Machine learning prediction of obstructive coronary artery disease using opportunistic coronary calcium and epicardial fat assessments from CT calcium scoring scans

Ammar Hoori; David E. Newby; David L. Wilson; Juhwan Lee; Justin N. Kim; Michelle C. Williams; Mohamed H. E. Makhlouf; Robert Gilkeson; Sanjay Rajagopalan; Tao Hu

arxiv: 2605.21762 · v1 · pith:NGF4KCPDnew · submitted 2026-05-20 · 💻 cs.LG

Machine learning prediction of obstructive coronary artery disease using opportunistic coronary calcium and epicardial fat assessments from CT calcium scoring scans

Juhwan Lee , Ammar Hoori , Tao Hu , Justin N. Kim , Mohamed H. E. Makhlouf , Michelle C. Williams , David E. Newby , Robert Gilkeson

show 2 more authors

Sanjay Rajagopalan David L. Wilson

This is my paper

Pith reviewed 2026-05-22 09:29 UTC · model grok-4.3

classification 💻 cs.LG

keywords machine learningcoronary artery diseaseCT calcium scoringepicardial fatobstructive CADfeature selectionCatBoostnon-contrast CT

0 comments

The pith

Machine learning using calcium and epicardial fat features from non-contrast CT calcium scans predicts obstructive coronary artery disease.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a machine learning framework that pulls quantitative details about coronary calcium deposits and surrounding epicardial fat from routine non-contrast CT calcium scoring images to forecast obstructive coronary artery disease. It begins with 424 candidate features drawn from 24 clinical variables plus hundreds of calcium-omics and fat-omics measurements in 1,324 patients, then uses CatBoost with SHAP to isolate 14 key predictors, the top two of which come from the fat data. The resulting model reaches 83 percent sensitivity, 94 percent specificity, and 85 percent accuracy while retaining performance even in zero-calcium-score cases that still show obstruction. A sympathetic reader sees this as a route to extract extra diagnostic information from low-cost, contrast-free scans and potentially limit the need for further imaging or invasive tests in lower-risk groups.

Core claim

The authors extract 189 calcium-omics and 211 epicardial fat-omics features plus 24 clinical variables from CTCS images of 1,324 SCOT-HEART patients. Using CatBoost with SHAP, they select 14 most predictive features, the top two from fat-omics. The CatBoost model then predicts obstructive CAD with sensitivity 83.1+/-4.6%, specificity 93.8+/-1.7%, accuracy 85.3+/-2.0%, and F1 73.9+/-3.3%. Performance holds across calcium score strata, including zero-calcium obstructive cases, and fat and calcium data add value beyond clinical variables alone.

What carries the argument

CatBoost gradient boosting combined with SHAP values to select and model 14 features from calcium-omics and epicardial fat-omics extracted from non-contrast CT calcium scoring scans.

If this is right

Inclusion of calcium-omics and fat-omics data significantly improves predictive performance over clinical variables alone.
The model maintains reliable accuracy across patients with varying coronary calcium scores, including those with obstructive disease despite a zero calcium score.
This approach may reduce reliance on contrast-enhanced CT angiography or invasive procedures for low-to-intermediate risk patients.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

Embedding these 14 features into standard calcium scoring reports could flag higher-risk patients during routine scans without extra imaging.
Re-training or fine-tuning the model on data from varied scanner vendors would likely be required before widespread clinical deployment.
The same style of omics feature extraction from non-contrast CT might be tested for predicting other cardiovascular events beyond obstruction.

Load-bearing premise

The 14 features chosen by CatBoost-SHAP and the resulting model performance will generalize to new patients and different CT scanners without major loss of accuracy.

