Comments and ControversiesWhy voxel-based morphometric analysis should be used with great caution when characterizing group differences
Section snippets
Diagnosis vs. investigation of spatial maps of morphological group differences
It is often believed that statistical decision methods are only suitable for recognition, as for example in clinical diagnosis, and not for scientific investigation of disease or other processes leading to groups that differ morphologically. Karl and John reiterate this issue in their response. However, this is not necessarily the case (see, for example, Lao et al., 2003). In particular, the vector w in Fig. 2 can be used to form a “difference image”, that is, a spatial map of the regions that
Focusing on morphological characteristics that matter
One of the strengths of some pattern classification techniques, including support vector machines (Burges, 1998), is that they focus on the interface between two groups, for example, on the boundaries drawn in Fig. 2, Fig. 3, and not on samples that are far away from the dividing boundary Lao et al., 2003, Golland et al., 2001. This allows them to zoom into the subtleties of group differences, and factor out morphological characteristics that are related to variation within each group and are
Classification vs. regression
In their response, Karl and John make an important point, using as an example the relationship between the parahippocampal gyrus volume and the number of tri-nucleotide repeats in fragile X. The point is that, whenever a relationship between two (continuous in this case) variables is sought, recognition models are not appropriate. Of course, there are regression implementations of support vector machines, which are used for describing such relationships (Hastie et al., 2001). The underlying
Acknowledgements
The author would like to thank Drs. Bilge Karacali and Dinggang Shen for helpful discussions, and grant support by NIH-R01AG14971 and NIH-N01-AG-3-2124.
References (18)
- et al.
Voxel-based morphometry: the methods
NeuroImage
(2000) ‘Voxel-based morphometry’ should not be used with imperfectly registered images
NeuroImage
(2001)- et al.
A unified statistical approach to deformation-based morphometry
NeuroImage
(2001) - et al.
Voxel-based morphometry using the RAVENS maps: methods and validation using simulated longitudinal atrophy
NeuroImage
(2001) - et al.
Automatic differentiation of anatomical patterns in the human brain: validation with studies of degenerative dementias
NeuroImage
(2002) - et al.
Structural gray matter differences between first-episode schizophrenics and normal controls using voxel-based morphometry
NeuroImage
(2002) - et al.
Voxel-based morphometric analysis of gray matter in first episode schizophrenia
NeuroImage
(2002) - et al.
Spatial pattern analysis of functional brain images using partial least squares
NeuroImage
(1996) - et al.
Evaluation of voxel-based morphometry for focal lesion detection in individuals
NeuroImage
(2003)
Cited by (329)
Effects of randomness in the development of machine learning models in neuroimaging studies of schizophrenia
2023, Schizophrenia ResearchOptimal transport features for morphometric population analysis
2023, Medical Image AnalysisA review on Alzheimer's disease classification from normal controls and mild cognitive impairment using structural MR images
2023, Journal of Neuroscience MethodsRapid volumetric brain changes after acute psychosocial stress
2023, NeuroImageUnpacking the Biases That Shape the Apparent Foci in the Meta-analysis of Voxel-Based Neuroimaging Studies
2022, Biological Psychiatry