Master Thesis: Electrical and Computer Engineering Department, Rutgers, The State University of New Jersey,
New Brunswick, USA, 2005
fMRI is an imaging technique that is used to understand
brain functionality. Scans are taken at intervals as a patient performs some
mental tasks, resulting in hundreds of datasets. It is an increasingly popular
technique in fields ranging from medicine, psychology or even marketing and
economics. However, these images tend to be noisy and new packages are
constantly being developed to analyze and filter these large datasets. Because
of the large data size and many analysis parameters, comparisons between
results or between experiments are difficult. We present a visualization tool
that allows interactive comparison of different analyzed datasets. Such
analyzed datasets can be results of different analytic methods used in fMRI
analysis, on data from one or more subjects and/or one or more experiments. We
treat every analysis result as a functional clustering of voxels mapped into
brain space and employ visualization techniques to allow the user to
interactively explore the similarity and differences between the different datasets.
This can provide valuable insight into the data or the analysis methodologies
being studied. Thus, the tool can be used as a visualization interface of a
data mining engine and could also support a "query-by-example"
approach to fMRI data retrieval.

Downloads:
Screeshots:
Comparing the result of several different analysis methodologies applied to the same dataset.

Comparing the results of using different parameters of an analysis algorithm (k-means).

Identifying common activations in multiple subjects of the same experiment.

Brodmann Area mapping

"Query-by-example" data retrieval

Investigating similarity reported by other methods (Brodmann Vector similarity)

Setting thresholds on parametric maps.
Nicu D. Cornea: cornea@caip.rutgers.edu |
Jul 13, 2005 |