Pulse AI

Our AI-powered, heart arrhythmia detection watches over every heartbeat. Silently guarding the moments and the people you cherish—because every heartbeat matters.

Comment

Pulse AI

Inspiration

Every year, thousands of people face serious health complications because they lack real-time insights into their heart health. Pulse AI began with a simple yet powerful vision: to empower individuals and healthcare providers with continuous, non-invasive, and intelligent monitoring of heart rhythms. Inspired by advancements in wearable ECG technology, we imagined an accessible, streamlined solution that captures, processes, and analyzes ECG data to proactively detect heart-related risks.

What it does

Pulse AI is a wearable ECG sensor and AI-powered monitoring platform that:

  • Captures high-frequency ECG (electrocardiogram) signals using a compact wearable device.
  • Sends ECG data securely to Firebase Realtime Database.
  • Filters and processes ECG signals in a Python-based backend to accurately calculate key metrics like beats per minute (BPM).
  • Utilizes an AI-driven model to identify cardiac anomalies and predict risks of arrhythmias.
  • Provides real-time alerts to users or healthcare professionals via an intuitive mobile app interface, enabling timely interventions.

How we built it

  • Hardware:

    • ESP32 microcontroller paired with the SparkFun AD8232 ECG sensor module to capture heart signals every 25 milliseconds.
    • Securely sends raw ECG data (voltage and timestamps) every 10 seconds.

  • Backend:

    • Firebase Realtime Database for securely storing incoming ECG data in JSON format.
    • Custom Python scripts handle filtering, BPM calculation, and data preprocessing for the AI model.

  • AI Model:

    • Logistic regression model (extendable to deep learning) trained on historical ECG datasets to detect arrhythmias and cardiac risk factors.

  • Mobile App:

    • Flutter-based mobile app provides real-time data visualization, heart rate trends, risk assessments, and actionable alerts.

Challenges we ran into

  1. Reliable Data Acquisition:
    Capturing consistent, high-quality ECG signals without interference was challenging, particularly during continuous real-time data streaming.

  2. Filtering ECG Signals and Noise Management:
    ECG data contains inherent noise and artifacts; implementing robust filtering (local maxima detection, thresholding, and refractory periods) required extensive fine-tuning.

  3. Real-Time Data and AI Integration:
    Achieving low-latency ECG signal processing and AI analysis demanded optimized backend computation and data transmission.

  4. Security and Privacy Compliance:
    Handling sensitive health data responsibly required secure Firebase authentication methods and strict adherence to data privacy standards, even within rapid development cycles.

Accomplishments that we're proud of

  • Fully Operational Prototype:
    Built a complete ECG monitoring pipeline—from hardware data collection to real-time visualization and AI-driven alerts.

  • High-Accuracy BPM Computation:
    Achieved precise BPM calculation closely matching clinical-grade ECG systems.

  • Proactive AI-based Detection:
    Integrated a working AI model capable of detecting potential arrhythmias early, providing actionable user alerts.

  • User-Friendly Mobile App:
    Designed an intuitive, visually appealing mobile interface clearly communicating heart health insights and risks.

What we learned

  • Wearable ECG Integration and Embedded Systems:
    Gained insights into challenges with ESP32 configuration and coding as well as ECG sensor calibration, data quality management, and continuous data capture.

  • Real-Time Database Management (Firebase):
    Learned efficient data structuring, concurrency management, and latency optimization for real-time health data.

  • Signal Processing Techniques:
    Expanded our understanding of noise reduction methods, accurate physiological metric extraction, and real-world signal processing.

  • Healthcare Machine Learning:
    Developed deeper appreciation for the complexity of building robust, accurate AI pipelines specifically tailored to medical data.

What's next for Pulse AI

  1. Hardware Enhancement:
    Miniaturize and refine the ECG device for enhanced comfort, long-term usability, and clinical accuracy.

  2. Advanced Machine Learning:
    Implement sophisticated deep learning architectures (LSTM, CNN) for improved precision, personalization, and predictive arrhythmia detection.

  3. Automated Cloud Infrastructure:
    Leverage cloud functions to fully automate data handling, signal processing, and alerting mechanisms, increasing scalability.

  4. Compliance and Certification:
    Pursue HIPAA compliance and FDA certification, positioning Pulse AI as a medical-grade monitoring solution.

  5. Clinical Validation and Trials:
    Conduct rigorous clinical trials to validate Pulse AI’s effectiveness and accuracy across diverse user demographics and medical contexts.

Built With

Share this project:

Updates