Neurophysiology Signal Codecs for the DICOM® Standard: Preliminary Results.

Abstract

In this paper we report experimental results on compression of neurophysiology signals obtained as part of the standardization activities conducted by the Working Group 32 (WG-32) of the Digital Imaging and Communications in Medicine (DICOM). WG-32 focuses on extending the DICOM^® standard for clinical neurophysiology data exchange. With this aim, several compression techniques specifically devised for neurophysiology signals, as well as audio codecs, have been investigated and compared using real-world datasets. Moreover, a web-based application, named EEGnet and developed specifically for viewing and annotating electroencephalography (EEG) data, has been exploited for determining the maximum distortion that can be tolerated on neurophysiology signals. Through the EEGnet framework, eight neurologists, affiliated to different universities and medical centers, identified signals where they observed a clinically-significant difference. As one of the main results of our study, we found that, in the case of EEG signals, a percentage root mean square difference (PRD) of 5% can be accepted by clinicians and experts. On the other hand, all experts agreed that distortion is unacceptable when the PRD is greater than 15%. Finally, surprisingly enough, experimental results showed that audio codecs provide performance levels that, in some cases, are comparable to those of state-of-the-art algorithms specifically devised for compression of EEG signals.