HRVFrame: Java-Based Framework for Feature Extraction from Cardiac Rhythm

1
HRVFrame Java-Based Framework for Feature
Extraction from Cardiac Rhythm

• Alan Jovic and Nikola Bogunovic
• Faculty of Electrical Engineering and Computing,
  University of Zagreb

2
Motivation

• Lack of agreement among experts upon the best
  heart rate variability (HRV) features used to
  classify cardiac arrhythmias
• Problem of results comparison
• Different datasets
• Different features -gt particularly problematic
  lots of proposed features!
• Different evaluation metrics
• What are the limits of HRV analysis for
  classification of cardiac rhythms and cardiac
  diseases?

3
Research goals

• Systematize existing HRV features
• Implement the features in a modular and easily
  upgradable framework
• Facilitate comparison of scientific work in
  biomedical time-series variability modeling
• Extract HRV features for automatic arrhythmia and
  heart diseases classification using freely
  available knowledge discovery platforms

4
Framework overview
Selection of features and features parameters
Feature calculation
Storing feature vectors in .arff file
Knowledge discovery platform
Cardiac rhythm records in textual format
Extracted feature vectors in .arff file
HRVFrame

• Input PhysioNet format (R peak times)
• Selection GUI-based selection of extraction
  parameters and features
• Calculation more than 30 linear time, frequency,
  time-frequency, and nonlinear features
• Output feature vectors in .arff file -gt Weka,
  RapidMiner

5
Comparison

• Other frameworks
• ECGLab - Matlab (ECGRR) linear, time-frequency,
  few nonlinear features
• KARDIA - Matlab (RR) linear, few non-linear
  features
• BioSig - C/Matlab (EEGRR) linear features
  only, aim is standardization of biomed. series
  processing tools and file formats
• Advantages of HRVFrame
• Implementation of numerous nonlinear features
• Preparation for data mining of cardiac disorders
  and arrhythmias
• Java-based, platform-independent
• Easily upgradeable to include additional novel
  HRV features
• Modifiable for analysis of other biomedical
  time-series
• Free for non-commercial purposes

6
Thank you!