Evaluation of MPEG-4 AAC Audio Codec Using EEG and EMG Signals for Use with DICOM® Neurophysiology

Modern neurophysiological recording technologies are producing large volumes of data. This study investigates the use of the Moving Picture Experts Group version 4 Advanced Audio Coding (MPEG-4 AAC) for compressing electroencephalography (EEG) and electromyography (EMG) signals. All EEG signals contained either seizures or interictal epileptiform discharges (IEDs), which are nonstationary patterns characterized by sharp transients. Differences between pairs of original single-channel signals and compressed/decompressed signals were explored using the Percentage Root Mean Square Difference (PRD), power spectral density (PSD) features, and clinical expert opinion of waveform morphology. PRD was used as the primary error measure and PSD features were explored across different frequency bands. The results showed substantial decline in signal quality based on human expert opinion for compressed then reconstructed EEG signals with PRD error greater than 15% and EMG signals with PRD greater than 1%. PSD analysis showed significant changes in the gamma and beta bands for both EEG and EMG signals. This study provides further evidence that standard audio codecs which use psychoacoustic models, such as MPEG-4 AAC, do not sufficiently support compression of neurophysiology waveforms.