SignalML for the INCF workshop report
Contents |
Reading multivarite time series
The most basic and primary challenge to the sustainability and sharing of electrophysiological and other time series data is the ability to properly read and graph the multivariate signals, stored in binary data files according to a variety of incompatible formats. It requires a sustainable description of the formats in a possibly generic form, other than e.g. compiled binary libraries, that may or may not function with the future versions of operating systems and compilers.
In the field of clinical electroencephalography, this probem was mentioned in "IFCN standards for digital recording of clinical EEG" [1]: (...) the vendor must make public the EEG file format and allow other vendors, or third party software vendors, to read and to translate the EEG record into another format readable by another vendor’s equipment. Clinical EEG data belongs to the health care providers or to the patients, not to the vendors.
We propose to store descriptions of these formats in a specially designed, yet generic XML-based language SignalML. Once written, an XML file containing SignalML description of a given format can be used by any SignalML-aware software to read and interpret the original data. Such an approch eliminates the loss of information and mutiplication of files, caused by conversions to any "standard" format. SignalML 1.0 was described in [2].
http://signalml.org hosts a wiki and a link to the newsgroup, set up for collaborative collection of information, needed for creation of such a definition of SignalML 2.0 specification, that will hopefully allow to describe all the relevant formats. To this aim, SignalML wiki hosts separate sections dedicated to the collection of existing relevant formats and discussion on the minimum set of parameters needed for their interpretation, as well as the structure needed for SignalML 2.0 to efficiently and elegantly serve its purpose. There are also sections with references to relevant ontologies and schemas existing in the field. We invite manufacturers of equipment and groups involved in relevant projects to contribute information, and take part in an open discussion on making SignalML possibly compatible with existing solutions, formats and Schemas.
Recommendations
In the discussions at the workshop, SignalML was initially accepted as a standard for the basic layer of metadescription of the time series data.
R1.
That the INCF establishes a Task Force coordinating the creation of the SignalML 2.0 standard, wthin a specified deadline.
R2.
That the INCF ensures worldwide visibility of the task from R1., so that all the potentialy relevant projects, standards, groups and manufacturers have the chance to contribute.
R3.
That the INCF ensures means for cooperation, parallell development and preserving compatibility between the SignalML and emerging ontologies and more complicated metadata schemes, possibly resulting from this workshop followups.
R4.
That the INCF promotes future development of SignalML-compliant software.
References:
[1] M.R. Nuwer et al. "IFCN standards for digital recording of clinical EEG", Electroencephalography and clinical Neurophysiology 106, pp. 259--261, 1998. http://www1.elsevier.com/homepage/sah/ifcn/pdf/106_259.pdf
[2] P.J. Durka and D. Ircha, "SignalML: metaformat for description of biomedical time series" Computer Methods and Programs in Biomedicine 76(3), pp. 253-259, 2004. http://signalml.org/SignalML/SignalML.html
