GeoPose: Patient-agnostic CTA-to-DSA registration through projection-space calibration
2026-08-17 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors developed GeoPose, a method to quickly and accurately align 3D CT images with 2D X-ray images taken during surgery without needing patient-specific training. GeoPose uses a neural network trained on many patients to estimate the camera position, then refines it to improve accuracy, doing so much faster than traditional methods. Their tests showed GeoPose performs better and faster than other approaches, even before additional optimization steps. This technique helps doctors better reconstruct 3D blood vessels from 2D images during procedures.
Digital subtraction angiography (DSA)Computed tomography angiography (CTA)3D-to-2D registrationC-arm pose estimationNeural networksImage registrationProjection-space calibrationResidual networkOptimizationBiplanar imaging
Authors
Rudolf L. M. van Herten, Robert Graf, Paula Feldman, Johannes C. Paetzold
Abstract
Aligning intraoperative biplanar digital subtraction angiography (DSA) to pre-procedural computed tomography angiography (CTA) requires rapid and accurate 3D-to-2D registration. Optimization-based methods are sensitive to initialization and may require hundreds of iterations, whereas learning-based approaches commonly rely on patient-specific training. We propose GeoPose, a population-trained framework that estimates the C-arm pose in a learned canonical frame and transfers it to the native frame of an unseen CTA through projection-space calibration and transform composition. A population-trained residual network refines the pose, followed optionally by low-budget image-driven optimization. GeoPose requires neither patient-specific adaptation nor explicit inter-volume preregistration. On 80 DSA observations from 20 held-out patients, optimization-free GeoPose achieved a carotid mean projected centerline distance (mPCD) of 5.8 mm and a clDice of 0.45, compared with 14.5 mm and 0.28 for the best-performing baseline, while requiring only 0.15 s. After 25 optimization iterations, GeoPose reached an mPCD of 4.6 mm and a clDice of 0.58 in approximately two seconds. Under the same budget, native-initialized optimization achieved 14.6 mm and 0.15, respectively. GeoPose thus provides rapid native-frame registration with fixed population-level weights and the geometric correspondence required for downstream biplanar 3D vascular reconstruction.