Predicting coronary artery calcified plaques using perivascular fat CT radiomics features and clinical risk factors.

ABSTRACT

OBJECTIVE: The purpose of this study was to develop a combined radiomics model to predict coronary plaque texture using perivascular fat CT radiomics features combined with clinical risk factors. METHODS: The data of 200 patients with coronary plaques were retrospectively analyzed and randomly divided into a training group and a validation group at a ratio of 73. In the training group, The best feature set was selected by using the maximum correlation minimum redundancy method and the least absolute shrinkage and selection operator. Radiomics models were built based on different machine learning algorithms. The clinical risk factors were then screened using univariate logistic regression analysis. and finally a combined radiomics model was developed using multivariate logistic regression analysis to combine the best performing radiomics model with clinical risk factors and validated in the validation group. The efficacy of the model was assessed by a receiver operating characteristic curve, the consistency of the nomogram was assessed using calibration curves, and the clinical usefulness of the nomogram was assessed using decision curve analysis. RESULTS: Twelve radiomics features were used by different machine learning algorithms to construct the radiomics model. Finally, the random forest algorithm built the best radiomics model in terms of efficacy, and this was combined with age to construct a combined radiomics model. The area under curve for the training and validation group were 0.98 (95% confidence interval, 0.95-1.00) and 0.97 (95% confidence interval, 0.92-1.00) with sensitivities of 0.92 and 0.86 and specificities of 0.99 and 1, respectively. The calibration curve demonstrated that the nomogram had good consistency, and the decision curve analysis demonstrated that the nomogram had high clinical utility. CONCLUSIONS: The combined radiomics model established based on CT radiomics features and clinical risk factors has high value in predicting coronary artery calcified plaque and can provide a reference for clinical decision-making.