Skip to content
Open access

Machine Learning Multiclass Models Based on Cardiac Segmentation Quantitative Parameters from Non-Contrast and Contrast-Enhanced Chest CT for the Differential Diagnosis of Acute Chest Pain

Nov 2026 · Medicina · Vol 62, pp. 2212 · 0 citations · 27 references

TL;DR

Machine learning models built on automatically segmented cardiac parameters showed moderate discriminatory potential for differentiating ACS, PE, and AD, and the contrast-enhanced model outperformed the non-contrast model, whereas non-contrast CT still carried complementary structural information that may support risk stratification before contrast administration.

Abstract

Objectives: Acute coronary syndrome (ACS), pulmonary embolism (PE), and aortic dissection (AD) present with overlapping symptoms but require markedly different treatments. We investigated whether cardiac quantitative parameters obtained by automated segmentation of non-contrast and contrast-enhanced chest CT can support machine learning-based multiclass differentiation of these conditions. Materials and Methods: This single-center retrospective study included 628 patients who underwent triple rule-out CT angiography (147 negative controls, 193 ACS, 147 PE, and 141 AD). Twenty-six features were extracted separately from automated cardiac segmentation of non-contrast and contrast-enhanced images. Five classifiers—multivariable logistic regression, random forest, XGBoost, LightGBM, and support vector machine—were developed with a 7:3 stratified train–test split and five-fold cross-validation. Performance was evaluated by accuracy, macro-averaged F1 score, and macro-averaged area under the receiver operating characteristic curve (AUC), and feature importance was assessed with SHapley Additive exPlanations (SHAP). Results: Multivariable logistic regression showed the most balanced performance for both modalities. The non-contrast model achieved an accuracy of 0.635, a macro-averaged F1 score of 0.637, and a macro-averaged AUC of 0.840, with class-specific AUCs of 0.814 for ACS, 0.878 for PE, and 0.887 for AD. The contrast-enhanced model achieved 0.683, 0.673, and 0.861, with AUCs of 0.881, 0.876, and 0.919. Recall for ACS increased from 58.6% to 70.7% and for AD from 74.4% to 83.7%, while AD cases misclassified as negative decreased from 11.6% to 2.3%. Left ventricular mass index and right-heart parameters ranked among the most contributory features in both models. Conclusion: Machine learning models built on automatically segmented cardiac parameters showed moderate discriminatory potential for differentiating ACS, PE, and AD. The contrast-enhanced model outperformed the non-contrast model, whereas non-contrast CT still carried complementary structural information that may support risk stratification before contrast administration.

Read PDF

Similar papers

Open access Nov 2026

Application of abdominal CT radiomic features for discriminating prevalent coronary heart disease in patients with type 2 diabetes mellitus

To develop and validate a hybrid model combining pancreatic-aortic radiomic features from non-contrast abdominal CT with clinical factors for discriminating prevalent coronary heart disease (CHD) status in patients with type 2 diabetes mellitus (T2DM). This retrospective study enrolled 407 T2DM patients (277...

H. Sui, Ying-Chun Wan, Ming Qu et al. · 0 citations
Open access Nov 2026

An End-to-End Cardiac MRI Segmentation and Multi-Class Pathology Classification Pipeline Using Ensemble U-Net Models and Explainable Machine Learning

Accurate delineation of cardiac structures in cardiac magnetic resonance (CMR) images is a prerequisite for quantifying cardiac function. This paper presents an end-to-end pipeline for automated cardiac segmentation and five-class pathology classification, evaluated on the 150 patients of the Automated Cardiac Diagnosi...

Alina Georgiana Badea, I. Tache, Lidia Băjenaru et al. · 0 citations
Open access Nov 2026

Development and internal validation of an interpretable machine learning model incorporating LVEF and SII for Predicting complex coronary lesions in patients with type 2 diabetes mellitus and coronary heart disease

Noninvasive preoperative risk stratification for complex coronary lesions remains difficult for people living with type 2 diabetes (T2DM) and coronary heart disease (CHD). While left ventricular ejection fraction (LVEF) and the systemic immune-inflammation index (SII) have each been associated with such lesions,...

Xiao-Fang Ma, M. Wang, Jin Zhang et al. · 0 citations
Open access Oct 2026

Machine learning CT-FFR for physiologic decision-making in diffuse coronary artery disease

Diffuse coronary artery disease (CAD) is characterized by gradual, long-segment pressure loss, which may not be fully captured by anatomy-based stenosis severity alone. Computed tomography-derived fractional flow reserve (CT-FFR) provides noninvasive physiological assessment, but its performance in diffuse diseas...

Z. An, Z. Ma, M. Zhang et al. · 0 citations
Sep 2026

Atherosclerosis diagnosis from magnetic resonance images: an empirical approach to automate carotid vessel wall segmentation and feature-based prediction.

The proposed hybrid DL-ML pipeline provides an accurate, interpretable, and clinically feasible solution for automated carotid atherosclerosis diagnosis and demonstrates that improved segmentation significantly enhances classification performance.

S. B. Baray, R. Sarmun, M. Islam et al. · 0 citations
Open access Oct 2026

An interpretable machine learning risk stratification model for the imaging burden of cerebral small vessel disease: integrating echocardiography, carotid ultrasound, and serological indicators

This study enrolled 401 subjects, who were stratified according to the total imaging burden score of Cerebral Small Vessel Disease (CSVD) and divided into a training set (280 cases) and a validation set (121 cases) at a ratio of 7:3. Six variables—age, left ventricular posterior wall end-diastolic thickness, E/e′ ratio...

H. Tan, Hou-Wei Li, Chu-Lin Deng et al. · 0 citations

We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.