Exploring the Potential Performance of Fibroscan for Predicting and Evaluating Metabolic Syndrome using a Feature
Kuan-Lin Chiu1, Yu-Da Chen1, Sen-Te Wang1,2,3
1Department of Family Medicine, Taipei Medical University Hospital, Taipei 110301, Taiwan.
Metabolites
|July 29, 2023
Summary
The controlled attenuation parameter (CAP) score, measured non-invasively, shows promise for identifying metabolic syndrome (MetS). Machine learning models indicate CAP is a key biomarker for early MetS detection.
Area of Science:
- Hepatology
- Cardiology
- Data Science
Background:
- Metabolic syndrome (MetS) increases cardiovascular risks.
- Non-alcoholic fatty liver disease (NAFLD) is the liver manifestation of MetS and a leading cause of cirrhosis.
- FibroScan® offers non-invasive assessment of liver steatosis and fibrosis using CAP and LSM scores.
Purpose of the Study:
- To evaluate the feasibility of using the controlled attenuation parameter (CAP) score with machine learning (ML) for metabolic syndrome (MetS) detection.
- To identify the most significant biomarkers for predicting MetS using ML and recursive feature elimination (RFE).
Main Methods:
- Application of multiple machine learning models with recursive feature elimination (RFE) algorithm.
- Analysis of controlled attenuation parameter (CAP) and liver stiffness measurement (LSM or E) scores.
- ANOVA testing to identify significant symptoms at various CAP and E score levels.
Main Results:
- All eight ML models achieved accuracy scores greater than 0.9.
- Treebag and random forest models exhibited the highest kappa values (0.6439 and 0.6533).
- The CAP score was identified as the most important variable across seven of the ML models.
Conclusions:
- Machine learning models utilizing RFE demonstrate the potential feasibility of using CAP scores for identifying patients with MetS.
- Combining CAP scores with other significant biomarkers may enable early detection and prediction of MetS.