What would settle it

Applying the trained model unchanged to CT calcium scoring scans from an independent external cohort and finding that sensitivity falls below 70 percent or specificity drops sharply.

read the original abstract

Non-contrast computed tomography calcium scoring (CTCS) is a cost-effective imaging modality widely used to detect coronary artery calcifications. This study aimed to develop an advanced machine learning framework that utilizes quantitative analyses of coronary calcium and epicardial fat from CTCS images to predict obstructive coronary artery disease (CAD). The study population consisted of 1,324 patients from the SCOT-HEART clinical trial who underwent both CTCS and coronary CT angiography. We extracted and analyzed a broad range of features, including 24 clinical variables, 189 calcium-omics, and 211 epicardial fat-omics features from the CTCS images. Feature selection was conducted using the CatBoost algorithm combined with SHapley Additive exPlanation (SHAP) values. Predictive modeling utilized the CatBoost gradient boosting method, focusing on the most informative features. From an initial set of 424 candidate features, 14 were identified as most predictive through the CatBoost-SHAP method. The top two predictive features originated from fat-omics, with the remaining 12 features derived from calcium-omics. The optimized model achieved robust predictive capabilities, demonstrating a sensitivity of 83.1+/-4.6%, specificity of 93.8+/-1.7%, accuracy of 85.3+/-2.0%, and an F1 score of 73.9+/-3.3%. Inclusion of calcium-omics and fat-omics data significantly improved predictive performance. Notably, the model also showed reliable predictive accuracy in patients with diverse coronary calcium scores, including cases with obstructive CAD despite a zero-calcium score. This innovative approach holds promise for improving clinical decision-making and potentially reducing dependence on contrast-enhanced or invasive diagnostic procedures, particularly within low-to intermediate-risk patient groups.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

Adding detailed epicardial fat and calcium features from routine CT calcium scans improves internal ML prediction of obstructive CAD on the SCOT-HEART cohort, but the results rest on untested generalization.

read the letter

This paper shows that you can get better predictions of obstructive coronary artery disease by pulling detailed calcium and epicardial fat measurements from routine non-contrast CT calcium scoring scans and feeding them into a CatBoost model. On the SCOT-HEART data they get sensitivity around 83% and accuracy 85% with the top 14 features, and the fat features turn out to be the strongest. What stands out is the scale of the feature set—over 400 candidates including 189 calcium-omics and 211 fat-omics—and the fact that the model still works in patients with zero calcium score. They demonstrate that including these omics features improves performance over clinical variables or calcium score alone, which is a practical step toward using existing scans more effectively. The main limitation is that all the results come from internal evaluation on the same 1,324 patients. Feature selection and model fitting happened within this cohort, so there's the usual risk that the selected features and performance numbers are tuned to this specific population and imaging setup. No external cohort or multi-site test is described, which makes it unclear how well this would hold up on different scanners or patient groups. Readers in cardiac imaging or machine learning applications in cardiology would find this relevant. It gives a concrete example of combining quantitative imaging features for risk stratification without needing contrast. The work is grounded enough in a real clinical trial dataset that it merits a full review rather than a desk reject. I would send it to peer review. Reviewers will likely focus on the validation approach and ask for more transparency on the exact splitting and any handling of class imbalance.

Referee Report

2 major / 3 minor

Summary. The manuscript develops a CatBoost gradient boosting model to predict obstructive coronary artery disease from non-contrast CT calcium scoring scans. Using 1,324 patients from the SCOT-HEART trial, the authors extract 424 features (24 clinical variables, 189 calcium-omics, 211 epicardial fat-omics), apply CatBoost-SHAP to select 14 top features (12 calcium-omics, 2 fat-omics), and report internal cross-validation performance of 83.1±4.6% sensitivity, 93.8±1.7% specificity, 85.3±2.0% accuracy, and 73.9±3.3% F1 score. They claim that adding omics features improves prediction over clinical variables alone and that the model remains reliable even in zero-calcium-score cases with obstructive disease.

