Identifying Metabolic Syndrome Easily and Cost Effectively Using Non-Invasive Methods with Machine Learning Models

Area of Science:

• Medical Informatics
• Machine Learning in Healthcare
• Public Health Screening

Background:

• Metabolic syndrome (MetS) poses a significant global health challenge.
• Early identification and intervention are crucial for managing MetS and its complications.
• Current screening methods can be invasive or costly, limiting large-scale application.

Purpose of the Study:

• To develop and validate machine learning (ML) models for early and low-cost identification of MetS.
• To utilize non-invasive factors for MetS detection in a large population undergoing physical examinations.
• To establish a convenient and cost-effective tool for MetS screening.

Main Methods:

• Utilized data from 9171 participants undergoing physical examinations.
• Collected non-invasive factors: gender, age, BMI, systolic blood pressure (SBP), and diastolic blood pressure (DBP).
• Trained and evaluated multiple ML models including logistic regression, k-NN, naive bayesian (NB), decision tree, random forest, ANN, and SVM.

Main Results:

• Most ML models demonstrated good performance in 10-fold cross-validation.
• The NB model achieved the highest performance in external validation.
• NB model reported an AUC of 0.976, accuracy of 0.923, sensitivity of 98.32%, and specificity of 91.32%.

Conclusions:

• Proposed a novel non-invasive ML-based method for early MetS identification.
• The NB model shows significant potential as a highly sensitive and cost-effective tool for large-scale MetS screening.
• This approach facilitates convenient and economical screening for metabolic syndrome.