TY - JOUR
T1 - Artificial Intelligence Prediction of Cardiovascular Events Using Opportunistic Epicardial Adipose Tissue Assessments From Computed Tomography Calcium Score
AU - Hu, Tao
AU - Freeze, Joshua
AU - Singh, Prerna
AU - Kim, Justin
AU - Song, Yingnan
AU - Wu, Hao
AU - Lee, Juhwan
AU - Al-Kindi, Sadeer
AU - Rajagopalan, Sanjay
AU - Wilson, David L.
AU - Hoori, Ammar
N1 - Publisher Copyright:
© 2024 The Authors
PY - 2024/9
Y1 - 2024/9
N2 - Background: Recent studies have used basic epicardial adipose tissue (EAT) assessments (eg, volume and mean Hounsfield unit [HU]) to predict risk of atherosclerosis-related, major adverse cardiovascular events (MACEs). Objectives: The purpose of this study was to create novel, hand-crafted EAT features, “fat-omics,” to capture the pathophysiology of EAT and improve MACE prediction. Methods: We studied a cohort of 400 patients with low-dose cardiac computed tomography calcium score examinations. We purposefully used a MACE-enriched cohort (56% event rate) for feature engineering purposes. We divided the cohort into training/testing sets (80%/20%). We segmented EAT using a previously validated, deep-learning method with optional manual correction. We extracted 148 initial EAT features (eg, morphologic, spatial, and HU), dubbed fat-omics, and used Cox elastic-net for feature reduction and prediction of MACE. Bootstrap validation gave CIs. Results: Traditional EAT features gave marginal prediction (EAT-volume/EAT-mean-HU/BMI gave C-indices 0.53/0.55/0.57, respectively). Significant improvement was obtained with the 15-feature fat-omics model (C-index = 0.69, test set). High-risk features included the volume-of-voxels-having-elevated-HU-[-50,-30-HU] and HU-negative-skewness, both of which assess high HU values in EAT, a property implicated in fat inflammation. Other high-risk features include kurtosis-of-EAT-thickness, reflecting the heterogeneity of thicknesses, and EAT-volume-in-the-top-25%-of-the-heart, emphasizing adipose near the proximal coronary arteries. Kaplan-Meyer plots of Cox-identified, high- and low-risk patients were well separated with the median of the fat-omics risk, with the high-risk group having an HR 2.4 times that of the low-risk group (P < 0.001). Conclusions: Preliminary findings indicate an opportunity to use finely tuned, explainable assessments on EAT for improved cardiovascular risk prediction.
AB - Background: Recent studies have used basic epicardial adipose tissue (EAT) assessments (eg, volume and mean Hounsfield unit [HU]) to predict risk of atherosclerosis-related, major adverse cardiovascular events (MACEs). Objectives: The purpose of this study was to create novel, hand-crafted EAT features, “fat-omics,” to capture the pathophysiology of EAT and improve MACE prediction. Methods: We studied a cohort of 400 patients with low-dose cardiac computed tomography calcium score examinations. We purposefully used a MACE-enriched cohort (56% event rate) for feature engineering purposes. We divided the cohort into training/testing sets (80%/20%). We segmented EAT using a previously validated, deep-learning method with optional manual correction. We extracted 148 initial EAT features (eg, morphologic, spatial, and HU), dubbed fat-omics, and used Cox elastic-net for feature reduction and prediction of MACE. Bootstrap validation gave CIs. Results: Traditional EAT features gave marginal prediction (EAT-volume/EAT-mean-HU/BMI gave C-indices 0.53/0.55/0.57, respectively). Significant improvement was obtained with the 15-feature fat-omics model (C-index = 0.69, test set). High-risk features included the volume-of-voxels-having-elevated-HU-[-50,-30-HU] and HU-negative-skewness, both of which assess high HU values in EAT, a property implicated in fat inflammation. Other high-risk features include kurtosis-of-EAT-thickness, reflecting the heterogeneity of thicknesses, and EAT-volume-in-the-top-25%-of-the-heart, emphasizing adipose near the proximal coronary arteries. Kaplan-Meyer plots of Cox-identified, high- and low-risk patients were well separated with the median of the fat-omics risk, with the high-risk group having an HR 2.4 times that of the low-risk group (P < 0.001). Conclusions: Preliminary findings indicate an opportunity to use finely tuned, explainable assessments on EAT for improved cardiovascular risk prediction.
KW - Cox
KW - CT calcium score
KW - epicardial adipose tissue
KW - machine learning
KW - major adverse cardiovascular event
KW - radiomics
KW - risk prediction
UR - http://www.scopus.com/inward/record.url?scp=85203726195&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=85203726195&partnerID=8YFLogxK
U2 - 10.1016/j.jacadv.2024.101188
DO - 10.1016/j.jacadv.2024.101188
M3 - Article
AN - SCOPUS:85203726195
SN - 2772-963X
VL - 3
JO - JACC: Advances
JF - JACC: Advances
IS - 9P2
M1 - 101188
ER -