First comparison between artificial intelligence-guided coronary computed tomography angiography versus single-photon...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Noninvasive cardiac testing is evolving with coronary computed tomography angiography (CCTA) and single-photon emission computed tomography (SPECT) as alternatives to invasive angiography.
• Obstructive coronary artery disease evaluation is increasingly relying on noninvasive methods.
• The study investigates a novel artificial intelligence (AI)-assisted CCTA program against SPECT for ischemic testing.

Purpose of the Study:

• To evaluate the comparability of a novel AI-assisted CCTA program to SPECT imaging for detecting myocardial ischemia.
• To assess the diagnostic performance of AI-QCTISCHEMIA-augmented CCTA against SPECT and invasive angiography.

Main Methods:

• Coronary computed tomography angiography (CCTA) images were analyzed using an AI convolutional neural network machine-learning model (AI-QCTISCHEMIA).
• 183 patients underwent AI-QCTISCHEMIA-augmented CCTA; 60 had concurrent SPECT, and 16 had invasive coronary angiograms.
• Eight studies were excluded due to incomplete data or coronary anomalies, resulting in 175 patients for analysis.

Main Results:

• AI-QCTISCHEMIA-driven CCTA demonstrated higher sensitivity (75%) and specificity (70%) for predicting coronary ischemia compared to SPECT (70% and 53%).
• The negative predictive value was significantly higher for female patients using AI-QCTISCHEMIA (91%) versus SPECT (68%).
• Areas under the receiver operating characteristic curves were similar (0.81 for AI-CCTA, 0.75 for SPECT), with a significant correlation (r = 0.71) between modalities.

Conclusions:

• AI-powered CCTA is a viable alternative to SPECT for detecting myocardial ischemia in low- to intermediate-risk coronary artery disease.
• Results from AI-CCTA and SPECT imaging were comparable for patients undergoing invasive angiography.
• Further prospective studies with larger, diverse populations are needed to confirm the benefits of AI-driven CCTA.