Machine learning models for screening carotid atherosclerosis in asymptomatic adults

Area of Science:

• Cardiovascular Medicine
• Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Carotid atherosclerosis (CAS) is a significant risk factor for cardiovascular and cerebrovascular events.
• Current medical guidelines do not recommend duplex ultrasonography for routine screening of asymptomatic populations.
• There is a need for effective screening methods for CAS in asymptomatic individuals.

Purpose of the Study:

• To develop and evaluate machine learning models for screening carotid atherosclerosis (CAS) in asymptomatic adults.
• To identify the most effective machine learning algorithm for CAS detection using routine physical examination data.

Main Methods:

• A total of 2732 asymptomatic subjects were included in the study.
• Machine learning models including decision tree, random forest (RF), extreme gradient boosting (XGBoost), support vector machine (SVM), and multilayer perceptron (MLP) were developed using 17 candidate features.
• Model performance was assessed on a testing dataset.

Main Results:

• The multilayer perceptron (MLP) model achieved the highest performance metrics: accuracy (0.748), positive predictive value (0.743), F1 score (0.742), AUC (0.766), and Kappa score (0.445).
• XGBoost and SVM models also showed promising results, ranking second and third, respectively.
• The MLP model proved superior in classifying subjects with or without CAS.

Conclusions:

• Machine learning models, particularly MLP, can effectively screen for asymptomatic carotid atherosclerosis (CAS) using data from routine physical examinations.
• These models offer a practical approach for physicians and primary care doctors to identify individuals at risk.
• Improved screening can enhance risk prediction and prevention strategies for cardiovascular and cerebrovascular events in the general population.