Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

3D CT-Based Coronary Calcium Assessment: A Feature-Driven Machine Learning Framework

Ayman Abaid, Gianpiero Guidone, Sara Alsubai, Foziyah Alquahtani, Talha Iqbal, Ruth Sharif, Hesham Elzomor, Emiliano Bianchini, Naeif Almagal, Michael G. Madden, Faisal Sharif, Ihsan Ullah

TL;DR

This work tackles automated coronary artery calcium (CAC) classification from non-contrast CCTA under limited annotations. It introduces a radiomics-based pipeline that uses pseudo-labeling via TotalSegmentator for ROI-centric feature extraction, and compares it against foundation-model embeddings from CT-FM and RadImageNet, using classical classifiers in a binary zero-versus-non-zero CAC task. On a clinical dataset of 182 CAC-scored patients, radiomics features achieved the strongest performance (up to $84\%$ accuracy, $0.95$ sensitivity), with statistically significant improvements over deep learning embeddings ($p<0.05$). The results support radiomics as a robust, interpretable, data-efficient approach for early CAD risk stratification in settings with limited annotations, and point to future directions in multimodal fusion and multi-class calcium scoring.

Abstract

Coronary artery calcium (CAC) scoring plays a crucial role in the early detection and risk stratification of coronary artery disease (CAD). In this study, we focus on non-contrast coronary computed tomography angiography (CCTA) scans, which are commonly used for early calcification detection in clinical settings. To address the challenge of limited annotated data, we propose a radiomics-based pipeline that leverages pseudo-labeling to generate training labels, thereby eliminating the need for expert-defined segmentations. Additionally, we explore the use of pretrained foundation models, specifically CT-FM and RadImageNet, to extract image features, which are then used with traditional classifiers. We compare the performance of these deep learning features with that of radiomics features. Evaluation is conducted on a clinical CCTA dataset comprising 182 patients, where individuals are classified into two groups: zero versus non-zero calcium scores. We further investigate the impact of training on non-contrast datasets versus combined contrast and non-contrast datasets, with testing performed only on non contrast scans. Results show that radiomics-based models significantly outperform CNN-derived embeddings from foundation models (achieving 84% accuracy and p<0.05), despite the unavailability of expert annotations.

3D CT-Based Coronary Calcium Assessment: A Feature-Driven Machine Learning Framework

TL;DR

This work tackles automated coronary artery calcium (CAC) classification from non-contrast CCTA under limited annotations. It introduces a radiomics-based pipeline that uses pseudo-labeling via TotalSegmentator for ROI-centric feature extraction, and compares it against foundation-model embeddings from CT-FM and RadImageNet, using classical classifiers in a binary zero-versus-non-zero CAC task. On a clinical dataset of 182 CAC-scored patients, radiomics features achieved the strongest performance (up to accuracy, sensitivity), with statistically significant improvements over deep learning embeddings (). The results support radiomics as a robust, interpretable, data-efficient approach for early CAD risk stratification in settings with limited annotations, and point to future directions in multimodal fusion and multi-class calcium scoring.

Abstract

Coronary artery calcium (CAC) scoring plays a crucial role in the early detection and risk stratification of coronary artery disease (CAD). In this study, we focus on non-contrast coronary computed tomography angiography (CCTA) scans, which are commonly used for early calcification detection in clinical settings. To address the challenge of limited annotated data, we propose a radiomics-based pipeline that leverages pseudo-labeling to generate training labels, thereby eliminating the need for expert-defined segmentations. Additionally, we explore the use of pretrained foundation models, specifically CT-FM and RadImageNet, to extract image features, which are then used with traditional classifiers. We compare the performance of these deep learning features with that of radiomics features. Evaluation is conducted on a clinical CCTA dataset comprising 182 patients, where individuals are classified into two groups: zero versus non-zero calcium scores. We further investigate the impact of training on non-contrast datasets versus combined contrast and non-contrast datasets, with testing performed only on non contrast scans. Results show that radiomics-based models significantly outperform CNN-derived embeddings from foundation models (achieving 84% accuracy and p<0.05), despite the unavailability of expert annotations.

Paper Structure

This paper contains 15 sections, 2 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Dataset Description
  3. Methodology
  4. Feature Extraction
  5. Radiomics:
  6. CT-FM:
  7. RadImageNet:
  8. Prediction Models
  9. Experiments and Results
  10. Preprocessing and Setup
  11. Results
  12. Conclusion
  13. Prospect of application
  14. Acknowledgments.
  15. Disclosure of Interests.

Figures (2)

  • Figure 1: Examples of axial slices from (a) contrast-enhanced and (b) non-contrast CT volumes.
  • Figure 2: Overview of the radiomics-based feature extraction pipeline.