Significance. If the internal performance generalizes, the work could support opportunistic use of routine CTCS scans for CAD risk stratification, potentially reducing reliance on contrast-enhanced CTA in low-to-intermediate risk groups. The emphasis on interpretable calcium- and fat-derived features plus SHAP analysis is a positive aspect, but the absence of external validation or multi-center testing substantially limits the strength of claims about clinical utility and robustness across scanners and populations.

major comments (2)

[Results] Results section (paragraph reporting the optimized model performance): the sensitivity, specificity, accuracy, and F1 metrics with ± variability estimates are presented after feature selection on the full SCOT-HEART cohort, yet no details are given on the train-test splitting procedure, whether feature selection was nested inside cross-validation, or confirmation that the reported numbers reflect performance on truly unseen data. This directly affects the reliability of the central performance claim.
[Methods] Methods (feature selection and predictive modeling subsections): the 14 features were identified via CatBoost-SHAP on the same 1,324-patient cohort used for final model evaluation. Without an independent external validation cohort or multi-center test set, the assumption that these calcium-omics and fat-omics features and decision boundaries will generalize to new scanners, protocols, or demographics remains untested and is load-bearing for the stated promise of improved clinical decision-making.

minor comments (3)

[Abstract] Abstract and Methods: the variability estimates (±4.6%, ±1.7%, etc.) are reported but the exact procedure (e.g., standard deviation across 5-fold or 10-fold CV) is not stated; please add this detail for reproducibility.
[Results] Results: the claim that inclusion of calcium-omics and fat-omics data 'significantly improved predictive performance' would be strengthened by an explicit comparison table or metrics for the clinical-variables-only baseline model.
[Discussion] Discussion: the prevalence of obstructive CAD in the cohort should be stated to help readers interpret the accuracy and F1 scores in light of possible class imbalance.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their detailed and constructive review of our manuscript. We address each of the major comments below and have revised the manuscript accordingly to improve clarity and transparency.

read point-by-point responses

Referee: [Results] Results section (paragraph reporting the optimized model performance): the sensitivity, specificity, accuracy, and F1 metrics with ± variability estimates are presented after feature selection on the full SCOT-HEART cohort, yet no details are given on the train-test splitting procedure, whether feature selection was nested inside cross-validation, or confirmation that the reported numbers reflect performance on truly unseen data. This directly affects the reliability of the central performance claim.

Authors: We appreciate this observation and agree that explicit details on the validation procedure are essential. Upon review, our analysis utilized a 5-fold stratified cross-validation framework. To address the concern, we have revised the Methods section to specify that feature selection via CatBoost-SHAP was nested within the cross-validation process, performed solely on the training portion of each fold. The performance metrics reported are the mean and standard deviation across the held-out test folds, ensuring evaluation on unseen data. We have also included a schematic diagram of the workflow in the revised manuscript. revision: yes
Referee: [Methods] Methods (feature selection and predictive modeling subsections): the 14 features were identified via CatBoost-SHAP on the same 1,324-patient cohort used for final model evaluation. Without an independent external validation cohort or multi-center test set, the assumption that these calcium-omics and fat-omics features and decision boundaries will generalize to new scanners, protocols, or demographics remains untested and is load-bearing for the stated promise of improved clinical decision-making.

Authors: We concur that external validation would provide stronger evidence for generalizability. The present work focuses on the SCOT-HEART cohort as a proof-of-concept using internal validation. In the revised manuscript, we have added a dedicated paragraph in the Discussion section acknowledging this limitation and outlining plans for future external validation studies. We maintain that the SHAP-based feature selection and the observed performance improvements over clinical variables alone offer valuable insights, but we have tempered the claims regarding immediate clinical utility pending further validation. revision: partial

Circularity Check

0 steps flagged

No circularity: standard empirical ML pipeline with internal validation

full rationale

The paper applies CatBoost-SHAP feature selection to 424 candidates drawn from the SCOT-HEART cohort, retains 14 features, trains a CatBoost model, and reports cross-validated metrics (sensitivity 83.1+/-4.6%, accuracy 85.3+/-2.0%). No equation, derivation, or result is shown to equal its own inputs by construction. Feature selection and model fitting are explicit data-driven steps whose outputs are evaluated on the same distribution via internal CV; this is ordinary supervised learning practice rather than a self-definitional loop, fitted parameter renamed as prediction, or load-bearing self-citation. The manuscript contains no first-principles derivation, uniqueness theorem, or ansatz smuggled via prior work that would trigger any of the enumerated circularity patterns. The central claim therefore remains an empirical performance report on the described cohort and does not reduce to its inputs by definition.

