This project aims to enhance the screening process for Congenital Hypothyroidism (CH) using advanced machine learning techniques. By leveraging a large dataset from Newborn Screening Ontario (NSO), the project evaluates various classifiers and resampling methods to improve the Positive Predictive Value (PPV) while maintaining 100% sensitivity. The ultimate goal is to reduce false positives, thereby minimizing unnecessary retests and alleviating stress for families.
The runtime component of this project is likely to be useful for other projects that require the use of machine learning models for classification tasks. The project is designed to be modular, allowing for easy integration of new classifiers, resampling methods, and datasets. The project is also designed to be user-friendly, with a small variety of example scripts that demonstrate how to use the runtime component.
- Python 3.9
- Pipenv
- Jupyter Notebook (optional)