Relevant Features in Nonalcoholic Steatohepatitis Determined Using Machine Learning for Feature Selection.
Rafael Garcia-Carretero1, Luis Vigil-Medina1, Oscar Barquero-Perez2
1Department of Internal Medicine, Mostoles University Hospital, Rey Juan Carlos University, Mostoles, Spain.
Metabolic Syndrome and Related Disorders
|November 2, 2019
Summary
Nonalcoholic steatohepatitis (NASH) affects 11.3% of hypertensive patients. Ferritin and insulin levels are key indicators for identifying NASH and assessing cardiovascular risk in this population.
Area of Science:
- Hepatology
- Cardiology
- Data Science
Background:
- Nonalcoholic fatty liver disease (NAFLD) encompasses nonalcoholic steatohepatitis (NASH), characterized by liver inflammation.
- NASH increases risks for cardiovascular disease, liver fibrosis, cirrhosis, and transplantation.
- Hypertension is a common comorbidity, potentially exacerbating NASH outcomes.
Purpose of the Study:
- To determine the prevalence of NASH in hypertensive patients.
- To identify the most relevant clinical features associated with NASH in this cohort.
- To evaluate the utility of machine learning in predicting NASH.
Main Methods:
- Analysis of data from 2239 hypertensive patients.
- Descriptive statistics to characterize the cohort.
- Supervised machine learning algorithms (LASSO, Random Forest) for feature selection.
Main Results:
- NASH prevalence was 11.3% among hypertensive patients.
- Associated factors included metabolic syndrome, type 2 diabetes, insulin resistance, and dyslipidemia.
- Ferritin and serum insulin were the most significant predictors (AUC 0.79).
Conclusions:
- Ferritin and insulin are significant predictors of NASH in hypertensive individuals.
- These biomarkers can aid clinicians in assessing cardiovascular risk and managing NASH.
- Machine learning offers valuable support for clinical decision-making in NASH management.
