Organizations: Vanke School of Public Health, Tsinghua University, Beijing, China · College of Computer Science and Electronic Engineering, Hunan University, Changsha, China · Department of Neurosurgery, the First Medical Center of Chinese PLA General Hospital, Beijing, China · Department of Neurosurgery, the Sixth Medical Center of Chinese PLA General Hospital, Beijing, China · Department of Neurosurgery, Air Force Medical Center of the Chinese PLA, Beijing, China
An X-ray angiogram sums every structure along a ray into one pixel. When vessels cross an aneurysm, they obscure part of it and mix their contrast into its filling curve. We present X-Strata, a network that separates a cine angiogram into three layers: the complete aneurysm sac, the vessels in front of it, and the vessels behind it. Contrast timing indicates how much each layer contributes to a pixel, and parallax between two views indicates their order, which a single view cannot determine. X-Strata first estimates the layers and their contrast curves, then uses the second view to assign vessels to the front or back layer, and lifts all layers into 3D. It learns from layer labels projected from same-session 3D rotational angiography, but needs only 2D cine runs and their acquisition geometry at inference. On 220 held-out patients from three centers (122 external), two-view X-Strata ranks first among 14 baselines, including VGGT and C2RV, in recovering the sac where vessels overlap it (Dice 0.693 internal, 0.641 external) and in ordering the layers (balanced accuracy 0.703, 0.659). Its order accuracy exceeds that of all 11 end-to-end baselines by at least 4.1 points internally and 3.3 externally (Holm-adjusted p<0.001) and grows with view separation. Recovered 3D sac surfaces have a median Chamfer distance of 0.34 mm on the internal cohort. X-Strata may therefore provide layered aneurysm and vessel information, including 3D sac shape, for patients who undergo only 2D angiography.
Figures & tables
Figure 2: Two-pass architecture of X-Strata. A ray-aware encoder processes cine frames and their C-arm geometry. Kinetic anchors initialize the layer time curves. The first-pass maps provide vessel centerlines and the visible sac rim for estimating parallax and shape cues. The second pass weights appearance, time, parallax, and shape to estimate layer path lengths and front-vessel probabilities. The layer maps are lifted into a shared 3D Gaussian representation for rendering. During training, projected labels from same-session 3D rotational angiography (3D-RA) supervise eligible pre-deployment runs. A pixel-wise gate controls path-length supervision, and depth-order labels are weighted separately.
Internal test Centers A and B ( n=98 )
External test Center C ( n=122 )
Method
Cues
Sac Dice ↑
O-Dice ↑
Vessel Dice ↑
DOA ↑
MAE ↓
O-Dice ↑
DOA ↑
One input run
nnU-Net 2D+t
.699
.607
.692
–
–
.533
–
SAM 2
.692
.598
.687
–
–
.569
–
SAM 3
.688
.595
.682
–
–
.546
–
Diffusion-VAS
.691
.598
.696
–
–
.536
–
Table 1: Amodal layer decomposition across cohorts. Squares denote appearance ( ), time ( ), parallax ( ), transmission physics ( ) and explicit shape ( ). O-Dice and DOA are defined in Sec. 5 . MAE is in mm. † Training-free solver using nnU-Net masks. Patients receive equal weight after averaging runs and seeds. Bold and underlining mark the best and next distinct displayed values within each fixed input count. Dashes indicate ineligible endpoints.
Figure 3: Two-run comparisons across cohorts. (a, b) Cohort means for O-Dice and balanced depth-order accuracy (DOA). (c) Patient-paired O-Dice differences for X-Strata against projected C 2 RV, projected DIF-Net, and VGGT. (d, e) Internal O-Dice by sac overlap and DOA by the view factor ∣sinφ∣ . Whiskers in (a–e) give pointwise 95% intervals from 3,000 patient bootstrap draws. A patient can contribute to several bins in (d, e). (f) Reliability curves pool eligible overlap pixels across seeds. Parentheses give the mean per-seed expected calibration error.
Variant
O-Dice ↑
DOA ↑
ECE ↓
CD ↓
TTP ↓
Full model (2 runs)
.693
.703
.030
0.34
0.144
Depth ordering
Attention-only parallax
–
.682
.052
–
–
Centerline parallax, frozen masks
–
.694
.050
–
–
Without moment anchor
–
.704
.047
–
–
With ordering constraint
–
.674
.054
–
–
Table 2: Two-run ablations in the internal cohort. Depth-only rows report DOA and expected calibration error (ECE). CD is median sac Chamfer distance in mm. TTP is patient-averaged absolute time-to-peak error in seconds on the mixture endpoint. Dashes indicate ineligible endpoints.
