AI-Defined Cardiac Anatomy Improves Risk Stratification of Hybrid Perfusion Imaging

Abstract

Background: Computed tomography attenuation correction (CTAC) improves perfusion quantification of hybrid myocardial perfusion imaging by correcting for attenuation artifacts. Artificial intelligence (AI) can automatically measure coronary artery calcium (CAC) from CTAC to improve risk prediction but could potentially derive additional anatomic features.

Objectives: The authors evaluated AI-based derivation of cardiac anatomy from CTAC and assessed its added prognostic utility.

Methods: The authors considered consecutive patients without known coronary artery disease who underwent single-photon emission computed tomography/computed tomography (CT) myocardial perfusion imaging at 3 separate centers. Previously validated AI models were used to segment CAC and cardiac structures (left atrium, left ventricle, right atrium, right ventricular volume, and left ventricular [LV] mass) from CTAC. They evaluated associations with major adverse cardiovascular events (MACEs), which included death, myocardial infarction, unstable angina, or revascularization.

Results: In total, 7,613 patients were included with a median age of 64 years. During a median follow-up of 2.4 years (IQR: 1.3-3.4 years), MACEs occurred in 1,045 (13.7%) patients. Fully automated AI processing took an average of 6.2 ± 0.2 seconds for CAC and 15.8 ± 3.2 seconds for cardiac volumes and LV mass. Patients in the highest quartile of LV mass and left atrium, LV, right atrium, and right ventricular volume were at significantly increased risk of MACEs compared to patients in the lowest quartile, with HR ranging from 1.46 to 3.31. The addition of all CT-based volumes and CT-based LV mass improved the continuous net reclassification index by 23.1%.

Conclusions: AI can automatically derive LV mass and cardiac chamber volumes from CT attenuation imaging, significantly improving cardiovascular risk assessment for hybrid perfusion imaging.

Keywords: artificial intelligence; hybrid imaging; myocardial perfusion imaging.

Figure 1:. Correlation between measurements

Correlation (left) between left ventricular (LV) mass index and LV volume by computed tomography (CT) and single photon emission computed tomography (SPECT). Bland-Altman plots for the same measures shown on the right.

Figure 2:. Case examples.

In panel A, results for 63-year-old female patient with normal left ventricular (LV) mass index by computed tomography (CT) but abnormal by single photon emission computed tomography (SPECT). All CT volumes were in the first or second quartile. The patient did not experience major adverse cardiovascular events during 1.8 years follow-up. In panel B, results for a 75-year-old male patient with abnormal LV mass index by CT, but normal by SPECT. The LV mass is normal by SPECT due to inability to track the inferolateral wall in the presence of extensive perfusion abnormality. All CT volumes were in the 3^rd or 4^th quartile. The patient underwent revascularization 3 days after SPECT. ANT – anterior, EDV – end-diastolic volume, ESV – end-systolic volume, INF – inferior, LA – left atrium, LAT – lateral, RA – right atrium, RV – right ventricle, SEPT - septal.

Figure 3:. Correlation between ungated and gated measurements.

Correlation between estimates of cardiac volumes and left ventricular (LV) mass from low-dose, ungated attenuation correction computed tomography (CT) and paired contrast-enhanced, ECG-gated, cardiac CT. Both scans were performed on the same day, during a single imaging session.

Figure 4:. Outcomes stratified by chamber volume.

Kaplan-Meier survival curves for major adverse cardiovascular events (MACE) by quartile of cardiac chamber volumes from computed tomography using deep learning. Patients in the highest quartile of chamber volume were more likely to experience MACE. HR – hazard ratio.

Figure 5:. Outcomes stratified by left ventricular mass

Kaplan-Meier curves stratified by the presence of abnormal left ventricular (LV) mass index from computed tomography using deep learning, stratified by sex. HR – hazard ratio.

Figure 6:. Results of the continuous net-reclassification analysis.

The base multivariable model included age, sex, body mass index, past medical history, coronary artery calcium, stress total perfusion deficit and left ventricular ejection fraction. For computed tomography (CT) measurements volume included left atrial, left ventricle, right atrial, and right ventricle volume. For single photon emission computed tomography (SPECT) measurements, volume included only left ventricular end-diastolic volume.

Central Illustration:

Overview of study design. We included patients who underwent hybrid perfusion imaging from 3 sites. Two previously developed artificial intelligence (AI) models were used to segment non-contrast computed tomography (CT) images. We evaluated associations with major adverse cardiovascular events (MACE), here showing patients stratified by the presence of abnormal left ventricular (LV) mass index. SPECT – single photon emission computed tomography.