Related Experiment Video
Updated: Dec 28, 2025
12:18
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
7.8K
Screening for chronic conditions with reproductive factors using a machine learning based approach
Siyu Tian1, Weinan Dong2,3, Ka Lung Chan4
1Department of Obstetrics and Gynecology, Li Ka Shing Faculty of Medicine, University of Hong Kong, Sassoon Road, Hong Kong, China.
Scientific Reports
|February 20, 2020
Summary
Reproductive factors may help identify undiagnosed chronic conditions like diabetes and hypertension. Earlier menarche and shorter reproductive lifespan are linked to higher general chronic condition scores in women.
Area of Science:
- Endocrinology and Metabolic Diseases
- Reproductive Health
- Biomarker Discovery
Background:
- Many chronic conditions, including diabetes, hypertension, and dyslipidemia, are often undiagnosed.
- Reproductive factors (RFs) are not currently utilized in screening guidelines for these conditions.
- Integrating RFs could potentially enhance the effectiveness of current screening strategies.
Purpose of the Study:
- To investigate the associations between reproductive factors and biomarkers of chronic conditions.
- To develop a comprehensive marker for general chronic conditions (GCC) using advanced machine learning.
- To evaluate the utility of reproductive factors in predicting this GCC marker.
Main Methods:
- Cross-sectional study involving 1,656 postmenopausal females.
- Collected data on demographics, reproductive factors, and metabolic biomarkers.
- Compared Principal Component Analysis (PCA) and autoencoder for dimensionality reduction, adopting the superior autoencoder to derive the GCC marker.
- Utilized multivariate linear regression to analyze the relationship between GCC and RFs.
Main Results:
- A multi-layer autoencoder demonstrated superior performance over PCA for dimensionality reduction.
- The derived GCC marker effectively represented three chronic conditions (AUCs: 0.844 for glycemia, 0.824 for hypertension, 0.805 for dyslipidemia).
- Earlier age at menarche (OR=0.9976) and shorter reproductive lifespan (OR=0.9895) were significantly associated with higher GCC.
Conclusions:
- Reproductive factors show potential for integration into screening tools for general chronic conditions.
- The autoencoder model effectively reduced complex metabolic biomarker data into a single predictive marker.
- Incorporating accessible reproductive factors could significantly enhance existing screening guidelines for chronic diseases.
