Recent Changes - Search:

PmWiki

pmwiki.org

edit SideBar

BooksteinVBM

Voxel-Based Morphometry Should Not Be Used With Imperfectly Registered Images

F. L. Bookstein, Neuroimage 14 p. 1454-1462, 2001

In 2000, John Ashburner and Karl Friston proposed a method for the groupwise comparison of local concentrations of grey matter between two groups of subjects [1], based on the voxel-wise Worsley approach to Gaussian random fields of differences. The method, known as voxel-based morphometry (VBM) grew rapidly in popularity, prompting Fred Bookstein to publish this paper in the "Comments and Controversies" section of Neuroimage, pointing out some of the inherent failings of the method. In order to produce a consistent coordinate system in which to perform the analysis, VBM requires the images under analysis to be registered. Bookstein analysed the statistical effects of imperfect registration, showing that VBM is only statistically valid when the registration is perfect, an impossible requirement to achieve, and thus via. reductio ad absurdam that VBM can never be statistically valid. The original authors' reply to Booksteins criticism was published alongside this paper [2], and broadly agrees with many of his points. Other authors have gone further, pointing out that, since cortical topology varies between individuals, voxel-level differences cannot be equated to meaningful volume changes (see e.g. [3]), and that, since VBM results in thousands of parameters per comparison, applying a suite of statistical analysis techniques to sub-sets of the results can "prove" almost any hypothesis.

The take-home message of this paper is implied by the somewhat subtle title: VBM can never be statistically valid and therefore should not be used.

References:

  1. J. Ashburner and K.J. Friston, Voxel-based Morphpmetry - The Methods, Neuroimage 11 p. 805-821, 2000.
  2. J. Ashburner and K.J. Friston, Why Voxel-based Morphometry should be Used, Neuroimage 14 p.1238-1243, 2001
  3. A Critical Analysis of VBM, Tina Memo no. 2003-011, N. Thacker, 2003.

PAB 8/3/2005

Edit - History - Print - Recent Changes - Search
Page last modified on March 05, 2012, at 03:59 PM