Biomedical Engineering Reference
In-Depth Information
In order to address problems arising from the multitude of formats and lack of
standard metadata, the BioSignalML project (www.embs.org/
TECHCOMM
/tc-cbap/
biosignal.html) is developing a framework for the exchange and storage of physio-
logical time-series data (biosignals). This framework:
•
encapsulates common features of biosignal file formats in an abstraction layer;
•
uses ontologies to define biosignal terms and attributes;
•
can be extended to incorporate domain-specific concepts and terms;
•
provides software tools and libraries that allow the use of disparate signal file
formats;
•
includes a repository component that allows recordings to be accessed using stan-
dard web software as a Linked Data resource.
The project is initially working with biosignal recordings associated with sleep re-
search. Polysomnograms contain a comprehensive range of physiological signals, in-
cluding those measuring brain, heart, breath, eye and muscle activity, often at differ-
ent sampling rates. Recordings are usually over several hours, with possibly several
recordings for a single patient. A variety of signal storage formats are used, includ-
ing proprietary formats, the European Data Format (EDF and EDF+; Kemp & Oli-
van, 2003), the Stanford Data Format (SDF; Sahul et al., 2003) and the WaveForm
DataBase (WFDB; Moody, 2011). BioSignalML will facilitate the use of biosignals
with physiological modelling languages such as CellML, with an HDF5-based file
format (Folk et al., 1999) specified for biosignal exchange and storage and a stream-
ing format used for real-time connections between simulation components.
BioSignal metadata
Some questions addressed are: What are general attributes of biosignals? What does
a particular signal represent? When was it recorded? How and by whom? What pro-
cessing has been applied? What is the purpose of a recording besides being a col-
lection of signals? Do other people know our meaning of the terms we have used to
describe properties? Current biosignal file formats usually have a limited number of
fields for metadata and these are usually only pertinent to the domain the format was
designed for. These metadata fields often contain free-format text with no controlled
vocabulary or ontology to specify content, leading to possible future ambiguity.
BioSignalML specifies a core set of terms and relationships for metadata. Some
of these are defined in the BioSignalML ontology; others are taken from existing
standard ontologies; additional domain-specific concepts and terms can easily be
added. Metadata statements about biosignals are made using RDF, with the Web
Ontology Language (OWL) being used to specify ontologies. The use of both RDF
and OWL allows for biosignal data to be integrated into the broader context of
the Semantic Web without restricting future applications and extensions. Some of
the ontologies applicable to biosignal annotation are well established international
standards; others are at different stages of development. Both the OBO Foundry
(www.obofoundry.org) and NCBO BioPortal (bioportal.bioontology.org) provide
repositories of publicly available biological and biomedical ontologies. Ontologies
used include:
