Abstract 13026: What Causes Speedy Growth (Within 3 to 6 Months) of Vulnerable Plaques in Acute Coronary Syndrome: An Angiographic and Machine Learning Analysis

• Published in: Circulation (New York, N.Y.) N.Y.), 2021-11, Vol.144 (Suppl_1), p.A13026-A13026
• Main Authors: Nguyen, Thach N, Le, Thoa, Le, Minh H, Vuong, Ngoc H, LE, DO H, Nguyen, Nguyet T, Le, Trang T, Nguyen, Thoai H, Cao, Thinh, Rigatelli, Gianluca, Zuin, Marco, Pham, Hung N, Ho, Dung T, Talarico, Ernest, Nguyen Thuong, Nghia T
• Format: Article
• Language: English
• ISSN: 0009-7322; 1524-4539
• DOI: 10.1161/circ.144.suppl_1.13026

IntroductionBased on damages by flow abnormalities seen in pipes, our research is built on the hypothesis that the flow disturbances start and grow atherosclerotic plaques. Our strategy is to use knowledge and practice of fluid mechanics to identify the different types of abnormal coronary flows and correlate them with the appearance and growth of lesions. HypothesisIn patient with fast tract ACS when a normal segment becomes severely stenotic in a matter of 3 to 6 months, what kind of abnormal flow could trigger this speedy growth of a plaque ? MethodsPatients were selected if there was a near normal segment in the first angiogram and severe lesion in the 2nd angiogram 3 to 6 months later. LDL cholesterol level and HbA1c were also measured. In our cardiac catheterization laboratories, in the past 5 years, all patients underwent a special new dynamic coronary angiogram protocol. The contrast was injected until the coronary artery was completely opacified. As the injection of contrast stopped, the blood in white color moved in and displaced the contrast in black. The morphologies, movements and interactions of the blood flow in white color above a black background were reviewed frame by frame by investigators. The main measurements were the types of flow, their durations and direction at the index segment of the 1rst angiogram. At the same time, an artificial intelligence (AI) program (mainly Machine Learning (ML) algorithm) searched by 2 models built on Python. Model 1 was based on U-net and Densenet-121 for vessel segmentation. Model 2 was used for classification of flow (laminar, turbulent, antegrade, retrograde or stagnant). Model 2 was trained based on the convolutional neural network. ResultsTen dual angiograms were selected. At the index segment of the 1st angiogram, 80% showed turbulent and retrograde flow. The duration of this turbulent flow lasted 80% of one cardiac cycle (diastole and systole). The duration of retrograde flow lasted 90% of one systolic phase. A high level of LDL cholesterol (120mg%) was statistically significant while the HbA1c was not. ConclusionsIn patient with fast tract ACS when a near normal segment became severely stenotic in a matter of 3 to 6 months, prolonged turbulent and retrograde flow with high LDL cholesterol were the culprits.