TESTING DISCRIMINANT POWER OF SONIC REPRESENTATION OF ECG SIGNALS
Introduction: Sonic representation of the ECG has the potential to add supplementary information and improve signal classification. The present study aimed to test the discriminant power of sonic representation.
Methods: Testing methodology. A group of 20 subjects was asked to score the difference between various sonic representations. A testing session consisted of several comparisons of pairs of signals, presented as short “clips”. For each clip, the subject could both hear the sound and see the graphical representation, and had to score the perceived differences by both sonification and visualization. The scoring scale had a dichotomous structure, each with 2 steps: Irrelevant difference [scores 0-1] / Relevant difference [2-3]. Signals. All clips were ten seconds epochs from four ECG signals (sinusal rhythm SR, arrhythmia AR, congestive heart failure CHF, and sleep apnea SA - all from Physiobank).
Sonification parameters. Three sonification parameters have been tested:
- sonification procedure: continuous [A] vs. step wise [Q]
- time scale: fix [f] vs. variable [v]
- temporal lens: with four fold magnification [4x] vs. without lens [1x].
Therefore, eight combinations for each signal could be obtained.
Study design. The major types of tests have been designed:
- to compare, for each of the four signals, all sonification combinations [SC]
- to compare, within each SC, all 4 ECG types.
Results: Our preliminary results showed that:
- in all cases, except arrhythmia, the use of [4x] temporal lens was preferred
- the difference between [f] and [v] is minor in almost all cases (except CHF 4x)
- the difference between A and Q is minor for 1x, but visible for 4x
- the differences noticed by sonification went in parallel with those by visualization.
Conclusion: Sonification of biosignals is a method with high discriminant power, similar to visualization; the two methods can complement each other.