Elsevier

NeuroImage

Volume 17, Issue 2, October 2002, Pages 583-591
NeuroImage

Regular Article
Cluster Significance Testing Using the Bootstrap

https://doi.org/10.1006/nimg.2002.1223Get rights and content

Abstract

Many of the statistical methods currently employed to analyze fMRI data depend on a response template. However, the true form of the hemodynamic response, and thereby the response template, is often unknown. Consequently, cluster analysis provides a complementary, template-free method for exploratory analysis of multidimensional fMRI data sets. Clustering algorithms currently being applied to fMRI data separate the data into a predefined number of clusters (k). A poor choice of k will result in erroneously partitioning well-defined clusters. Although several clustering algorithms have been successfully applied to fMRI data, techniques for statistically testing cluster separation are still lacking. To address this problem we suggest a method based on Fisher's linear discriminant and the bootstrap. Also introduced in this paper is a measure based on the projection of multidimensional data from two clusters onto the vector, maximizing the ratio of the between- to the within-cluster sums of squares. The resulting one-dimensional distribution may be readily visualized and used as a heuristic for estimating cluster homogeneity. These methods are demonstrated for the self-organizing maps clustering algorithm when applied to event-related fMRI data.

References (43)

  • C. Chatfield et al.

    Introduction to Multivariate Analysis

    (1980)
  • V. Cherkassky et al.

    Learning from Data: Concepts, Theory, and Methods

    (1998)
  • B. Efron et al.

    A leisurely look at the bootstrap, the jackknife and cross-validation

    Am. Stat.

    (1983)
  • B. Efron et al.

    An Introduction to the Bootstrap

    (1998)
  • L. Engelman et al.

    Percentage points of a test for clusters

    J. Am. Stat. Assoc.

    (1969)
  • H. Fischer et al.

    Neural network-based analysis of MR time series

    Magn. Reson. Med.

    (1999)
  • P.T. Fox et al.

    Focal physiological uncoupling of cerebral blood flow and oxidative metabolism during somatosensory stimulation in human subjects

    Proc. Natl. Acad. Sci. USA

    (1986)
  • K. Friston et al.

    Spatial registration and normalization of images

    Hum. Brain Mapping

    (1995)
  • K.J. Friston et al.

    Statistical parametric maps in functional imaging: A general linear approach

    Hum. Brain Mapping

    (1995)
  • Golay, X, Kollias, S, Meier, S. D, and, Boesiger, P. 1996, Optimization of a fuzzy clustering technique and comparison...
  • D. Hawkins

    Identification of Outliers

    (1980)
  • Cited by (22)

    • Enhanced limbic/impaired cortical-loop connection onto the hippocampus of NHE rats: Application of resting-state functional connectivity in a preclinical ADHD model

      2017, Behavioural Brain Research
      Citation Excerpt :

      Resampling is a powerful technique, allowing to assess data reliability and reproducibility and to identify, through a significance threshold, the brain “active” voxels. By using the bootstrap or jack-knife method, it is possible, for fMRI analyses in particular, to overcome the problem of the test-retest method [28] and/or to assess significance of activation clustering [29]. Present bootstrap resampling is an alternative, computer-intensive method that provides a strong control of the family wise control rate, which is more conservative than the false discovery rate among the combined Type I error rate [30].

    • False positive control of activated voxels in single fMRI analysis using bootstrap resampling in comparison to spatial smoothing

      2013, Magnetic Resonance Imaging
      Citation Excerpt :

      The jackknife method [11,12] was used to compute the reliability and confidence intervals of fMRI parameters during bilateral finger tapping. Auffermann et al. [14] applied the bootstrap method to assess the significance of the self-organizing maps in clustering algorithm applied on event-related data. In another study, the bootstrap analysis was used to investigate the stability of the clusters found by K-means algorithm for resting-state networks [15].

    • Uncertainty estimations for quantitative in vivo MRI T<inf>1</inf> mapping

      2012, Journal of Magnetic Resonance
      Citation Excerpt :

      The variation found in fits of these datasets is a measure of the uncertainty of the fit. Wild bootstrapping has already been applied successfully in the assessment of diffusion tensor imaging (DTI) parameters [3–8] and the determination of functional MRI resting state network nodes [9]. T1 fitting and conventional bootstrapping (i.e. random permutations from repeated measurements) were used in spectroscopic imaging to assess the standard error of T1 fitting [10].

    • Dissecting cognitive stages with time-resolved fMRI data: a comparison of fuzzy clustering and independent component analysis

      2007, Magnetic Resonance Imaging
      Citation Excerpt :

      Other studies compared FCM to alternative techniques in the field of fMRI, like correlation [18] or PCA [19]. Several studies dealt with other aspects, like the cluster validity problem [20–22] or the influence of higher fields on FCM results [23]. Clustering on features extracted from the fMRI time series at each voxel was also investigated [24,25].

    View all citing articles on Scopus
    View full text