Article Impact Level: HIGH Data Quality: STRONG Summary of JACC: Cardiovascular Imaging, 16(6), 800–816. https://doi.org/10.1016/j.jcmg.2022.11.018 Henry W. West et al.
Points
- A deep-learning network was developed to automate the quantification of epicardial adipose tissue (EAT) volume from coronary computed tomography angiograms (CCTA).
- EAT volume measured by the automated tool strongly correlated with human measurements, indicating reliability.
- Even after adjusting for risk factors, EAT volume was significantly associated with coronary artery disease, atrial fibrillation, all-cause mortality, myocardial infarction, and stroke.
- The automated EAT volume measurement demonstrated prognostic value, independently predicting adverse outcomes in both short-term and long-term follow-ups.
- This research highlights the potential of using machine-learning algorithms to improve cardiovascular risk stratification by accurately assessing metabolically unhealthy visceral obesity.
Summary
This research paper presents a deep-learning network developed to automate the quantification of epicardial adipose tissue (EAT) volume from coronary computed tomography angiograms (CCTA). The study assessed the clinical value of integrating EAT volume measurement in routine CCTA interpretation. The network was trained and validated using a large cohort of CCTA scans and tested in patients with challenging anatomy and scan artifacts. The prognostic value of EAT volume was also validated in patients post-cardiac surgery and the SCOT-HEART trial cohort.
The results showed a high concordance correlation coefficient between the machine and human measurements, indicating strong agreement. Even after adjusting for risk factors, EAT volume was associated with coronary artery disease, atrial fibrillation, all-cause mortality, myocardial infarction, and stroke. The prognostic value of EAT volume was independent of traditional risk factors and predicted adverse outcomes such as mortality, myocardial infarction, stroke, and atrial fibrillation in both short-term and long-term follow-ups.
The findings suggest that automated assessment of EAT volume using CCTA is feasible and reliable, even in challenging cases. EAT volume is a robust marker of metabolically unhealthy visceral obesity and can enhance cardiovascular risk stratification. The study highlights the potential of using machine-learning algorithms to improve the interpretation of cardiac imaging studies, particularly in non-expert centers and large datasets, facilitating early risk identification and potentially improving patient outcomes. Further research is needed to explore the clinical utility of noninvasive markers for identifying and managing cardiovascular risk associated with visceral obesity.
Link to the article: https://www.sciencedirect.com/science/article/pii/S1936878X22007227
References West, H. W., Siddique, M., Williams, M. C., Volpe, L., Desai, R., Lyasheva, M., Thomas, S., Dangas, K., Kotanidis, C. P., Tomlins, P., Mahon, C., Kardos, A., Adlam, D., Graby, J., Rodrigues, J. C. L., Shirodaria, C., Deanfield, J., Mehta, N. N., Neubauer, S., … Antoniades, C. (2023). Deep-learning for epicardial adipose tissue assessment with computed tomography. JACC: Cardiovascular Imaging, 16(6), 800–816. https://doi.org/10.1016/j.jcmg.2022.11.018
