Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2025 Apr 14;7(7):936-947.
doi: 10.1016/j.cjco.2025.03.024. eCollection 2025 Jul.

Prediction of Atrial Fibrillation Using Radiomic Features of Left Atrial Epicardial Adipose Tissue on Noncontrast Cardiac Computed Tomography

Shayna Cohen-Dor  1   2 , Moshe Rav-Acha  1   2 , Fauzi Shaheen  1 , Boris Chutko  1 , Hadas Labrisch-Kaye  2 , Zohar Ben-Haim  1 , Yoav Michowitz  1   2 , Hilla Gérard  1 , Naama Bogot  3 , Shemi Carraso  1   2 , Itzhak Vitkon-Barkay  4 , Laurian Copel  5 , Michael Glikson  1   2 , Arik Wolak  1   2
Affiliations

Prediction of Atrial Fibrillation Using Radiomic Features of Left Atrial Epicardial Adipose Tissue on Noncontrast Cardiac Computed Tomography

Shayna Cohen-Dor et al. CJC Open. .

Abstract
in English, French

Background: Early detection of atrial fibrillation (AF) can prevent AF-related complications. Radiomic analysis of epicardial adipose tissue (EAT) was shown to predict AF recurrence postablation, but only limited data exist regarding left atrial EAT (LA-EAT) radiomic analysis for predicting AF in patients with yet unknown AF. Our aim was to develop prediction model for AF, based on the association of machine learning-based radiomic analysis of LA-EAT and AF.

Methods: Retrospective matched case-control study of patients with and without AF, undergoing noncontrast electrocardiographic (ECG)-gated cardiac computed tomography (CT). Segmentation of LA-EAT and extraction of LA-EAT radiomic features were performed using syngo.via Frontier (Siemens Healthineers, Forchheim, Germany). Univariate analysis identified radiomic features associated with AF. Predictive models for AF were developed via logistic regression and machine learning-based random forest analyses. Models were validated on external cohort of patients with 1:1 AF : control ratio and deployed in a real-world setting with an AF : control ratio of 15:85.

Results: The study included 280 patients, 120 with documented AF and 160 matched controls. Based on LA-EAT radiomic features, which were significantly associated with AF, logistic regression and random forest models were constructed and tested on separate internal cohort of patients, yielding area under the curve (AUC) of 0.88 and 0.86, respectively, for prediction of AF. External validation verified these results (AUC 0.84 and 0.78, respectively). Both models were further validated in a real-world setting cohort (AUC 0.85 and 0.81, respectively).

Conclusions: Models, based on LA-EAT radiomic features extracted from noncontrast ECG-gated cardiac CT, could accurately predict AF, suggesting a potential widespread noninvasive method for predicting the presence of AF.

Clinical registration number: 0281-23-ASF.

Contexte: La détection précoce de la fibrillation auriculaire (FA) peut prévenir les complications qui y sont associées. L’analyse radiomique du tissu adipeux épicardique (TAE) a déjà démontré son utilité pour prédire la récidive de FA après ablation, mais il n'existe que peu de données concernant l'analyse radiomique du tissu adipeux de l'oreillette gauche (TAE-OG) pour prédire la FA chez les patients sans diagnostic préalable de FA. Notre objectif était de développer un modèle de prédiction de la FA, basé sur l'association de la FA avec l'analyse radiomique du TAE-OG assistée par apprentissage automatique.

Méthodologie: Étude rétrospective cas-témoins appariés incluant des patients avec et sans FA, soumis à une tomodensitométrie cardiaque sans contraste, avec synchronisation ECG. La segmentation du TAE-OG et l'extraction des caractéristiques radiomiques du TAE-OG ont été réalisées à l'aide du logiciel syngo.via Frontier (Siemens Healthineers, Forchheim, Allemagne). Une analyse univariée a permis d'identifier les caractéristiques radiomiques associées à la FA. Des modèles prédictifs de la FA ont été développés par régression logistique et par des analyses de forêt aléatoire basées sur l'apprentissage automatique. Les modèles ont été validés sur une cohorte externe de patients avec un ratio AF:contrôle de 1:1 et déployés dans un environnement réel avec un ratio AF:contrôle de 15:85.

Résultats: L'étude a porté sur 280 patients, dont 120 avec une FA documentée et 160 témoins appariés. Sur la base des caractéristiques radiomiques du TAE-OG qui étaient significativement associées à la FA, des modèles de régression logistique et par forêt aléatoire ont été construits et testés sur une cohorte interne de patients distincte, produisant une aire sous la courbe (ASC) de 0,88 et 0,86, respectivement, pour la prédiction de la FA. Une validation externe a confirmé ces résultats (ASC de 0,84 et 0,78, respectivement). Les deux modèles ont été validés dans une cohorte de patients en conditions réelles (AUC de 0,85 et 0,81, respectivement).

