Performance of machine learning-based coronary computed tomography angiography for selecting revascularization...

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Limited research exists on the diagnostic accuracy of machine learning-based coronary computed tomography angiography (ML-CCTA) for therapeutic decision-making compared to conventional CCTA.
• Accurate assessment of coronary artery disease is crucial for guiding appropriate patient treatment strategies.

Purpose of the Study:

• To evaluate and compare the performance of ML-CCTA versus CCTA in guiding therapeutic decisions for patients with coronary artery disease.
• To assess the accuracy of ML-CCTA in identifying candidates for revascularization procedures.

Main Methods:

• A study involving 322 patients with stable coronary artery disease was conducted.
• The SYNTAX score was calculated using ML-CCTA results. Therapeutic decisions were made based on ML-CCTA and SYNTAX scores.
• Treatment strategies were independently determined using ML-CCTA, CCTA, and invasive coronary angiography (ICA).

Main Results:

• ML-CCTA showed higher accuracy (91.93%) than CCTA (86.65%) in identifying revascularization candidates, with a significantly higher area under the receiver operating characteristic curve (AUC) (0.917 vs. 0.866, P=0.016).
• ML-CCTA also demonstrated superior AUCs for selecting candidates for percutaneous coronary intervention (PCI) (0.883 vs. 0.777, P<0.001) and coronary artery bypass graft (CABG) (0.912 vs. 0.826, P=0.003).
• Sensitivity, specificity, and predictive values favored ML-CCTA over CCTA when compared against ICA as the gold standard.

Conclusions:

• ML-CCTA effectively differentiates patients who require revascularization from those who do not.
• ML-CCTA offers a slightly superior approach compared to CCTA for making appropriate therapeutic decisions and selecting optimal revascularization strategies in coronary artery disease patients.