Axiom & Free-Parameter Ledger

2 free parameters · 2 axioms · 0 invented entities

The central claim rests on standard machine learning assumptions plus properties of the SCOT-HEART clinical dataset. No new physical entities are postulated.

free parameters (2)

CatBoost hyperparameters
Model hyperparameters for gradient boosting are selected or tuned and directly affect reported performance.
Top feature count
Choice to retain exactly 14 features after SHAP ranking from 424 candidates.

axioms (2)

domain assumption Quantitative omics features extracted from CTCS images accurately capture coronary calcium and epicardial fat properties relevant to obstructive CAD.
Invoked when defining the 189 calcium-omics and 211 fat-omics features used for modeling.
domain assumption The 1,324 SCOT-HEART patients form a representative sample for training and evaluating a generalizable predictive model.
Underlying assumption for all performance claims.

pith-pipeline@v0.9.0 · 5906 in / 1614 out tokens · 75379 ms · 2026-05-22T09:29:25.996644+00:00 · methodology

discussion (0)

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear

?

unclear
Relation between the paper passage and the cited Recognition theorem.

From an initial set of 424 candidate features, 14 were identified as most predictive through the CatBoost-SHAP method... sensitivity of 83.1+/-4.6%, specificity of 93.8+/-1.7%

What do these tags mean?

matches: The paper's claim is directly supported by a theorem in the formal canon.
supports: The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends: The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses: The paper appears to rely on the theorem as machinery.
contradicts: The paper's claim conflicts with a theorem or certificate in the canon.
unclear: Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

28 extracted references · 28 canonical work pages

[1]

Cury, R. C. et al. CAD-RADSTM 2.0 – 2022 Coronary Artery Disease – Reporting and Data System An Expert Consensus Document of the Society of Cardiovascular Computed Tomography (SCCT), the American College of Cardiology (ACC), the Ameri can College of Radiology (ACR) and the North America Society of Cardiovascular Imaging (NASCI). Radiology: Cardiothoracic ...

work page 2022
[2]

Nam, K. et al. Prognostic value of coronary artery disease-reporting and data system (CAD-RADS) score for cardiovascular events in ischemic stroke. Atherosclerosis 287, 1–7 (2019)

work page 2019
[3]

Bittner, D. O. et al. Prognostic Value of Coronary CTA in Stable Chest Pain. JACC: Cardiovascular Imaging 13, 1534–1545 (2020)

work page 2020
[4]

Williams, M. C. et al. Standardized reporting systems for computed tomography coronary angiography and calcium scoring: A real-world validation of CAD-RADS and CAC-DRS in patients with stable chest pain. Journal of Cardiovascular Computed Tomography 14, 3–11 (2020)

work page 2020
[5]

Orringer, C. E. et al. The National Lipid Association scientific statement on coronary artery calcium scoring to guide preventive strategies for ASCVD risk reduction. Journal of Clinical Lipidology 15, 33–60 (2021)

work page 2021
[6]

Golub, I. S. et al. Major Global Coronary Artery Calcium Guidelines. JACC: Cardiovascular Imaging 16, 98– 117 (2023)

work page 2023
[7]

Lee, J. et al. Prediction of obstructive coronary artery disease using coronary calcification and epicardial adipose tissue assessments from CT calcium scoring scans. Journal of Cardiovascular Computed Tomography 19, 224–231 (2025)

work page 2025
[8]

Hoori, A. et al. Enhancing cardiovascular risk prediction through AI-enabled calcium-omics. Sci Rep 14, 11134 (2024)

work page 2024
[9]

Hu, T. et al. AI prediction of cardiovascular events using o pportunistic epicardial adipose tissue assessments from CT calcium score. ArXiv arXiv:2401.16190v1 (2024)

work page arXiv 2024
[10]

