Music Information Research (MIR) Techniques for Cardiac Data Analysis

Elaine Chew
CNRS-STMS (IRCAM)


Abstract

The explosion in digital music information has led to new research communities and computer algorithms and technologies for automating the analysis of musical structures such as rhythm, pitch (fundamental frequency), and temporal structures. These same kinds of features exist in cardiovascular data and manifest as rhythm, dominant frequency, and episodic form. I argue that because music and heart signals share many common features, the techniques developed for mathematical representation of music and for computational music structure analysis transfer readily to, but has yet to be exploited in, the analysis of cardiac information. Here, I demonstrate the ease of this transference through specific examples so as to promote this cross-domain exchange. First, computing tools developed for music data analysis, visualization, and annotation can be applied directly and meaningfully to electrocardiographic recordings. Second, music notation can accurately capture the rhythmic variations in arrhythmia, thus providing an avenue for the transfer of music information research (MIR) rhythm analysis techniques to electrocardiographic recordings. Finally, an prominent aspect of MIR is the discovery of musical structure, which at the rudimentary level refers to sectional, repeating forms; the episodic and repeating structures of many arrhythmias lend themselves to the same methods for characterizing and analyzing temporal structure. Such musical structures at multiple time scales provide descriptors for MIR tasks such as similarity assessment and genre classification, which have parallels in the classification and stratifying of cardiac conditions. Thus, many opportunities exist to port MIR computing techniques to applications in cardiac information research.