Fetal Health Risk Detection using Machine Learning

AI-powered web app for early detection of fetal health risks, enabling timely interventions for safer pregnancies and healthier outcomes.

Comment 1

Inspiration

The idea for this project stemmed from the pressing need to address neonatal mortality rates globally. Statistics from UNICEF reveal that millions of newborns face preventable health risks during their first month of life. This inspired us to develop a solution that could aid healthcare professionals in detecting fetal health risks early, leveraging machine learning and accessible technology to make an impact.

What it does

The Fetal Health Risk Detection Web App is a user-friendly platform that predicts fetal health risks based on Cardiotocography (CTG) data. Users input key parameters such as fetal heart rate and uterine contractions, and the app classifies the fetal condition into three categories: Normal, Suspected, or Pathological. This provides healthcare professionals with quick, data-driven insights for timely intervention.

How we built it

  1. Data Collection: We used a dataset of 2,126 CTG records obtained from the University of California Irvine repository. Each record included parameters analyzed and classified by medical experts.

  2. Preprocessing:

    • Cleaned and normalized the data to ensure consistency.
    • Explored feature relationships using visualizations like correlation matrices and boxplots.

  3. Model Development:

    • Experimented with various classification algorithms (Logistic Regression, KNN, Decision Tree, SVM, etc.).
    • Selected K-Nearest Neighbors (KNN) due to its simplicity and high accuracy (90%+).

  4. Web Application:

    • Built the interface using Flask for backend operations and Bootstrap for a clean, responsive design.
    • Integrated the trained model with the web app for real-time predictions.

  5. Deployment: Deployed the application on a cloud platform to ensure accessibility and scalability.

Challenges we ran into

  • Data Quality: Cleaning and preprocessing CTG data while maintaining its integrity was a time-intensive process.
  • Model Interpretability: Ensuring healthcare professionals could understand the decision-making process of the model required additional work on explainability.
  • Integration: Merging the machine learning model with a seamless web interface without compromising performance was a technical hurdle.
  • Deployment: Configuring the app for reliable cloud hosting required troubleshooting scalability and latency issues.

Accomplishments that we're proud of

  • Successfully achieving a prediction accuracy of over 90%.
  • Designing a web app that combines advanced machine learning with an intuitive interface.
  • Creating a tool that has the potential to positively impact maternal and neonatal health outcomes.
  • Implementing interpretability features to foster trust in the predictions made by the model.

What we learned

  • Data Science: Advanced skills in data preprocessing, feature selection, and model evaluation.
  • Web Development: Building user-friendly applications using Flask, HTML, CSS, and JavaScript.
  • Deployment: Navigating the complexities of deploying machine learning applications to the cloud.
  • Collaboration: The importance of integrating technical skills with domain knowledge in healthcare.

What's next

  • Feature Expansion: Incorporate more features like maternal data to enhance predictive accuracy.
  • Model Optimization: Explore deep learning techniques for even better performance.
  • Mobile Integration: Develop a mobile version of the application for wider accessibility.
  • Clinical Validation: Partner with healthcare institutions to test and refine the app in real-world settings.
  • Continuous Monitoring: Implement a feedback loop to monitor model performance and improve predictions over time.
Share this project:

Updates