Cardiac Arrest Prediction

Introduction

Code Blue Situation is an predictive ML model designed to anticipate cardiac arrest in hospitalized patients. Leveraging machine learning models, particularly Neural Networks, the system analyzes real-time patient data such as ECG measurements, vital signs, and demographic information to predict potential cardiac arrest events before they occur. It aims to enhance patient survival rates and optimize critical care resources.

Problem Statement

Cardiac arrest remains a leading cause of in-hospital mortality. Traditional early warning systems like the Modified Early Warning Score (MEWS) are often rigid, prone to errors, and less sensitive to subtle physiological changes. There is an urgent need for a more dynamic, intelligent system capable of earlier and more accurate predictions to enable proactive interventions.

Objectives

  • Predict the likelihood of cardiac arrest based on patient data, particularly ECG signals.
  • Develop multiple machine learning models and finalize the one with the highest accuracy.
  • Compare the predictive performance of traditional algorithms vs. advanced neural networks.
  • Implement a flexible, scalable system that can integrate with hospital monitoring setups.

Product/Project Analysis

  • Data Collection: Used a specialized cardiac patient dataset containing 279 features (linear and nominal).
  • Feature Engineering: Reduced and selected 175 key attributes for model training.
  • Comparison Studies: Benchmarked models against MEWS and traditional approaches.
  • Model Development:
    • Built models using Neural Networks (Keras/TensorFlow), K-Nearest Neighbors, Random Forest, SVM, and Naive Bayes.
    • Achieved the highest predictive accuracy (~76%) using a Keras-based Neural Network with optimized activation functions and dropout layers.

Architecture

Patient Data (ECG + Vitals)
↓
Data Preprocessing
↓
Feature Engineering
↓
KNN Classifier
Naive Bayes, SVM
Random Forest
Decision Trees
Neural Network (Keras)
Deep Learning Model
↓
Model Evaluation & Comparison
↓
Best Model Selection
↓
Cardiac Arrest Risk Prediction Output

Impact

  • Early Intervention: Predicting cardiac arrests hours in advance gives critical time to healthcare teams to intervene and save lives.
  • Enhanced Efficiency: Reduces unnecessary alarms and false positives compared to traditional scoring systems.
  • Scalable and Adaptable: Can be adapted for different hospital environments and integrated with various real-time patient monitoring devices.
  • Data-Driven Care: Moves critical care decisions toward proactive, data-backed actions rather than reactive measures.

Conclusion

Code Blue Situation exemplifies the transformative power of AI in healthcare. By combining robust patient monitoring data with machine learning algorithms, the system enhances clinical decision-making, saves lives, and pushes hospitals toward a smarter, more predictive future in critical care management.

Let's Connect!

I enjoy connecting with like-minded professionals passionate about technology, strategy, and impact. Feel free to reach out!

Chicago, IL
(312) 871-8022
k.teckchandani1703@gmail.com