The NEMO Analysis Pipeline:
                     EEG Pattern Extraction and Ontology-based Classification
                                             Gwen A. Frishkoff1,2 and Robert M. Frank2
            1
                Department	
  of	
  Psychology	
  &	
  Neuroscience	
  Institute,	
  Georgia	
  State	
  University,	
  Atlanta,	
  Georgia,	
  U.S.A.	
  
                               2	
  NeuroInformatics	
  Center,	
  University	
  of	
  Oregon,	
  Eugene,	
  Oregon,	
  U.S.A.	
  
                                                                              	
  


*
    ABSTRACT                                                                           REFERENCES

    In	
  this	
  software	
  demonstration,	
  we	
  will	
  show	
                   Liu, H., Frishkoff, G., Frank, R. M., & Dou, D.
                                                                                         (2012, in press). Integration of Cognitive Neu-
how	
  formal	
  ontologies	
  can	
  be	
  used	
  to	
  label	
  in-­‐                 roscience Data: Metric and Pattern Matching
stances	
   of	
   neural	
   (ERP)	
   patterns	
   that	
   have	
                     across Heterogeneous ERP Datasets. Journal of
been	
   extracted	
   from	
   multiple	
   datasets	
   using	
   a	
                  Neurocomputing.
novel	
   pipeline	
   for	
   pattern	
   and	
   metric	
   extrac-­‐
tion	
  (see	
  Figure	
  1,	
  next	
  page).	
  The	
  entire	
  dem-­‐              Frishkoff, G., Frank, R., & LePendu, P. (2011).
onstration	
   will	
   last	
   ~15	
   minutes.	
   We	
   will	
   begin	
            Ontology-based Analysis of Event-Related Po-
                                                                                         tentials. Proceedings of the International Con-
with	
  a	
  5-­‐minute	
  introduction	
  to	
  ERP	
  data	
  from	
                   ference on Biomedical Ontology (ICBO'11).
several	
   cross-­‐laboratory	
   studies	
   of	
   word	
   com-­‐                    July 26-30, 2011. Buffalo, NY.
prehension.	
   This	
   overview	
   will	
   motivate	
   our	
  
demonstration	
   by	
   showing	
   that	
   ERP	
   data	
   are	
                   Frishkoff, G., Frank, R., Sydes, J., Mueller, K., &
complex	
   and	
   heterogeneous,	
   which	
   explains	
                              Malony, A. (2011). Minimal Information for
                                                                                         Neural Electromagnetic Ontologies (MI-
the	
  radical	
  challenge	
  of	
  making	
  valid	
  compari-­‐                       NEMO): A standards-compliant workflow for
sons	
   across	
   different	
   studies	
   within	
   our	
   do-­‐                   analysis and integration of human EEG. Stan-
main.	
  We	
  will	
  then	
  give	
  a	
  2-­‐minute	
  description	
                  dards in Genomic Sciences (SIGS), 5(2).
of	
  the	
  pipeline	
  for	
  analysis,	
  which	
  has	
  two	
  main	
  
components:	
   (1)	
   a	
   set	
   of	
   pattern	
   extraction	
                  Liu, H., Frishkoff, G.A., Frank, R., and Dou, D
(signal	
  decomposition,	
  temporal	
  segmentation)	
                                 (2010). Ontology-based mining of brainwaves:
                                                                                         sequence similarity technique for mapping al-
methods;	
   and	
   (2)	
   code	
   to	
   extract	
   a	
   variety	
   of	
          ternative descriptions of patterns in event-
simple	
   metrics	
   (e.g.,	
   min	
   and	
   max	
   intensity	
   at	
   a	
       related potentials (ERP) data. Proceedings of the
particular	
   electrode)	
   and	
   to	
   express	
   these	
                         14th Pacific-Asia Conference on Knowledge
summary	
   features	
   as	
   N-­‐triples,	
   which	
   are	
   sub-­‐                Discovery and Data Mining (PAKDD'10), 12
sequently	
   stored	
   in	
   RDF.	
   Finally,	
   we	
   demon-­‐                    pages.
strate	
   how	
   the	
   NEMO	
   ontology	
   can	
   be	
   used	
   to	
  
                                                                                       Frishkoff, G.A., Dou, D., Frank, R., LePendu, P.,
reason	
   over	
   these	
   data.	
   We	
   highlight	
   both	
   ex-­‐              and Liu, H. (2009). Development of Neural
pected	
  and	
  novel	
  findings	
  for	
  the	
  test	
  datasets	
                   Electromagnetic Ontologies (NEMO): Repre-
and	
   note	
   that	
   large-­‐scale	
   application	
   of	
   this	
                sentation and integration of event-related brain
method	
   could	
   lead	
   to	
   major	
   breakthroughs	
   in	
                    potentials. Proceedings of the International Con-
understanding	
   neurological	
   patterns	
   that	
   are	
                           ference on Biomedical Ontologies (ICBO09).
linked	
   to	
   sensory,	
   motor,	
   and	
   cognitive	
   pro-­‐                   July 24-26, 2009. Buffalo, NY.
cesses	
   in	
   neurologically	
   healthy	
   and	
   brain-­‐
injured	
  children	
  and	
  adults.	
  

*   To whom correspondence should be addressed: gfrishkoff@gsu.edu



                                                                                                                                                             1
Figure 1. [1] ERP pattern extraction. [2] Extraction of summary metrics. [3] Formatting of metrics in N-triples, with arguments defined by NEMO on-
tology classes and relations. [4] Capture of metadata about the experiment context (e.g., participants, measurement methods, experiment paradigm). [5]
Interchange between NEMO ontology an NEMO ERP database.



                                                                                                                       2