Heart Failure and Atrial Fibrillation Prediction from Non-gated Chest CT Using AI-Based Cardiac Chamber Volumetry — An AI-CVD Study within MESA

Can a routine non-gated chest CT help predict future heart failure and atrial fibrillation?

A new HeartLung.AI AI-CVD study, to be presented at ESC Congress 2026, investigates whether AI-derived cardiac chamber measurements from routine non-gated chest CT can predict future heart failure (HF) and atrial fibrillation (AF) as effectively as measurements obtained from ECG-gated cardiac CT.

This question has significant clinical implications. Approximately 20 million chest CTs are performed annually in the U.S., while fewer than 5% are ECG-gated. If reliable cardiac measurements can be extracted from these routine scans, a vast number of existing CT examinations could potentially provide additional cardiovascular risk information.

Key Findings

The study compared AI-based cardiac chamber volumetry from non-gated versus ECG-gated CT in 2,053 MESA participants, evaluating measures including left atrial volume, left ventricular volume and mass, and LV/RV ratio.

Agreement between the two approaches was high:

• LA volume: ICC 0.94

• LV volume: ICC 0.95

• LV mass: ICC 0.96

• LV/RV ratio: ICC 0.89

• 
  

Most importantly, AI-derived measurements from non-gated CT were non-inferior to ECG-gated CT for predicting incident heart failure and atrial fibrillation.

For example, the C-index for predicting heart failure using LV volume index was 0.77 vs. 0.77, while prediction of atrial fibrillation using left atrial volume index was 0.74 vs. 0.75 for non-gated and gated CT, respectively.

The risk curves also closely overlapped: participants in the highest quartiles of cardiac chamber measurements showed substantially greater incidence of HF and AF, regardless of whether measurements came from non-gated or ECG-gated imaging.

Why It Matters

The opportunity goes beyond improving a dedicated cardiac scan.

Routine chest CTs are already being performed every day for many different clinical indications. AI could potentially analyze these existing scans to extract cardiac chamber information and identify patients at increased risk for future heart failure or atrial fibrillation—without requiring a dedicated cardiac CT.

Because non-gated scans greatly outnumber gated scans and already exist within health-system imaging archives, this approach could significantly expand opportunistic cardiovascular screening.

One routine chest CT. Additional cardiac insight. No additional scan, cost, or radiation.

📍 To be presented at ESC Congress 2026 · Munich

By Seyed Reza Mirjalili, MD