Williams Michelle C. et al. Low-Attenuation Noncalcified Plaque on Coronary Computed Tomography Angiography Predicts Myocardial Infarction. Circulation 141, 1452–1462 (2020)

work page 2020
[11]

Williams, M. C. et al. Coronary Artery Plaque Characteristics As sociated With Adverse Outcomes in the SCOT-HEART Study. J Am Coll Cardiol 73, 291–301 (2019)

work page 2019
[12]

Gulati, M. et al. 2021 AHA/ACC/ASE/CHEST/SAEM/SCCT/S CMR Guideline for the Evaluation and Diagnosis of Chest Pain: A Report of the American Co llege of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. Circulation 144, e368–e454 (2021)

work page 2021
[13]

Knuuti, J. et al. 2019 ESC Guidelines for the diagnosis and management of chronic coronary syndromes: The Task Force for the diagnosis and management of chronic coronary syndromes of the European Society of Cardiology (ESC). European Heart Journal 41, 407–477 (2020)

work page 2019
[14]

Hoori, A. et al. Deep learning segmentation and quantification method for assessing epicardial adipose tissue in CT calcium score scans. Sci Rep 12, 2276 (2022)

work page 2022
[15]

Prokhorenkova, L., Gusev, G., Vorobev, A., Dorogush, A. V. & Gulin, A. CatBoost: unbiased boosting with categorical features. in Advances in Neural Information Processing Systems vol. 31 (Curran Associates, Inc., 2018)

work page 2018
[16]

Lundberg, S. M. & Lee, S.-I. A unified approach to interpreting model predictions. in Proceedings of the 31st International Conference on Neural Information Processing Systems 4768–4777 (Curran Associates Inc., Red Hook, NY, USA, 2017)

work page 2017
[17]

Singh, P. et al. Leveraging calcium score CT radiomic s for heart failure risk prediction. Sci Rep 14, 26898 (2024)

work page 2024
[18]

Hu, T. et al. Artificial Intelligence Prediction of Cardiovascular Events Using Opportunistic Epicardial Adipose Tissue Assessments From Computed Tomography Calcium Score. JACC: Advances 3, 101188 (2024)

work page 2024
[19]

Lee, J. et al. Detection of arterial remodeling using epicardial adipose tissue assessment from CT calcium scoring scan. Front Cardiovasc Med 12, 1543816 (2025)

work page 2025
[20]

Song, Y. et al. Pericoronary adipose tissue feature analysis in computed tomography ca lcium score images in comparison to coronary computed tomography angiography. JMI 12, 014503 (2025). 9

work page 2025
[21]

Kim, J. N. et al. Improving coronary artery segmentation with self-supervised learning and automated pericoronary adipose tissue segmentation: a multi-institutional study on coronary computed tomography angiography images. J Med Imaging (Bellingham) 12, 016002 (2025)

work page 2025
[22]

Wu, H. et al. Cardiac CT Perfusion Imaging of Pericoronary Adipose Tissue (PCAT) Highlighting Potential Confounds in CTA Analysis. Journal of Clinical Medicine 14, 769 (2025)

work page 2025
[23]

Goeller, M. et al. Epicardial adipose tissue density and volume are related to subclinical atherosclerosis, inflammation and major adverse cardiac events in asymptomatic subjects. Journal of Cardiovascular Computed Tomography 12, 67–73 (2018)

work page 2018
[24]

McLaughlin, T. et al. Relationship Between Coronary Atheroma, Epicardial Adipose Tissue Inflammation, and Adipocyte Differentiation Across the Human Myocardial Bridge. Journal of the American Heart Association 10, e021003 (2021)

work page 2021
[25]

Cosson, E. et al. Epicardial adipose tissue volume and coronary calcification among people living with diabetes: a cross-sectional study. Cardiovascular Diabetology 20, 35 (2021)

work page 2021
[26]

Agha, A. M. et al. The Prognostic Value of CAC Zero Among Individuals Presenting With Chest Pain. JACC: Cardiovascular Imaging 15, 1745–1757 (2022)

work page 2022
[27]

