Skip to main content

Application of Deep Learning Techniques for Real-Time ECG Interpretation on Wearable Smart Devices

Primary supervisor

Jackie Rong

Recent advancements in wearable technology have enabled continuous health monitoring, significantly expanding the capabilities of devices like the Apple Watch, Fitbit, and Samsung Watch in medical diagnostics. Among these, Electrocardiography (ECG) interpretation is a critical function, traditionally requiring expert analysis. This project proposes the use of deep learning (DL) algorithms to automate ECG interpretation on these devices, enhancing diagnostic accuracy and accessibility. The integration of DL could revolutionize how cardiac abnormalities are detected in everyday settings, potentially reducing emergency response times, and improving preventative healthcare.

A preliminary review of existing studies highlights the application of algorithms like Convolutional Neural Networks (CNNs) and Recurrent Neural Networks (RNNs) in ECG interpretation. Recent research demonstrates promising results in accuracy and speed using deep learning frameworks. However, there is a gap in research specifically targeting accurate interpretation on low-power wearable devices. This project will explore scalable and efficient DL architectures that operate within these constraints.

Student cohort

Double Semester

Aim/outline

Objectives

  • To extract ECG data from heart rhythm reports generated by smartwatches.
  • To develop a deep learning model that accurately interprets ECG data obtained from smartwatches.
  • To evaluate the model's effectiveness in detecting various cardiac abnormalities compared to traditional monitoring systems.
  • To assess the model's integration potential including identifying specific rhythm for individual rhythm analysis

Expected Outcomes

  • A validated deep learning model that can be deployed on smartwatches to provide real-time, accurate ECG interpretations.
  • Performance benchmarks for various model architectures, including processing time and diagnostic accuracy.
  • A framework for future research into other physiological signal interpretations on wearable devices.

Required knowledge

  • Solid knowledge about machine learning and deep learning.
  • Programming skills with Python, R, or MATLAB.
  • Experience with data processing and image processing is preferred.