ARIADNE: A Perception-Reasoning Synergy Framework for Trustworthy Coronary Angiography Analysis

Coronary artery disease (CAD) is a leading cause of morbidity and mortality worldwide. Invasive X-ray coronary angiography (XCA) serves as the primary tool for CAD diagnosis and guidance of percutaneous coronary interventions (PCI), offering high temporal resolution necessary for visualizing hemodynamic flow. However, current clinical workflows rely heavily on manual interpretation, a process characterized by significant inter-observer variability and susceptibility to clinician fatigue. As healthcare institutions universally adopt Picture Archiving and Communication Systems (PACS), a critical gap persists between passive image storage and active, automated clinical interpretation. While hospitals have implemented digital image storage, they lack automated systems capable of transforming raw imaging data into actionable clinical insights. The growing volume of interventional procedures makes purely manual interpretation increasingly unsustainable, creating demand for computer-aided diagnosis systems that can bridge the gap between data acquisition and clinical decision-making.

Accurate segmentation of the coronary vascular tree represents a fundamental prerequisite for automated coronary analysis. Over the past decade, convolutional neural networks (CNNs), particularly U-Net and its attention-enhanced variants such as CS-Net and SA-UNet, have dominated the field. More recently, vision transformers (ViTs) have been introduced to capture global spatial relationships. Despite achieving high pixel-level performance metrics, these models face a critical limitation in preserving vascular topology. Traditional loss functions, including cross-entropy and Dice loss, optimize pixel-level accuracy independently without explicitly penalizing topological errors. Consequently, these models frequently produce fragmented vessel trees where distal branches appear disconnected, particularly due to signal loss in thin vessels during downsampling operations. In coronary hemodynamics analysis, topological connectivity is essential; a segmentation with a high Dice score remains insufficient for clinical use if discontinuities prevent accurate centerline extraction and subsequent geometric analysis.

The core problem in coronary angiography analysis is ensuring topological consistency in vessel segmentation while maintaining high pixel-level accuracy. Conventional pixel-wise loss functions fail to enforce topological constraints in coronary vessel segmentation, producing fragmented vascular trees despite high pixel-level accuracy. Existing automated frameworks predominantly follow a sequential approach where segmentation and stenosis detection are performed as independent tasks. However, these deterministic algorithms lack the ability to distinguish pathological stenosis from common anatomical artifacts, resulting in elevated false positive rates.

The core innovations of the ARIADNE framework include the integration of a preference-aligned perception module and an RL-based reasoning module. • The perception module employs DPO to fine-tune the Sa2VA vision-language model using Betti number constraints as preference signals, ensuring geometrically complete vessel structures. • The reasoning module formulates stenosis localization as a Markov Decision Process with an explicit rejection mechanism, autonomously deferring ambiguous anatomical candidates like bifurcations and vessel crossings. • This represents the first application of DPO for topological alignment in medical imaging, demonstrating that preference-based learning mitigates topological violations while maintaining diagnostic sensitivity.

The ARIADNE framework is designed to emulate the hierarchical decision-making process of human experts through two biomimetic stages: a perception module for anatomically consistent vascular reconstruction and a reasoning module for context-aware lesion localization. • The perception module employs the Sa2VA foundation model with a progressive training strategy designed to enforce topological continuity throughout the segmentation process. We integrate DPO into the training pipeline to align model outputs toward geometrically complete vessel structures rather than fragmented pixel-level predictions. • The reasoning module operates as a structure-guided diagnostic agent that navigates the extracted vessel skeleton to identify stenotic lesions. We develop an RL agent that analyzes local geometric features, including radius gradients and curvature patterns, to perform context-aware lesion localization.

The experimental design was structured to address two primary objectives: evaluation of topological consistency in vascular segmentation across diverse angiographic conditions, and assessment of stenosis detection accuracy and false positive management in anatomically complex scenarios. • A proprietary dataset was curated from coronary angiography video sequences acquired at Guizhou Aviation Industry Group 302 Hospital using a Siemens angiography system. • The dataset comprises 1,400 high-resolution images at 512×512 resolution from 35 patients, with an average of 40 frames extracted per patient to capture varying vessel angulations and contrast conditions. • To prevent data leakage inherent in video-based acquisitions, the dataset was partitioned at the patient level rather than the image level.

On 1,400 clinical angiograms, ARIADNE achieves a state-of-the-art centerline Dice of 0.838, reducing false positives by 41% compared to geometric baselines. • External validation on multi-center benchmarks ARCADE and XCAD confirms generalization across acquisition protocols. • This represents the first application of DPO for topological alignment in medical imaging, demonstrating that preference-based learning mitigates topological violations while maintaining diagnostic sensitivity.

The ARIADNE framework can be directly applied to interventional cardiology workflows to enhance the automation of coronary angiography analysis. • The method significantly reduces false positive rates while maintaining high sensitivity, improving diagnostic reliability. • Future exploration could extend the framework's application to other medical imaging fields, such as cerebral angiography or pulmonary CT scans.

While ARIADNE performs well in coronary angiography analysis, the model's topological consistency remains unstable in low-contrast and complex anatomical structures like vessel crossings. • The preference-aligned perception module requires a large amount of annotated data for training, which may limit its practical deployment. • In extreme cases, the rejection mechanism may defer too many ambiguous candidates, affecting diagnostic efficiency.