The TopAneu 2026 challenge requires intracranial aneurysms to be segmented at the voxel level and assigned to one of more than fifty vessel-specific classes, on CT and MR angiography alike. We address it with a six-stage anatomy-aware pipeline rather than a single segmentation network. A logistic-regression classifier routes each scan to a modality-specific branch; networks segment the vasculature, first as a binary mask and then over the full 36-class arterial taxonomy; a candidate localiser turns raw angiography and vessel priors into suspicious regions; a candidate-centred network segments the lesion in each crop; connected components become instances; and a per-instance classifier assigns each its anatomical location. Two principles organise the design. Whenever an upstream model output becomes a downstream training input it is generated out-of-fold, so no stage trains on unrealistically clean predictions of itself. And the anatomical decision is made once per lesion rather than once per voxel, which fits the data: of 406 connected aneurysm components in the full release, 405 carry exactly one location label. Task 1 is read off the Task 2 mask, making the two consistent by construction. We report stage-wise performance and a five-part ablation isolating what vessel priors and class-aware sampling contribute at each stage.
Gniewosz Drwiega, Wojciech Szymanski, Marek Wodzinski
Sano – Centre for Computational Personalised Medicine, International Research Foundation, Kraków, Poland · AGH University of Krakow, Kraków, Poland
Multi-view reasoning in coronary X-ray angiography is inherently a cross-projection geometric problem, yet automated report generation in this setting remains largely unexplored. The 3D vascular topology leads to projection-dependent branch overlap and foreshortening, rendering single-view modeling fundamentally incomplete and unstable for lesion localization and stenosis grading. Although multi-view fusion appears promising, learning anatomically consistent fusion from real angiograms is impeded by a critical limitation: cross-view alignment is unobservable and cannot be explicitly supervised. Consequently, conventional fusion relies on implicit correlations rather than verified anatomical correspondence. We address this by reformulating multi-view stenosis reporting as an alignment-constrained aggregation problem. A controllable synthetic angiography generation strategy is introduced to expose geometry-derived patch-level correspondence supervision unavailable in real data. An anatomy-correspondence module learns cross-view correspondence matrices that explicitly align auxiliary features within the main-view coordinate space prior to fusion, thereby constraining evidence aggregation to anatomically consistent regions. Experiments on synthetic data and zero-shot transfer to real angiograms show that this alignment-constrained design improves correspondence consistency and structured stenosis reporting compared to single-view modeling and conventional multi-view fusion methods. The code will be publicly available upon publication.
Chen Jia, Baochang Zhang, Fatia Kusuma Dewi +4
Computer Aided Medical Procedures, Technical University of Munich, Munich, Germany · relAI – Konrad Zuse School of Excellence in Reliable AI, Munich, Germany · German Heart Center Munich, Munich, Germany +3
X-ray angiography relies on iodinated contrast agents to visualize vascular structures during image-guided interventions. However, contrast administration carries risks of adverse events, motivating the development of contrast-free alternatives. Generating X-ray angiograms directly from non-contrast X-ray images offers a potential solution, but existing approaches remain limited by (i) insufficient control over vascular localization and (ii) inefficient modeling of redundant background content. To address these challenges, we propose VeCAS, a two-stage vessel-focused contrast-free angiogram synthesis framework that separates vascular structure localization from angiographic appearance synthesis. In Stage I, a discriminative model localizes vascular structures in non-contrast X-ray images, while cross-modality latent distillation transfers vessel-sensitive knowledge from X-ray angiograms during training. In Stage II, a vessel-focused inpainting model synthesizes angiographic appearance within the localized vascular regions while preserving the non-vascular background. Experiments on an in-house lower-limb vascular intervention dataset show that VeCAS outperforms the comparison methods in terms of vascular structural fidelity and image quality. Visual Turing tests and physician assessments indicate the perceptual realism of the synthesized angiograms. In addition, robotic guidewire navigation experiments in vascular phantoms show that VeCAS guidance reduces the time to target by 41.4% and the number of operation steps by 40.7% compared with non-contrast guidance. Together, these results suggest the potential of VeCAS to serve as ``meta contrast agent'' for vascular interventions.
De-Xing Huang, Chen-Yu Wang, Hao Liang +10
State Key Laboratory of Multimodal Artificial Intelligence Systems, Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China · School of Artificial Intelligence, University of Chinese Academy of Sciences, Beijing 100049, China · School of Advanced Interdisciplinary Sciences, University of Chinese Academy of Sciences, Beijing 101408, China +1