Anagrafe della ricerca

Acute coronary syndrome (ACS) most commonly arises from thrombosis of coronary artery lesions that angiographically appear mild, typically with non-flow limiting stenosis, but pathologically contain large plaque burden with an organized lipid-rich necrotic core that is separated from the lumen by a thin fibrous cap. These thin cap fibroatheromas place patients at risk for future unstable angina, acute myocardial infarction (MI) and cardiac death, and they have been termed vulnerable plaques (1-3). In this context, optical coherence tomography (OCT) has emerged as one of the most promising tools to assess patients with coronary lesions and to detect key features of plaques at high risk for rupture and consequently responsible of future cardiovascular events (4-6). However, whether prophylactic revascularization of non–flow-limiting vulnerable plaques might improve patient prognosis is unknown (7). To date, the diagnostic yield of invasive and noninvasive imaging techniques in predicting future major adverse cardiovascular events (MACE) among patients with vulnerable plaques remain low and the treatment of non-flow limiting stenosis with high-risk features is still controversial (8). Although the presence of high-risk coronary lesion features confer a higher and exponential risk of adverse events, there are no available imaging and clinical based risk scores that predict the risk of MACE at follow-up in patients with non-flow limiting coronary artery stenosis. The aim of our project is to predict with the aid of artificial intelligence (AI) and machine learning techniques the natural history of non-flow limiting coronary artery stenoses, and to develop and validate a machine learning risk score capable of estimate the risk of MACE during follow-up based on the OCT findings observed at the coronary plaques not undergoing percutaneous coronary intervention (PCI) and the clinical characteristics of the patients (9,10). We will collect information on all the coronary lesions not undergoing PCI evaluated by performing OCT in non-culprit vessels of patients presenting with ACS at the index procedure. All the OCT runs will be digitalized and analyzed with the aid of AI. The OCT appearance of the coronary plaques will be analyzed with AI and related with the risk of MACE during follow-up (a composite of cardiac death, and myocardial infarction). A predictive risk AI model based on the OCT images of the coronary plaques and the clinical characteristics of the patients will be developed to estimate the risk of the incidence of MACE at follow-up. State-of-the-art machine learning algorithms, including convolutional neural networks, random forests, and support vector machines, will be exploited and we will calculate the area under the curve (AUC) of the receiver operating characteristic curve for the internal validation dataset to select a probability threshold, which we will apply to the testing dataset.