MAC-XA: Multi-view Anatomy-Correspondence Fusion for Coronary Stenosis Reporting from X-ray Angiography

2026-07-08 · Source: cs.CV updates on arXiv.org · Field: Science & Research — Artificial Intelligence & Machine Learning, Health & Medical Research · Depth: Expert, long

Summary

MAC-XA is a multi-view anatomy-correspondence fusion framework designed for coronary stenosis reporting from X-ray angiography. It addresses the challenge of unobservable cross-view alignment in real angiograms by reformulating the problem as an alignment-constrained aggregation. The system introduces a controllable synthetic angiography generation strategy to provide geometry-derived patch-level correspondence supervision, which is unavailable in real data. An Anatomy-Correspondence Module (ACM) learns explicit cross-view correspondence matrices, aligning auxiliary features within the main-view coordinate space prior to fusion. Experiments on 68,096 synthetic DRR angiograms from 38 patients (9,728 cases) and zero-shot transfer to 50 real angiogram cases demonstrate that MAC-XA improves correspondence consistency and structured stenosis reporting compared to single-view modeling and conventional multi-view fusion methods like Cross-Attention and DuoDuo.

Key takeaway

For AI Scientists developing automated multi-view coronary angiography reporting systems, you should prioritize explicit anatomical correspondence. Utilizing geometry-derived supervision from synthetic data, as demonstrated by MAC-XA, can overcome the lack of real-world alignment ground truth. This approach significantly improves structured stenosis reporting and correspondence consistency, enabling more reliable clinical decision support. Consider integrating similar alignment-constrained aggregation methods to enhance the robustness and generalizability of your models in complex real-world scenarios.

Key insights

MAC-XA uses geometry-derived synthetic data to explicitly supervise cross-view anatomical correspondence for robust multi-view coronary stenosis reporting.

Principles

• Multi-view reasoning in angiography requires explicit anatomical correspondence.
• Synthetic data can provide unobservable ground truth for alignment supervision.
• Alignment-constrained aggregation improves multi-view evidence fusion.

Method

Reformulate stenosis reporting as alignment-constrained aggregation. Generate synthetic angiograms with geometry-derived patch-level correspondence. Train an Anatomy-Correspondence Module (ACM) to learn cross-view alignment matrices, then fuse aligned features for report generation.

In practice

• Synthesize multi-view X-ray angiograms from 3D CTA volumes.
• Use RAD-DINO and Pose-ViT for view encoding.
• Apply focal BCE, attention CE, and negative-mass penalty for ACM training.

Topics

• Coronary Angiography
• Stenosis Reporting
• Multi-view Fusion
• Anatomical Correspondence
• Synthetic Data Generation
• X-ray Imaging

Best for: Computer Vision Engineer, AI Scientist, Research Scientist

Related on AIssential

• MAC-XA: Multi-view Anatomy-Correspondence Fusion for Coronary Stenosis Reporting from X-ray Angiography — MAC-XA improves multi-view coronary stenosis reporting by explicitly learning and applying anatomical correspondence for fusion.
• Transformation-driven generation of comparable projection images from multimodal anatomical scenes — A transformation-driven framework generates comparable synthetic projection images from multimodal anatomical scenes by explicitly modeling observation processes.
• The Slop Paradox: How Synthetic Standardization Erodes Clinical Uncertainty and Cross-Modal Alignment in AI-Rewritten Radiology Reports — AI-rewriting of clinical text for standardization can paradoxically degrade cross-modal alignment despite appearing cleaner.
• Clinically Aligned Geometry Constraints for Robust IVUS Vessel Boundary Segmentation — GeoCat improves IVUS segmentation by integrating geometry consistency loss for clinical accuracy.
• A Patient-Specific Pulmonary Arterial Tree Digital Twin to Extract Pulmonary Embolism Biomarkers — Automated digital twins of pulmonary arteries can rapidly quantify PE severity and clot distribution.
• Echo2ECG: Enhancing ECG Representations with Cardiac Morphology from Multi-View Echos — Aligning global ECGs with multi-view Echos resolves representational mismatch, yielding powerful, lightweight cardiac morphology features.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.CV updates on arXiv.org.