Haishan Liu1, Gwen Frishkoff2, Robert Frank1, Dejing Dou1
1 University of Oregon
2 Georgia State University

ERP (Event-Related Potentials): a direct
measure of neural activity
 Lack of meta-analysis across experiment
 NEMO (Neural ElectroMagnetic Ontologies) for
data sharing and integration

Goal of the presented study
 Mapping alternative sets of ERP spatial and
temporal measures
Alternative sets of ERP metrics

Semi-structured data

Semi-structured data

Uninformative column
headers

Semi-structured data

Uninformative column
headers

Numerical values






Value vector to point-sequence curve
Use clustering analysis to identify
subsequences
Assign meaningful labels to subsequences
Align subsequences across different datasets
Sequence post-processing
Evaluate similarities of the curves to
determine the mapping (cross-spatial join)
Point-sequence curve
Meaningful
Cluster labels
labels

Process all pointsequence curves

Calculate Euclidean
distance between
sequences in the
Cartesian product set
(Cross-spatial join)
Metric Set1
Metric Set2
●●●

the two datasets contain the same or similar
ERP patterns


the two datasets contain the same or similar
ERP patterns
1-to-1 mapping between metrics

the two datasets contain the same or similar
ERP patterns
1-to-1 mapping between metrics

Minimum sum of distances

4.01 + 3.74 > 4.08 + 3.57
Gold standard mapping falls along the diagonal cells
Wrong Mappings. Precision = 9/13

3-Factor design of experiment data (Fully
factorial: 2 x 2 x 2)
 2 simulated “subject groups” (samples)
▪ SG1 = sample 1
▪ SG2 = sample 2
 2 data decompositions
▪ tPCA = temporal PCA decomposition
▪ sICA = spatial ICA decomposition
 2 sets of alternative metrics
▪ m1 = metric set 1
▪ m2 = metric set 2
Overall Precision: 84.6%

We describe a method for identifying
mappings between alternative sets of ERP
measures.
 Use of an ontology to assign meaningful labels to
ERP patterns
 Application of sequence similarity search in
discovering mappings across alternative metrics
 Extension of the instance-level approach in
schema matching
 Articulation of a global minimum heuristic

Assuming datasets to contain the dissimilar
ERP patterns

From 1-to-1 mapping to complex mapping

Approximation of the global minimum
heuristics

More tests on real-world data
Questions and comments?
Please contact Haishan Liu ([email protected])