Contralateral artery enlargement predicts carotid plaque progression based on machine learning algorithm models in apoE(-/-) mice.

BACKGROUND: This study specifically focused on anatomical MRI characterization of the low shear stress-induced atherosclerotic plaque in mice. We used machine learning algorithms to analyze multiple correlation factors of plaque to generate predictive models and to find the predictive factor for vulnerable plaque.

METHODS: Branches of the left carotid artery in apoE(-/-) and C57BL/6J mice were ligated to produce the partial left carotid artery model. Before surgery, and 7, 14, and 28 days after surgery, in vivo serial MRI measurements of carotid artery diameter were obtained. Meanwhile, proximal blood flow was evaluated. After image acquisition and animal sacrifice, carotid arteries were harvested for histological analysis. Support vector machine (SVM) and decision tree (DT) were used to select features and generate predictive models of vulnerable plaque progression.

RESULT: Seven days after surgery, neointima formation was visualized on micro-MRI in both apoE(-/-) and C57BL/6J mice. Ultrasonography showed that blood flow had significantly decreased compared to that in the contralateral artery. Partial ligation of the carotid artery for 4 weeks in apoE(-/-) mice induced vulnerable plaque; however, in C57BL/6J mice this same technique performed for 4 weeks induced arterial stenosis. Contralateral carotid artery diameter at 7 days after surgery was the most reliable predictive factor in plaque progression. We achieved over 87.5% accuracy, 80% sensitivity, and 95% specificity with SVM. The accuracy, sensitivity, and specificity for the DT classifier were 90, 90, and 90%, respectively.

CONCLUSIONS: This study is the first to demonstrate that SVM and DT methods could be suitable models for identifying vulnerable plaque progression in mice. And contralateral artery enlargement can predict the vulnerable plaque in carotid artery at the very early stage. It may be a valuable tool which helps to optimize the clinical work flow process by providing more decision in selecting patients for treatment.