Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CardioSyntax: end-to-end SYNTAX score prediction -- dataset, benchmark and method

Alexander Ponomarchuk, Ivan Kruzhilov, Galina Zubkova, Artem Shadrin, Ruslan Utegenov, Ivan Bessonov, Pavel Blinov

TL;DR

CardioSYNTAX tackles automatic SYNTAX score estimation from invasive coronary angiography by introducing a public, multi-view angiography dataset of 3,018 patients and an end-to-end Multi-View Look (MVL) framework that fuses 3D video features from multiple views with separate RCA/LCA heads to predict the total score. The approach achieves an $R^2$ of $0.51$ for score regression and $77.3\%$ zero-score accuracy, while demonstrating domain-shift robustness and strong coronary dominance classification performance. The work also quantifies inter-expert disagreement (around $R^2\approx0.7$) and provides a domain-shift benchmark, highlighting challenges such as score distribution effects and cross-system generalization. Overall, CardioSYNTAX establishes a public benchmark and a viable path toward end-to-end, data-driven SYNTAX estimation with practical implications for CAD decision-making.

Abstract

The SYNTAX score has become a widely used measure of coronary disease severity, crucial in selecting the optimal mode of the revascularization procedure. This paper introduces a new medical regression and classification problem - automatically estimating SYNTAX score from coronary angiography. Our study presents a comprehensive CardioSYNTAX dataset of 3,018 patients for the SYNTAX score estimation and coronary dominance classification. The dataset features a balanced distribution of individuals with zero and non-zero scores. This dataset includes a first-of-its-kind, complete coronary angiography samples captured through a multi-view X-ray video, allowing one to observe coronary arteries from multiple perspectives. Furthermore, we present a novel, fully automatic end-to-end method for estimating the SYNTAX. For such a difficult task, we have achieved a solid coefficient of determination R2 of 0.51 in score value prediction and 77.3% accuracy for zero score classification.

CardioSyntax: end-to-end SYNTAX score prediction -- dataset, benchmark and method

TL;DR

CardioSYNTAX tackles automatic SYNTAX score estimation from invasive coronary angiography by introducing a public, multi-view angiography dataset of 3,018 patients and an end-to-end Multi-View Look (MVL) framework that fuses 3D video features from multiple views with separate RCA/LCA heads to predict the total score. The approach achieves an of for score regression and zero-score accuracy, while demonstrating domain-shift robustness and strong coronary dominance classification performance. The work also quantifies inter-expert disagreement (around ) and provides a domain-shift benchmark, highlighting challenges such as score distribution effects and cross-system generalization. Overall, CardioSYNTAX establishes a public benchmark and a viable path toward end-to-end, data-driven SYNTAX estimation with practical implications for CAD decision-making.

Abstract

The SYNTAX score has become a widely used measure of coronary disease severity, crucial in selecting the optimal mode of the revascularization procedure. This paper introduces a new medical regression and classification problem - automatically estimating SYNTAX score from coronary angiography. Our study presents a comprehensive CardioSYNTAX dataset of 3,018 patients for the SYNTAX score estimation and coronary dominance classification. The dataset features a balanced distribution of individuals with zero and non-zero scores. This dataset includes a first-of-its-kind, complete coronary angiography samples captured through a multi-view X-ray video, allowing one to observe coronary arteries from multiple perspectives. Furthermore, we present a novel, fully automatic end-to-end method for estimating the SYNTAX. For such a difficult task, we have achieved a solid coefficient of determination R2 of 0.51 in score value prediction and 77.3% accuracy for zero score classification.
Paper Structure (12 sections, 6 figures, 7 tables)

This paper contains 12 sections, 6 figures, 7 tables.

Table of Contents

  1. Introduction
  2. Related works
  3. SYNTAX datasets
  4. SYNTAX datasets
  5. CardioSYNTAX dataset
  6. Data labeling
  7. Multi-view look method
  8. Models and training details for SYNTAX score estimation
  9. Ablation study
  10. Results
  11. Discussion
  12. Conclusions

Figures (6)

  • Figure 1: SYNTAX score evaluation scheme. (a) Weighting factors for artery branches farooq2013widening. (b) Constituent factors of SYNTAX score.
  • Figure 2: An example of anomalies in right coronary artery. (a) Stenosis. (b) Bifurcation. (c) Calcific stenosis. (d) Total occlusion.
  • Figure 3: CardioSYNTAX dataset structure.
  • Figure 4: Proposed end-to-end SYNTAX score estimation scheme.
  • Figure 5: Example of RMSE loss (Log of SYNTAX score)
  • ...and 1 more figures