Exploring Medical Data by Sonification. Applications in Cardiology

Introduction. Sonification is defined as “the use of non-speech audio to convey information or perceptualize data”. For medical applications the data space comprises all types of data, but those related to dynamic processes (like biological signals – ECG, EEG etc.) or structured data (like molecular sequences – DNA, proteins) are most appropriate.
Methods. The central paradigm of sonification, called “parameter mapping” is to establish a correspondence between the parameters of the input data and the characteristics of the audio display as output. There are several sonification techniques classified on various taxonomic axes. We have used a “three level approach”: level A (acoustic) – with a continuous spectrum of frequencies, level S (sonic) – with a discrete scale of frequencies and durations and level M (musical) –adding harmony and rhythm. Most of our studies have been focused on the S level. Data: both our own recordings as well as data from Physiobank have been used – ECG, heart rate (HR) and pulse wave from both health subjects or patients with various CVD (arrhythmia, atrial fibrillation, ventricular extrasystolae, obstructive sleep apnea) or during exercise test. Algorithms: a simple natural mapping was chosen: sound pitch (frequency) to be a discrete function of normalized signal amplitude. For sound duration we have introduced tempolenses for dilating or compressing the signal, including an original tempolens with variable magnification, for ECG. Software: a MATLAB set of programs has been done for all operations – preprocessing (normalization), parameter mapping, tempolenses and producing the audio display.
Results. For ECG the results were modest: easy to recognize the obstructive episodes and some ischemic modifications (flat T wave), but insufficient discriminant power for other cases. Much better perception of deviations was shown by HR sonification and still better for HR monitoring during the exercise. For the latter application a warning software module was added and a couple of warning designs was tested.
Discussions refer to potential directions of medical applications and the limits of the method.
Conclusions Sonification is a new way to represent medical data, still insufficiently explored, but with high potential to complement the present list of investigation methods.
Up to 350 words. No references allowed. Abstracts may be submitted at a later date.