Winther, S. et al. Coronary Calcium Scoring Improves Risk Pred iction in Patients With Suspected Obstructive Coronary Artery Disease. J Am Coll Cardiol 80, 1965–1977 (2022)

work page 1965
[28]

P., Lakshmanan, S., Lichtenstein, S

Sheppard, J. P., Lakshmanan, S., Lichtenstein, S. J., Budoff, M. J. & Roy, S. K. Age and the power of zero CAC in cardiac risk assessment: overview of the literature and a cautionary case. Br J Cardiol 29, 23 (2022). 10 Tables Table 1 Baseline clinical characteristics of patients with obstructive (n=334) and non-obstructive CAD (n=990) (BMI: body mass ind...

work page 2022

[1] [1]

Cury, R. C. et al. CAD-RADSTM 2.0 – 2022 Coronary Artery Disease – Reporting and Data System An Expert Consensus Document of the Society of Cardiovascular Computed Tomography (SCCT), the American College of Cardiology (ACC), the Ameri can College of Radiology (ACR) and the North America Society of Cardiovascular Imaging (NASCI). Radiology: Cardiothoracic ...

work page 2022

[2] [2]

Nam, K. et al. Prognostic value of coronary artery disease-reporting and data system (CAD-RADS) score for cardiovascular events in ischemic stroke. Atherosclerosis 287, 1–7 (2019)

work page 2019

[3] [3]

Bittner, D. O. et al. Prognostic Value of Coronary CTA in Stable Chest Pain. JACC: Cardiovascular Imaging 13, 1534–1545 (2020)

work page 2020

[4] [4]

Williams, M. C. et al. Standardized reporting systems for computed tomography coronary angiography and calcium scoring: A real-world validation of CAD-RADS and CAC-DRS in patients with stable chest pain. Journal of Cardiovascular Computed Tomography 14, 3–11 (2020)

work page 2020

[5] [5]

Orringer, C. E. et al. The National Lipid Association scientific statement on coronary artery calcium scoring to guide preventive strategies for ASCVD risk reduction. Journal of Clinical Lipidology 15, 33–60 (2021)

work page 2021

[6] [6]

Golub, I. S. et al. Major Global Coronary Artery Calcium Guidelines. JACC: Cardiovascular Imaging 16, 98– 117 (2023)

work page 2023

[7] [7]

Lee, J. et al. Prediction of obstructive coronary artery disease using coronary calcification and epicardial adipose tissue assessments from CT calcium scoring scans. Journal of Cardiovascular Computed Tomography 19, 224–231 (2025)

work page 2025

[8] [8]

Hoori, A. et al. Enhancing cardiovascular risk prediction through AI-enabled calcium-omics. Sci Rep 14, 11134 (2024)

work page 2024

[9] [9]

Hu, T. et al. AI prediction of cardiovascular events using o pportunistic epicardial adipose tissue assessments from CT calcium score. ArXiv arXiv:2401.16190v1 (2024)

work page arXiv 2024

[10] [10]

Williams Michelle C. et al. Low-Attenuation Noncalcified Plaque on Coronary Computed Tomography Angiography Predicts Myocardial Infarction. Circulation 141, 1452–1462 (2020)

work page 2020

[11] [11]

Williams, M. C. et al. Coronary Artery Plaque Characteristics As sociated With Adverse Outcomes in the SCOT-HEART Study. J Am Coll Cardiol 73, 291–301 (2019)

work page 2019

[12] [12]

Gulati, M. et al. 2021 AHA/ACC/ASE/CHEST/SAEM/SCCT/S CMR Guideline for the Evaluation and Diagnosis of Chest Pain: A Report of the American Co llege of Cardiology/American Heart Association Joint Committee on Clinical Practice Guidelines. Circulation 144, e368–e454 (2021)

work page 2021

[13] [13]

Knuuti, J. et al. 2019 ESC Guidelines for the diagnosis and management of chronic coronary syndromes: The Task Force for the diagnosis and management of chronic coronary syndromes of the European Society of Cardiology (ESC). European Heart Journal 41, 407–477 (2020)

work page 2019

[14] [14]

