[Nipy-devel] Fwd: Statistics in Neuroscience Talk (01/15/10)

Arno Klein binarybottle@gmail....
Mon Jan 11 14:30:01 CST 2010


a talk with ramifications for pipeline environments!

cheers,
@rno


---------- Forwarded message ----------
From: <martin@stat.columbia.edu>
Date: Mon, Jan 11, 2010 at 3:20 PM
Subject: Statistics in Neuroscience Talk (01/15/10)
To:

Hi everyone,

On Friday, January 15th at 11am the Department of Statistics will be hosting
a talk by Stephen Strother from the Rotman Research Institute entitled
'Better BOLD fMRI Across the Lifespan from Children to the Elderly'.

The talk will be held in 903 School of Social Work (9th floor) and is part
of the Department's 2009-10 Special Focus Series on 'Statistical Methods in
Neuroscience'.

Best,

Martin Lindquist


Better BOLD fMRI Across the Lifespan from Children to the Elderly

Stephen Strother
Rotman Research Institute
Toronto, Canada

Researchers using BOLD fMRI typically operate with an implicit hypothesis
that: 'My BOLD fMRI results are robust to the exact techniques, algorithms,
software packages and pipeline combinations I used and are not a significant
function of these choices'. In this talk I will present evidence for the
alternative hypothesis that: 'BOLD fMRI results are a significant function
of many pipeline choices, particularly early and late across the lifespan'.

I will argue that we actually know little about what constitutes an optimal
BOLD pipeline for many cognitive and clinically relevant tasks, particularly
in children, middle-aged and older subjects who represent the age-matched
controls relevant for many clinical populations. I will present evidence for
the alternative hypothesis from three data sets: Fusiform face processing in
children and young adults (Evans et al., Neuroimage, in press);
neuropsychological testing using the Trails A/B task in young normal adults
(Churchill et al., in preparation); a cognitive multitask activation study
(Grady et al., Cerebral Cortex, in press). Results will be generated using
both univariate GLM and multivariate analysis models, together with
performance metrics for prediction and activation pattern reproducibility
(NPAIRS, Strother et al., Neuroimage, 2004), and a new measure of mutual
information extraction (Afshinpoor et al., Human Brain Mapp., in revision).

I will conclude that: 1) We do not adequately understand the impact of the
pipeline choices that we make to produce BOLD fMRI results; 2) with better
understanding of pipeline choices variability across published BOLD fMRI
studies could be significantly reduced;  3) 1 and 2 are particularly true
outside of the study of primary sensory systems in young normal controls,
because BOLD fMRI results are a strong function of pipeline choices,
particularly in the young, middle-aged and elderly.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.scipy.org/pipermail/nipy-devel/attachments/20100111/4486c068/attachment.html 


More information about the Nipy-devel mailing list