Machine Learning with ¹⁸F-Sodium Fluoride PET and Quantitative Plaque Analysis on CT Angiography for the Future Risk of Myocardial Infarction

Abstract

Coronary 18 F-sodium fluoride ( 18 F-NaF) positron emission tomography (PET) and computed tomography (CT) angiography-based quantitative plaque analysis have shown promise in refining risk stratification in patients with coronary artery disease. We combined both of these novel imaging approaches to develop an optimal machine-learning model for the future risk of myocardial infarction in patients with stable coronary disease. MethodsPatients with known coronary artery disease underwent coronary 18 F-NaF PET and CT angiography on a hybrid PET/CT scanner. Machine-learning by extreme gradient boosting was trained using clinical data, CT quantitative plaque analysis measures and 18 F-NaF PET, and it was tested using repeated 10-fold hold-out testing. ResultsAmong 293 study participants (65±9 years; 84% male), 22 subjects experienced a myocardial infarction over the 53 [40][41][42][43][44][45][46][47][48][49][50][51][52][53][54][55][56][57][58][59] months of follow-up. On univariable receiver-operator-curve analysis, only 18 F-NaF coronary uptake emerged as a predictor of myocardial infarction (cstatistic 0.76, 95% confidence interval (CI) 0.68-0.83). When incorporated into machine-learning models, clinical characteristics showed limited predictive performance (c-statistic 0.64, 95% CI 0.53-0.76;) and were outperformed by a quantitative plaque analysis-based machine-learning model (c-statistic 0.72, 95% CI 0.60-0.84). After inclusion of all available data (clinical, quantitative plaque and 18 F-NaF PET), we achieved a substantial improvement (p=0.008 versus 18 F-NaF PET alone) in the model performance (c-statistic 0.85, 95% CI 0.79-0.91). ConclusionsBoth 18 F-NaF uptake and quantitative plaque analysis measures are additive and strong predictors of outcome in patients with established coronary artery disease. Optimal risk stratification can be achieved by combining clinical data with these approaches in a machinelearning model.

A Machine-Learning-Based Risk-Prediction Tool for HIV and Sexually Transmitted Infections Acquisition over the Next 12 Months

“…Machine learning algorithms have advantages for developing predictive models, such as not requiring statistical inferences or assumptions, being data driven, automatically learning from data that identifies complex nonlinear patterns, and exploiting complex interactions between risk factors [ 13 ]. Machine learning models have been used to predict the future risk of other conditions such as suicide [ 14 , 15 ], type 2 diabetes [ 16 ], Alzheimer’s disease [ 17 ], and myocardial infarction [ 18 ]. Two studies using electronic health-record data from the USA reported that machine learning could accurately predict future HIV infection.…”

PET-based artificial intelligence applications in cardiac nuclear medicine

“…CCTA radiomics identified invasive and radionuclide imaging markers of plaque vulnerability with good to excellent diagnostic accuracy, significantly outperforming conventional quantitative and qualitative high-risk plaque features. Kwiecinski et al [13] developed a machine-learning model for prediction of the future risk of myocardial infarction in patients with stable CAD undergoing [ 18 F]-NaF PET/CC-TA. The machine learning included clinical data, CT quantitative plaque analysis measures and [ 18 F]-NaF PET findings of 293 subjects.…”

Clinical Molecular Imaging for Atherosclerotic Plaque

“…Ishiwata et al, who examined 34 patients, showed that the accumulation of sodium fluoride can predict the progression of vascular calcification, which in turn is a predictor of vascular catastrophes during the year after the PET/CT investigation [69]. A study of 293 patients, using PET with fluoride, showed not only its prognostic significance but also the possibility of using machine learning systems to predict the risk of cardiovascular events [70].…”