Hoori, A. et al. Deep learning segmentation and quantification method for assessing epicardial adipose tissue in CT calcium score scans. Sci Rep 12, 2276 (2022)

work page 2022

[15] [15]

Prokhorenkova, L., Gusev, G., Vorobev, A., Dorogush, A. V. & Gulin, A. CatBoost: unbiased boosting with categorical features. in Advances in Neural Information Processing Systems vol. 31 (Curran Associates, Inc., 2018)

work page 2018

[16] [16]

Lundberg, S. M. & Lee, S.-I. A unified approach to interpreting model predictions. in Proceedings of the 31st International Conference on Neural Information Processing Systems 4768–4777 (Curran Associates Inc., Red Hook, NY, USA, 2017)

work page 2017

[17] [17]

Singh, P. et al. Leveraging calcium score CT radiomic s for heart failure risk prediction. Sci Rep 14, 26898 (2024)

work page 2024

[18] [18]

Hu, T. et al. Artificial Intelligence Prediction of Cardiovascular Events Using Opportunistic Epicardial Adipose Tissue Assessments From Computed Tomography Calcium Score. JACC: Advances 3, 101188 (2024)

work page 2024

[19] [19]

Lee, J. et al. Detection of arterial remodeling using epicardial adipose tissue assessment from CT calcium scoring scan. Front Cardiovasc Med 12, 1543816 (2025)

work page 2025

[20] [20]

Song, Y. et al. Pericoronary adipose tissue feature analysis in computed tomography ca lcium score images in comparison to coronary computed tomography angiography. JMI 12, 014503 (2025). 9

work page 2025

[21] [21]

Kim, J. N. et al. Improving coronary artery segmentation with self-supervised learning and automated pericoronary adipose tissue segmentation: a multi-institutional study on coronary computed tomography angiography images. J Med Imaging (Bellingham) 12, 016002 (2025)

work page 2025

[22] [22]

Wu, H. et al. Cardiac CT Perfusion Imaging of Pericoronary Adipose Tissue (PCAT) Highlighting Potential Confounds in CTA Analysis. Journal of Clinical Medicine 14, 769 (2025)

work page 2025

[23] [23]

Goeller, M. et al. Epicardial adipose tissue density and volume are related to subclinical atherosclerosis, inflammation and major adverse cardiac events in asymptomatic subjects. Journal of Cardiovascular Computed Tomography 12, 67–73 (2018)

work page 2018

[24] [24]

McLaughlin, T. et al. Relationship Between Coronary Atheroma, Epicardial Adipose Tissue Inflammation, and Adipocyte Differentiation Across the Human Myocardial Bridge. Journal of the American Heart Association 10, e021003 (2021)

work page 2021

[25] [25]

Cosson, E. et al. Epicardial adipose tissue volume and coronary calcification among people living with diabetes: a cross-sectional study. Cardiovascular Diabetology 20, 35 (2021)

work page 2021

[26] [26]

Agha, A. M. et al. The Prognostic Value of CAC Zero Among Individuals Presenting With Chest Pain. JACC: Cardiovascular Imaging 15, 1745–1757 (2022)

work page 2022

[27] [27]

Winther, S. et al. Coronary Calcium Scoring Improves Risk Pred iction in Patients With Suspected Obstructive Coronary Artery Disease. J Am Coll Cardiol 80, 1965–1977 (2022)

work page 1965

[28] [28]

P., Lakshmanan, S., Lichtenstein, S

Sheppard, J. P., Lakshmanan, S., Lichtenstein, S. J., Budoff, M. J. & Roy, S. K. Age and the power of zero CAC in cardiac risk assessment: overview of the literature and a cautionary case. Br J Cardiol 29, 23 (2022). 10 Tables Table 1 Baseline clinical characteristics of patients with obstructive (n=334) and non-obstructive CAD (n=990) (BMI: body mass ind...

work page 2022