Predicting breast screening attendance using machine learning techniques.
Vikraman Baskaran1, Aziz Guergachi, Rajeev K Bali
1Ryerson University, TRSM, Toronto, Canada. vikraman@ryerson.ca
Summary
This study introduces a novel machine learning algorithm to predict breast screening attendance, achieving nearly 80% accuracy. Further research is needed to improve negative predictive value and specificity for this important healthcare application.
Area of Science:
- Healthcare Informatics
- Machine Learning Applications
- Biomedical Data Science
Background:
- Machine learning is increasingly used in healthcare prediction.
- Predicting breast screening attendance before mammography is an emerging research area.
- Accurate prediction can optimize screening program resource allocation and patient outreach.
Purpose of the Study:
- To introduce novel predictor attributes for breast screening attendance prediction.
- To present a new hybrid machine learning algorithm combining back-propagation and radial basis function neural networks.
- To evaluate the algorithm's accuracy and efficiency using a large, long-term dataset.
Main Methods:
- Development of a hybrid neural network algorithm integrating back-propagation and radial basis function networks.
- Utilization of a comprehensive 13-year dataset (1995-2008) for algorithm training and testing.
- Validation of algorithm performance across different computational platforms.
Main Results:
- The developed algorithm achieved approximately 80% accuracy.
- High positive predictive value (88%) and sensitivity (88%) were recorded.
- Negative predictive value and specificity were in the range of 40-50%, indicating areas for improvement.
Conclusions:
- The hybrid machine learning algorithm shows promising results for predicting breast screening attendance.
- The recorded accuracy, sensitivity, and positive predictive value support further large-scale testing.
- Enhancements to negative predictive value and specificity are recommended for future iterations.