Epicardial adipose tissue radiomics on cardiac CT predicts incident HF

A novel automated AI algorithm for radiomic phenotyping of epicardial adipose tissue shows good discrimination for new-onset HF among adults without prior HF or MI who underwent routine CCTA.

This summary is based on the publication of Oikonomou EK, Chan K, Patel P, et al. Early Prediction of Heart Failure From Routine Cardiac CT Using Radiomic Phenotyping of Epicardial Fat. J Am Coll Cardiol. 2026 Apr 1:S0735-1097(26)05652-4. [Online ahead of print]. doi: 10.1016/j.jacc.2026.02.5116.

Introduction and methods

Background

Epicardial adipose tissue (EAT) is a biologically active fat depot surrounding the human myocardium [1-2]. It exerts paracrine effects on the heart and can also sense inflammatory signals originating from the myocardium [3-4]. In response to early disease signals, EAT modifies its texture and composition. Whether these alterations in texture and composition can capture adverse cardiac remodeling and enable early risk stratification for HF remains unclear.

Aim of the study

The aim of the study was to develop and externally validate a radiomic signature of EAT for the prediction of incident HF.

Methods

This large multicenter cohort study included 72,751 adults without prior HF or MI from the ORFAN (Oxford Risk Factors and Noninvasive Imaging) cohort who underwent routine CCTA at 9 UK centers from 2007 to 2022. A fully automated artificial intelligence pipeline extracted 1,655 radiomic features from EAT to derive a fat radiomic profile for HF (FRP_HF). The model was developed in 59,327 patients from 7 centers (n=47,461 for training and n=11,866 for internal validation) and externally validated in 13,424 individuals from 2 geographically distinct centers. Median follow-up was 5.1 years (Q1-Q3: 3.4-7.6) in the development cohort and 4.0 years (Q1-Q3: 3.0-6.9) in the external validation cohort.

Outcomes

The primary outcome was the incidence of new-onset HF, defined as the first appearance of an HF-specific ICD-10 code ≥30 days after the index CCTA scan.

Main results

• In the training and internal validation cohort, 1,737 (2.9%) participants developed HF during follow-up. In the external validation cohort, 363 (2.7%) participants developed HF during follow-up.
• The FRP_HF model demonstrated strong discrimination for new-onset HF:
  • C-statistic: 0.869 (95%CI: 0.850-0.889) in the internal validation set and 0.850 (95%CI: 0.831-0.870) in the external validation set.
• Each 25-percentile increase in FRP_HF was associated with nearly a 4-fold higher risk of new-onset HF (internal adjusted HR: 3.90; 95%CI: 3.13-4.84; external adjust HR: 3.79; 95%CI: 3.01-4.76; both P<0.001).
• Individuals in the >90th percentile had almost a 20-fold higher HF risk compared with those in the ≤10th percentile (adjusted HR: 19.96; 95%CI: 7.10-56.11; P<0.001).
• Adding FRP_HF to a baseline model (consisting of age, sex, conventional risk factors, and CAD-RADS) significantly improved discrimination and risk stratification:
  • Increase in 5-year AUC by 0.071 (95%CI: 0.051-0.092) in internal validation set and 0.047 (95%CI: 0.029-0.065) in external validation set (both P<0.001).
  • Net reclassification improvement of 0.39 (95%CI: 0.29-0.48).

Conclusion

Automated radiomic phenotyping of EAT from cardiac CT enables early identification of individuals at high risk for HF. The FRP_HF model showed robust discrimination for new-onset HF, and adding the model to conventional risk models significantly improved discrimination. The authors highlight that “these findings establish a new dimension to CCTA interpretation, highlighting its promising role as a screening tool for early signs of HF by uncovering clinically and prognostically relevant phenotypic information hidden in the epicardial fat depot.”

Find this article online at J Am Coll Cardiol.

References

1. Oikonomou EK, Antoniades C. The role of adi pose tissue in cardiovascular health and disease. Nat Rev Cardiol. 2019;16:83–99.
2. Iacobellis G. Epicardial adipose tissue in contemporary cardiology. Nat Rev Cardiol. 2022;19:593–606.
3. Carena MC, Badi I, Polkinghorne M, et al. Role of human epicardial adipose tissue-derived miR 92a-3p in myocardial redox state. J Am Coll Car diol. 2023;82:317–332.
4. Antonopoulos AS, Margaritis M, Verheule S, et al. Mutual regulation of epicardial adipose tis sue and myocardial redox state by PPAR-γ/adi ponectin signalling. Circ Res. 2016;118:842–855.