Abstract Body (Do not enter title and authors here): BACKGROUND
While non-contrast CT scans are routinely used to assess the risk of major adverse cardiac events by quantifying coronary artery calcium (CAC), there is growing interest in leveraging cardiac chamber volumetry derived from these scans to improve the prediction of heart failure (HF). However, it is unknown whether cardiac chamber volumes from varying phases of the cardiac cycle differ in ability to predict HF risk.
RESEARCH QUESTION
Does the phase of cardiac cycle during CT acquisition affect the predictive value of chamber volumes for heart failure risk?
METHODS
This is a retrospective cohort of 5,325 asymptomatic patients aged 45-75 years, without known cardiac disease, who underwent CT imaging for CAC scoring between 2010-2023. CT images were analyzed using a previously validated convolutional neural network-based autoencoder model to derive chamber volumes. Patients were divided into groups based on cardiac phase during CT acquisition (diastolic or systolic). Cox proportional hazard regression was used to assess the association between chamber volumes and heart failure. Hazard ratios (HR) were adjusted for clinical risk factors.
RESULTS
Patients were imaged during diastole (n=3,172, 58±9 years old, 44% women, 81% White) and during systole (n=2,153, 58±10 years old, 46% women, 84% White). Over a mean follow-up period of 4.7±2.6 years, 7.2% (n=384) developed HF within 10 years. There were 215 HF diagnoses in the diastolic group (6.8%) and 169 in the systolic group (7.8%). The strongest predictors of HF were left atrium (LA) and myocardial volumes. Optimal cutoffs for risk stratification were higher for diastolic chamber volumes (LA: >90.4 mL vs >82.1 mL; myocardium: >136.4 mL vs >102.2 mL). Greater LA and myocardial diastolic chamber volumes were independently associated with higher risk of HF than systolic volumes (LA: HR 1.379 [95% CI, 1.233-1.542] vs HR 1.223 [95% CI, 1.084-1.380], P<0.0001; myocardium: HR 1.513 [95% CI, 1.329-1.764] vs HR 1.426 [95% CI, 1.169-1.740], P<0.0001). Diastolic volumes had significant predictive improvement over systolic volumes (LA: AUC 0.712 [0.669-0.753) vs 0.703 [0.659-0.745]; myocardium: AUC 0.717 [0.674-0.757] vs 0.705 [0.662-0.745]).
CONCLUSION
AI-derived diastolic chamber volumes from CAC CT scans outperformed systolic volumes in prediction of HF risk. These findings highlight the need for consideration of cardiac cycle phase during CT acquisition in the use of chamber volumetry for HF risk stratification.