Michael H. Chappell

University of Canterbury

Private Bag 4800

Christchurch, 8140
New Zealand
+64 3 364 2404

Michael Chappell's Website
CV
NIH Biosketch
NSF Biosketch

Keywords:
DTI, fMRI, pediatric MRI

Statement:
I would like to become involved in a cutting-edge development in diffusion tensor analysis called log Euclidean analysis. This overcomes many of the present shortcomings of tensor analysis such as being unable to combine the full tensors from different voxels in the same subject, or from the same voxels in different subjects. In conventional tensor analysis, only scalar derivatives of the tensors can be compared, which loses important directional information. This problem arises when the tensor is analysed in Euclidean (“flat”) space, whereas the log Euclidean approach analyses the tensor in Riemannian (“curved”) space, allowing it to retain more of its inherent features. Pediatric MRI, and especially fMRI and DTI, are rapidly growing areas of research. They pose two particular problems that I would alike to address. Firstly, in fMRI, an understanding of the expected hemodynamic response (HDR) to a stimulus is an essential part of the analysis. While it is known that this response is different with children than adults, there is no clear understanding of how to include these differences in the analysis. Failure to do so, however, will almost certainly lead to erroneous results. Standard fMRI analysis software such as SPM and AFNI use a mathematical model of the assumed adult HDR in the general linear model to discover areas of activation. This model assumes the HDR is the same over the whole brain. Pediatric fMRI requires a different approach. Little work has been done on this critical area. I would like to investigate a voxel-specific, or at least region-specific, “recognition model” approach. This non-parametric approach would obtain the HDR from the data themselves rather than using a mathematical model, and would allow variation in the HDR over the brain. The second problem area of pediatric MRI needing urgent research is obtaining suitable templates and normalization procedures to enable inter-subject comparisons. I would like to be involved in specialised registration techniques of infant brains aimed at forming infant brain atlases. Finally, if suitable results can be obtained from fMRI and DTI scans of the same child, then pediatric connectivity is an obvious study. This will be important both to compare with adult connectivity, and to lay the groundwork for a longitudinal studies comparing developmental connectivity changes in the same subject. I am also interested in exploring the possibility of using a spatial mapping model from ecological research to interpret voxel based results. This could facilitate a move away from binary hypothesis testing, with all the problems associated with visually presenting the results of such analysis. An index of clustering will give a measure of how significant a particular group of voxels are based on their regression statistic or t-statistic, and the extent of the cluster. The advantage of this approach over present methodologies is that it does not use an arbitrary threshold, nor have to make allowances for multiple comparison problems.

I sincerely apologize that we have a long-standing and unresolved problem that users are unable to modify the database contents using their logins and passwords. I hope that we can fix this problem soon. In the meantime, I will try to do this manually as best I can.

-- Mark Cohen

We encourage you to add your own name to this list (without obligation or cost...). You may use this list to let others know of your interests, your contact information, your biographical and professsional history.

FAQ and HELP

We appreciate ideas to make brainmapping.org more useful. Submit your suggestions for changes and links here: [ suggestions ]