FlowReg is a versatile end-to-end framework for spinal CT/X-ray registration that supports both single- and multi-vertebrae scenarios. It leverages CT volume reconstruction from intraoperative biplanar X-rays as a spatial prior, learns deterministic pose gradients on the SE(3) manifold, and bridges cross-modal gaps with Difference-Aware Feature Fusion.

MedCenterDet is an anchor-free, center-based object detection framework for 3D medical images. It predicts object centers through heatmap regression, decodes bounding box dimensions with anisotropic Gaussian target generation, and uses Counterfactual Attention Learning plus two-stage refinement for precise localization.

Computed Tomography (CT)/X-ray registration in image-guided navigation remains challenging because of its stringent requirements for high accuracy and real-time performance. We introduce RadGS-Reg, a novel framework for vertebral-level CT/X-ray registration through joint 3D Radiative Gaussians (RadGS) reconstruction and 3D/3D registration.

Tumour-infiltrating lymphocytes (TILs), as a new prognostic biomarker, are of important clinical value and have an association with improved survival rates. We build a multimodal fusion pipeline framework based on CrossTransformer and KNN omics feature screening.

We determine whether the patient has the infraorbital posterior ethmoid cells through CT segmentation, assisting doctors in diagnosis. This work is in cooperation with the Department of Otolaryngology of Nanjing Tongren Hospital.

We introduce an abductive learning mechanism for semi-supervised vertebral localization. The approach combines visual object detection with first-order logic predicate reasoning, improving the recognition rate through logical reasoning.