by 40%, and enhances mIoU by up to 1.1%. Moreover, ESTJ can adjust the merging threshold during inference to adapt to scenarios that prioritize efficiency or accuracy. Compared to existing methods, ESTJ achieves a better balance between computational efficiency and segmentation accuracy.

Image-based 2D/3D registration is a critical technique for fluoroscopic guided surgical interventions. Conventional intensity-based 2D/3D registration approa- ches suffer from a limited capture range due to the presence of local minima in hand-crafted image similarity functions. In this work, we aim to extend the 2D/3D registration capture range with a fully differentiable deep network framework that learns to approximate a convex-shape similarity function. The network uses a novel Projective Spatial Transformer (ProST) module that has unique differentiability with respect to 3D pose parameters, and is trained using an innovative double backward gradient-driven loss function. We compare the most popular learning-based pose regression methods in the literature and use the well-established CMAES intensity-based registration as a benchmark. We report registration pose error, target registration error (TRE) and success rate (SR) with a threshold of 10mm for mean TRE. For the pelvis anatomy, the median TRE of ProST followed by CMAES is 4.4mm with a SR of 65.6% in simulation, and 2.2mm with a SR of 73.2% in real data. The CMAES SRs without using ProST registration are 28.5% and 36.0% in simulation and real data, respectively. Our results suggest that the proposed ProST network learns a practical similarity function, which vastly extends the capture range of conventional intensity-based 2D/3D registration. We believe that the unique differentiable property of ProST has the potential to benefit related 3D medical imaging research applications. The source code is available at https://github.com/gaocong13/Projective-Spatial-Transformers.

., surgical navigation) would benefit from a real-time method that can simultaneously track the endoscope and reconstruct the dense 3D geometry of the observed anatomy from a monocular endoscopic video. To this end, we develop a Simultaneous Localization and Mapping system by combining the learning-based appearance and optimizable geometry priors and factor graph optimization. The appearance and geometry priors are explicitly learned in an end-to-end differentiable training pipeline to master the task of pair-wise image alignment, one of the core components of the SLAM system. In our experiments, the proposed SLAM system is shown to robustly handle the challenges of texture scarceness and illumination variation that are commonly seen in endoscopy. The system generalizes well to unseen endoscopes and subjects and performs favorably compared with a state-of-the-art feature-based SLAM system. The code repository is available at https://github.com/lppllppl920/SAGE-SLAM.git.

Extracting geometric features from 3D models is a common first step in applications such as 3D registration, tracking, and scene flow estimation. Many hand-crafted and learning-based methods aim to produce consistent and distinguishable geometric features for 3D models with partial overlap. These methods work well in cases where the point density and scale of the overlapping 3D objects are similar, but struggle in applications where 3D data are obtained independently with unknown global scale and scene overlap. Unfortunately, instances of this resolution mismatch are common in practice, e.g., when aligning data from multiple sensors. In this work, we introduce a new normalization technique, Batch-Neighborhood Normalization, aiming to improve robustness to mean-std variation of local feature distributions that presumably can happen in samples with varying point density. We empirically demonstrate that the presented normalization method’s performance compares favorably to comparison methods in indoor and outdoor environments, and on a clinical dataset, on common point registration benchmarks in both standard and, particularly, resolution-mismatch settings. The source code and clinical dataset are available at https://github.com/lppllppl920/NeighborhoodNormalization-Pytorch.

Stereo depth estimation relies on optimal correspondence matching between pixels on epipolar lines in the left and right images to infer depth. In this work, we revisit the problem from a sequence-to-sequence correspondence perspective to replace cost volume construction with dense pixel matching using position information and attention. This approach, named STereo TRansformer (STTR), has several advantages: It 1) relaxes the limitation of a fixed disparity range, 2) identifies occluded regions and provides confidence estimates, and 3) imposes uniqueness constraints during the matching process. We report promising results on both synthetic and real-world datasets and demonstrate that STTR generalizes across different domains, even without fine-tuning.

Image stitching is a prominent challenge in medical imaging, where the limited field-of-view captured by single images prohibits holistic analysis of patient anatomy. The barrier that prevents straight-forward mosaicing of 2D images is depth mismatch due to parallax. In this work, we leverage the Fourier slice theorem to aggregate information from multiple transmission images in parallax-free domains using fundamental principles of X-ray image formation. The details of the stitched image are subsequently restored using a novel deep learning strategy that exploits similarity measures designed around frequency, as well as dense and sparse spatial image content. Our work provides evidence that reconstruction of orthographic mosaics is possible with realistic motions of the C-arm involving both translation and rotation. We also show that these orthographic mosaics enable metric measurements of clinically relevant quantities directly on the 2D image plane.

High-quality 3D reconstructions from endoscopy video play an important role in many clinical applications, including surgical navigation where they enable direct video-CT registration. While many methods exist for general multi-view 3D reconstruction, these methods often fail to deliver satisfactory performance on endoscopic video. Part of the reason is that local descriptors that establish pair-wise point correspondences, and thus drive reconstruction, struggle when confronted with the texture-scarce surface of anatomy. Learning-based dense descriptors usually have larger receptive fields enabling the encoding of global information, which can be used to disambiguate matches. In this work, we present an effective self-supervised training scheme and novel loss design for dense descriptor learning. In direct comparison to recent local and dense descriptors on an in-house sinus endoscopy dataset, we demonstrate that our proposed dense descriptor can generalize to unseen patients and scopes, thereby largely improving the performance of Structure from Motion (SfM) in terms of model density and completeness. We also evaluate our method on a public dense optical flow dataset and a small-scale SfM public dataset to further demonstrate the effectiveness and generality of our method. The source code is available at https://github.com/lppllppl920/DenseDescriptorLearning-Pytorch.

We present a self-supervised approach to training convolutional neural networks for dense depth estimation from monocular endoscopy data without a priori modeling of anatomy or shading. Our method only requires monocular endoscopic videos and a multi-view stereo method, e.g., structure from motion, to supervise learning in a sparse manner. Consequently, our method requires neither manual labeling nor patient computed tomography (CT) scan in the training and application phases. In a cross-patient experiment using CT scans as groundtruth, the proposed method achieved submillimeter mean residual error. In a comparison study to recent self-supervised depth estimation methods designed for natural video on in vivo sinus endoscopy data, we demonstrate that the proposed approach outperforms the previous methods by a large margin. The source code for this work is publicly available online at https://github.com/lppllppl920/EndoscopyDepthEstimation-Pytorch.

Robot-assisted needle insertion technology is now widely used in minimally invasive surgeries. The success of these surgeries highly depends on the accuracy of the position and orientation estimation of the needle tip. In this paper, we proposed an optical tracker based registration method for the marker-robot-needle tip registration in a robot-assisted needle insertion surgery. The method consists of two steps. First, with the guide of the optical tracker, we use the motion vector from the marker's current position to its target point as a feedback control to automatically register the robot-marker and the robot-tracker coordinates respectively. Then, in the procedure of marker-needle tip registration, we use two points on the needle holder to achieve a high accuracy needle orientation registration. We have conducted a series of experiments for the proposed method on a verified platform. Results show that the position estimation error is below 0.5mm and the orientation estimation error is below 0.5 degrees.