Faithful representation of separable distributions
Neural Computation
Information-theoretic approach to blind separation of sources in non-linear mixture
Signal Processing - Special issue on neural networks
On the eve of the 21st century: Statistical science at a crossroads
Computational Statistics & Data Analysis
Pattern Recognition with Fuzzy Objective Function Algorithms
Pattern Recognition with Fuzzy Objective Function Algorithms
Fuzzy Models and Algorithms for Pattern Recognition and Image Processing
Fuzzy Models and Algorithms for Pattern Recognition and Image Processing
Motor Area Activity During Mental Rotation Studied by Time-Resolved Single-Trial fMRI
Journal of Cognitive Neuroscience
Concepts in Magnetic Resonance: an Educational Journal - Functional magnetic resonance imaging
A quantitative comparison of functional MRI cluster analysis
Artificial Intelligence in Medicine
Editorial: Exploratory data analysis in functional neuroimaging
Artificial Intelligence in Medicine
A novel, direct spatio-temporal approach for analyzing fMRI experiments
Artificial Intelligence in Medicine
Hi-index | 0.00 |
The purpose of Exploratory Data Analysis (EDA) is to investigate and discover salient and novel features of complex, high-dimensional data. We describe a particular realization of EDA, the three-stage strategy EROICA (Exploring Regions Of Interest with Cluster Analysis), specifically designed to analyze functional MR neuroimaging data. The first stage consists of an Initial Partition of the data into three groups: a group of "trend" time-courses (TCs), a group of "potentially interesting" TCs, and a group that contains the remaining, putative "noise" TCs. The initial grouping is achieved by first normalizing (scaling) the TCs, followed by selection procedures based on specific "trend" and "noise" tests. The second stage is the Principal Partition, where fuzzy clustering analysis (FCA) is applied to the group of "potentially interesting" TCs. The third stage, Significance Testing, "validates" the second-stage results by first removing those TCs from the original clusters that fail special statistical tests. and then by attempting to allocate to the clusters some of the initially excluded "trend" and "noise" TCs. We assessed the consequences of this three-stage strategy on the quality of the clustering results. We show that employing this strategy both improves results relative to clustering that did not use the initial partitioning, and also speeds up execution significantly. We report detailed analyses on several phantom datasets and on a multi-slice, real fMRI dataset. Based on detailed studies carried out on sixteen fMRI datasets, the execution time of EROICA scales sublinearly both with T (scans) and N (number of TCs). We propose robustness (noise resistance, reproducibility) flexibility/versatility, and speed as the three major requirements that any practically viable EDA method ought to satisfy. We show that the EROICA process, and EvIdent®, its software implementation, fulfill these requirements.