Conclusions: Les modèles, basés sur les caractéristiques radiomiques du TAE-OG extraites de la tomodensitométrie cardiaque sans contraste, avec synchronisation ECG, pouvaient prédire avec précision la FA, ce qui suggère une méthode non invasive potentiellement applicable à grande échelle pour prédire la présence de FA.

PubMed Disclaimer

Figures

Figure 1
Figure 1
Flow chart demonstrating the process of training, testing, external validation, and verifying results on a clinical simulation process. Initially, we trained the algorithm on 120 patients from Shaare Zedek Medical Center (SZMC). Following this, we tested the algorithm on an additional separate cohort of 60 patients from SZMC. Subsequently, an external validation of the same algorithm was performed on an additional 60 patients from Shamir Medical Center. In the final stage, 60 patients were taken from the testing and external validation cohorts, along with 40 new patients, culminating in a comprehensive analysis.
Figure 2
Figure 2
Flow process of left atrium epicardial adipose tissue (LA-EAT) segmentation using syngo.via Frontier (Siemens Healthineers, Forchheim, Germany). The left atrium (LA) is visible in all panels. (A) Sagittal cardiac computed tomography view with epicardial adipose tissue (EAT) automatically detected in red using the cardiac risk analysis tool. (B) Green mask from (A) imported into the radiomic tool. (C) Yellow representation of EAT detected using a –150 to –50 HU threshold within the mask borders from (B). (D, E) Blue landmarks segmenting the area of interest around the LA: (D) superior border marked by the main pulmonary artery (MPA), and (E) inferior border marked by the coronary sinus (CS). (F) Depiction of the adipose tissue segmented around the LA post-previous steps (yellow).
Figure 3
Figure 3
Ten top radiomic features as found in the univariate analysis of LA-EAT differentiating between patients with and without AF. The radiomic features are presented as a heat map: Each row corresponds to 1 radiomic feature presented on the right side of the map, and each column corresponds to 1 patient. The color keys of z scores are depicted below the heat map on the left. The letter in bold (X) represents the kind of radiomic feature: (M) = morphologic, (T) = texture, (FO) = first order. AF, atrial fibrillation; LA-EAT, left atrial-epicardial adipose tissue.
Figure 4
Figure 4
ROC curves of 2 models based on radiomic features from LA-EAT applied on internal testing and external validation cohorts. (A) AUC of logistic regression model on internal testing cohort. (B) AUC of random forest model on internal testing cohort. (C) AUC of logistic regression model on external validation cohort. (D) AUC of random forest model on external validation cohort. AUC, area under the curve; LA-EAT, left atrial-epicardial adipose tissue; ROC, receiver-operating characteristic.
Figure 5
Figure 5
ROC curves of 2 models based on radiomic features from LA-EAT applied on clinical setting model. (A) AUC of logistic regression model. (B) AUC of random forest model. AUC, area under the curve; LA-EAT, left atrial-epicardial adipose tissue; ROC, receiver-operating characteristic.
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. Staerk L., Wang B., Preis S.R., et al. Lifetime risk of atrial fibrillation according to optimal, borderline, or elevated levels of risk factors: cohort study based on longitudinal data from the Framingham Heart Study. BMJ. 2018;361 - PMC - PubMed
    1. Lloyd-Jones D.M., Wang T.J., Leip E.P., et al. Lifetime risk for development of atrial fibrillation: the Framingham Heart Study. Circulation. 2004;110:1042–1046. - PubMed
    1. Oladiran O., Nwosu I. Stroke risk stratification in atrial fibrillation: a review of common risk factors. J Community Hosp Intern Med Perspect. 2019;9:113–120. - PMC - PubMed
    1. Hindricks G., Potpara T., Dagres N., et al. 2020 ESC guidelines for the diagnosis and management of atrial fibrillation developed in collaboration with the European Association for Cardio-Thoracic Surgery (EACTS): the task force for the diagnosis and management of atrial fibrillation of the European Society of Cardiology (ESC). Developed with the special contribution of the European Heart Rhythm Association (EHRA) of the ESC. Eur Heart J. 2021;42:373–498. - PubMed
    1. Menezes A.R., Lavie C.J., DiNicolantonio J.J., et al. Atrial fibrillation in the 21st century: a current understanding of risk factors and primary prevention strategies. Mayo Clin Proc. 2013;88:394–409. - PubMed

LinkOut - more resources