Related Experiment Video
Updated: Jul 10, 2025
07:34
Fetal Mouse Cardiovascular Imaging Using a High-frequency Ultrasound 30/45MHZ System
Published on: May 5, 2018
11.7K
A cluster-based ensemble approach for congenital heart disease prediction
1Sri Guru Tegh Bahadur Khalsa College, University of Delhi, Delhi, India.
Computer Methods and Programs in Biomedicine
|November 20, 2023
Summary
This study developed a novel data analysis model for predicting congenital heart disease (CHD) risk in expectant mothers. The model achieved 99% accuracy by effectively handling imbalanced and missing data, improving upon existing methods.
Area of Science:
- Medical Informatics
- Data Mining
- Machine Learning
Background:
- Congenital heart diseases (CHD) are common birth defects, yet risk factors and prediction models using population data are underexplored.
- Existing research often struggles with imbalanced datasets, hindering accurate CHD risk forecasting.
Purpose of the Study:
- To develop a robust data analysis model for predicting CHD risk, addressing challenges of missing and imbalanced data.
- To identify and create cohorts of expectant mothers with shared lifestyle characteristics for targeted analysis.
Main Methods:
- Employed Density-Based Spatial Clustering of Applications with Noise (DBSCAN) for cohort identification.
- Utilized Random Forest for CHD prediction, incorporating DBSCAN-generated clusters.
- Applied k-NN imputation for missing data and SMOTE for balancing imbalanced datasets.
Main Results:
- DBSCAN identified three distinct patient cohorts.
- The clustering approach significantly improved Random Forest model accuracy for CHD prediction.
- The model achieved 99% accuracy and 0.91 AUC, outperforming state-of-the-art methods.
Conclusions:
- Unsupervised learning via DBSCAN provides valuable insights for CHD prediction classifiers.
- The proposed data-driven approach enhances classification performance and reveals complex influencing factors.
