fissure, and (5) atrophy of the corpus callosum. These
multivariate results are consistent with the SPM
between-group differences presented previously and
with other common findings of patients who have
developed the pathology [12], such as the enlargement
of the ventricular system. Therefore, the use of the
multivariate approach has allowed not only the
simultaneous identification of localised between-group
differences but also distributed ones that are often
measured separately in the voxel-wise statistical
approaches.
4. Conclusion
We have presented a general PCA+MLDA
multivariate linear framework to identify and analyse
the most discriminating hyper-plane separating two
populations. The statistical analysis generates a
detailed description of the neuroanatomical changes
due to diseases and can facilitate the studies of the
brain disorders, such as Alzheimer, through
understanding of the captured anatomical changes.
The idea of using PCA plus an LDA-based
approach to discriminate patterns of interest is not new.
In this paper we have added to the functionality of this
approach the following important points for medical
image analysis. The use of full rank version of PCA
transformation matrix that allows valuable low
representation of high dimensional data, providing
optimal reconstruction of the most discriminant
intensity features without adding any artefacts on the
patterns when mapped back into the original image
space. By selecting a slightly biased within-class
scatter matrix composed of the most informative
dispersions we resolve not only the LDA singularity
problem but also we stabilise the maximisation of the
Fisher’s criterion on limited sample size problems.
The conceptual and mathematical simplicity of the
approach, which pivotal step is spatial normalisation,
involves the same operations irrespective of the
complexity of the experiment or nature of the data,
giving multivariate results that are easy to interpret.
Although the approach has been demonstrated in
two-class problems, it is extensible to several classes.
The only difference is the visual analysis of the
discriminant features, which might be performed pair-
wisely.
References
[1] G. F. Busatto, G. E. J. Garrido, O. P. Almeida, C. C.
Castro, C. H. P. Camargo, C. G. Cid, C. A. Buchpiguel, S.
Furuie, and C. M. Bottino, “A voxel-based morphometry
study of temporal lobe gray matter reductions in
Alzheimer’s disease”, Neurobiology of Aging, 24, pp.
221-231, 2003.
[2] L. Chen, H. Liao, M. Ko, J. Lin, and G. Yu, “A new
LDA-based face recognition system which can solve the
small sample size problem”, Pattern Recognition, 33 (10),
pp. 1713-1726, 2000.
[3] K. J. Friston and J. Ashburner, “Generative and
recognition models for neuroanatomy”, NeuroImage, vol.
23, pp. 21–24, 2004.
[4] K. J. Friston, A. P. Holmes, K. J. Worsley, J. P. Poline,
C. D. Frith, and R. S. J. Frackowiak, “Statistical
Parametric Maps in Functional Imaging: A General Linear
Approach”, Human Brain Mapping 2, pp. 189–210, 1995.
[5] K. Fukunaga, Introduction to Statistical Pattern
Recognition, second edition. Boston: Academic Press,
1990.
[6] P. Golland, W. Grimson, M. Shenton, and R. Kikinis,
“Detection and Analysis of Statistical Differences in
Anatomical Shape”, Medical Image Analysis, vol. 9, pp.
69-86, 2005.
[7] C. D. Good, I. S. Johnsrued, J. Ashburner, R. N. Henson,
K. J. Friston, and R. S. Frackowiak, “A voxel-based
morphometric study of ageing in 465 normal adult human
brains, NeuroImage, vol. 14, pp. 21-36, 2001.
[8] Z. Lao, D. Shen, Z. Xue, B. Karacali, S. Resnick, and C.
Davatzikos, “Morphological classification of brains via
high-dimensional shape transformations and machine
learning methods”, NeuroImage, vol. 21, pp. 46-57, 2004.
[9] J.C. Mazziotta, A.W. Toga, A. Evans, P. Fox, and J.
Lancaster, “A probabilistic atlas of the human brain:
Theory and rationale for its development”, NeuroImage,
vol. 2, pp. 89-101, 1995.
[10] C. E. Thomaz and D. F. Gillies, “A Maximum
Uncertainty LDA-based approach for Limited Sample
Size problems - with application to Face Recognition”, in
Proceedings of SIBGRAPI’05, IEEE CS Press, pp. 89-96,
2005.
[11] C. E. Thomaz, D. F. Gillies and R. Q. Feitosa, “A New
Covariance Estimate for Bayesian Classifiers in Biometric
Recognition”, IEEE Transactions on Circuits and
Systems for Video Technology, vol. 14, no. 2, pp. 214-
223, 2004.
[12] P. M. Thompson, J. Moussai, S. Zohoori, A. Goldkorn,
A. Khan, M.S. Mega, G. Small, J. Cummings, and A. W.
Toga, “Cortical variability and asymmetry in normal
aging and Alzheimer’s disease”, Cerebral Cortex, vol. 8,
pp. 492-509, 1998.
[13] J. Yang and J. Yang, “Why can LDA be performed in
PCA transformed space? ”, Pattern Recognition, vol. 36,
pp. 563-566, 2003.
[14] H. Yu and J. Yang, “A direct LDA algorithm for high
dimensional data – with application to face recognition”,
Pattern Recognition, vol. 34, pp. 2067-2070, 2001.