| Article Access Statistics|
| Viewed||2721 |
| Printed||195 |
| Emailed||0 |
| PDF Downloaded||70 |
| Comments ||[Add] |
Click on image for details.
|Year : 2011 | Volume
| Issue : 3 | Page : 327-328
Spatial normalization of BOLD fMRI data in cases of qualitative gross neuroanatomical changes resulting from pathology
Basant K Puri
Department of Imaging, Hammersmith Hospital, Imperial College London, United Kingdom
|Date of Submission||23-Mar-2011|
|Date of Decision||23-Mar-2011|
|Date of Acceptance||23-Mar-2011|
|Date of Web Publication||7-Jul-2011|
Basant K Puri
Department of Imaging, Hammersmith Hospital, Du Cane Road, London W12 0HS, England
Source of Support: None, Conflict of Interest: None
|How to cite this article:|
Puri BK. Spatial normalization of BOLD fMRI data in cases of qualitative gross neuroanatomical changes resulting from pathology. Neurol India 2011;59:327-8
In this month's issue of Neurology India, Nagata and colleagues present a paper entitled "Non-normalized individual analysis of statistical parametric mapping for clinical fMRI," in which 10 patients with brain tumours underwent pre-surgical blood oxygen level-dependent (BOLD) fMRI while performing, in turn, motor and verbal tasks.  This paper is instructive for those considering the clinical use of fMRI, as not all the patients showed significant activation of the hand area of the contralateral primary motor cortex (M1) during a hand grasping motor task, and not all the patients showed significant frontal lobe language area activation during silent word generation. Yet, numerous lesion and stimulation studies have left modern neurology in no doubt that activation of these areas of the brain would almost certainly have been expected to have occurred during performance of these tasks. So, assuming that the non-activation results were not the consequence of highly unusual central neurophysiological processes occurring in the non-activating patients, the cause(s) of the negative results must be sought elsewhere.
The authors suggest that the culprit is likely to lie with the pre-processing of the BOLD fMRI data. This had been conducted by them using the commonly employed program SPM, which in the context of neuroimaging variously stands for statistical parametric mapping, statistical parametric map and, from the world of electroencephalography, significance probability mapping. In particular, the authors point the finger at the spatial normalization procedure. This is an operational component of BOLD fMRI data analysis in which the images are translated and morphed (or "warped") so that they conform to some idealized or standard brain, thereby allowing the implementation of voxel-based analysis of the functional imaging data.  This procedure was first employed for positron emission tomography (PET) data, and is now routinely used with SPM when processing fMRI data.
The key creator of the statistical parametric mapping procedures used in PET and fMRI analyses, Professor Karl J. Friston, has pointed out the difficulties that can occur in the spatial normalization of brains that have gross anatomical pathology, but it would appear that this message has not yet been taken fully on board by neurologists and neurosurgeons. This may be partly because it tends to be confined to discussions centered on the mathematics and physics of BOLD fMRI data pre-processing. Friston considers two types of pathology.  The first, such as cortical atrophy, can be considered to be primarily a quantitative change in the amount of a particular tissue compartment, and this should not usually prove problematic when carrying out spatial normalization, since spatial location in reference to an anatomical template such as that of the Montreal Neurological Institute (MNI) should by and large be either unaffected or easily dealt with by construction of a disease-specific template.  It is in respect of the second type of pathology, however, that particular problems can arise. These are considered by Friston as being qualitative changes in neuroanatomy involving the insertion or deletion of tissue compartments, as would occur in the case of brain tumours as in the case series by Nagata and colleagues.  Pathological lesions such as brain tumours can introduce bias into the spatial normalization process; the mathematical generative model for this algorithm is a canonical image or template which is distorted to give rise to the subject-specific image (with spatial normalization inverting this model) but, as Friston points out, the generative model does not have a lesion.  With the increasing use of clinical fMRI, the paper by Nagata and colleagues does the clinical community a great service by drawing attention to the need to exercise extreme care if attempting to make clinical decisions (e.g. neurosurgical planning) based on fMRI data from patients with, for example, tumours or cerebral ischemic areas resulting from stroke.
The question naturally arises as to whether anything can be done to mitigate the adverse effects of qualitative cerebral pathological lesions on spatial normalization during BOLD fMRI pre-processing. One possibility is suggested by Friston, who suggests that special precautions be taken: "These usually involve imposing constraints on the warping to ensure that the pathology does not bias the deformation of undamaged tissue."  Thus, lesion masking could be employed during this pre-processing stage. However, the paper in this issue of Neurology India suggests the possibility of another, simpler, approach, namely avoiding spatial normalization altogether. As the authors conclude: "application of SPM (version 8) analysis to non-normalized individual data for the purpose of performing pre-operative fMRI is a useful method for investigation of functional localization."
| » References|| |
|1.||Nagata T, Tsuyuguchi N, Uda T, Ohata K. Non-normalized individual analysis of statistical parametric mapping for clinical fMRI. Neurol India 2011;59:339-43. |
|2.||Friston KJ. Analysing brain image: Principles and overview. In: Frackowiak RS, Friston KJ, Frith CD, Dolan RJ, Mazziotta JC, editors. Human brain function. London: Academic Press; 1997. p. 25-41. |
|3.||Friston K. Statistical parametric mapping. In: Friston KJ, Ashburner JT, Kiebel SJ, Nichols TE, Penny WD, editors. Statistical parametric mapping: The analysis of functional brain images. London: Academic Press; 2007. p. 10-31. |