Wednesday December 5th, 2.30pm-5.00pm RSS SIAP/East Midlands Meeting.
University of Nottingham, Room C14 Pope Building, University Park.
Coffee/tea available 2.00-2.30pm
2.30 Inference in fMRI experiments using spectral domain methods
Jonathan Marchini
University of Oxford
3.10 Statistical Parametric Mapping
Karl Friston
Institute of Neurology, University College London
3.50-4.20 Coffee/tea
4.20 Mammographic Image Analysis
Francesco de Pasquale
Istituto per le Applicazioni del Calcolo (CNR), Rome, Italy
and
University of Plymouth, UK
--------------------------------------------------------------------------
Inference in fMRI experiments using spectral domain methods
Jonathan Marchini
University of Oxford
fMRI experiments allow neuroscientists to 'watch' the brain function of
a subject as they perform some task or receive some stimulus. These
experiments produce large spatio-temporal datasets consisting of
repeated measurements of neuronal activation on a 3D grid of locations
throughout a subjects brain. In the analysis of each individual
experiment interest lies in identifying areas that exhibit significant
activation in response to the given stimulus or task. The fine spatial
and temporal resolution of the datasets implies that significant
spatio-temporal correlations occur within the datasets. How these
correlations are modeled delineates the many proposed methods of
analysis that currently exist in the literature. Our work has focussed
on the application of spectral time series methods to the analysis of
fMRI experiments. We find that it is important to use robust methods of
estimation to avoid the deleterious effects of imaging artefacts that
can occur within these datasets.
----------------------------------------------------------------------
Statistical Parametric Mapping
Karl Friston
Institute of Neurology, University College London
Functional brain mapping studies are usually analyzed with some form of
statistical parametric mapping. Statistical parametric mapping refers
to the construction of spatially extended statistical processes to test
hypotheses about regionally specific effects.
Statistical parametric maps (SPMs) are continuous image processes with
voxel values that are, under the null hypothesis, distributed according
to a known probability density function, usually the Student's T or F
distributions. These are known colloquially as T- or F-maps. The
success of statistical parametric mapping is due largely to the
simplicity of the idea. Namely, one analyses each and every voxel
using any standard (univariate) statistical test. The resulting
statistical parameters are assembled into an image - the SPM. SPMs are
interpreted as spatially extended statistical processes by referring to
the probabilistic behavior of Gaussian fields (Adler 1981, Worsley et
al 1992, Friston et al 1994a, Worsley et al 1996). Gaussian random
fields model both the univariate probabilistic characteristics of a SPM
and any non-stationary spatial covariance structure. 'Unlikely'
excursions of the SPM are interpreted as regionally specific effects,
attributable to the sensorimotor or cognitive process that has been
manipulated experimentally.
Over the years statistical parametric mapping has come to refer to
the conjoint use of the general linear model (GLM) and Gaussian random
field (GRF) theory to analyze and make classical inferences about
spatially extended data. The GLM is used to estimate some parameters
that could explain the data in exactly the same way as in conventional
analysis of discrete data. GRF theory is used to resolve the multiple
comparison problem that ensues when making inferences over a volume of
the brain. GRF theory provides a method for correcting p values for
the search volume of a SPM and plays the same role for continuous data
(i.e. images) as the Bonferonni correction for the number of
discontinuous or discrete statistical tests.
----------------------------------------------------------------------
Mammographic Image Analysis
Francesco de Pasquale
Istituto per le Applicazioni del Calcolo (CNR), Rome, Italy
and
University of Plymouth, UK
We discuss the problem of Bayesian image reconstruction and
classification for dynamic MRI images of the breast. This imaging
technique produces a temporal sequence of images of the same breast
slice acquired after the injection of a contrast agent into the blood
stream. Radiologists wish to extract as much clinical information as
possible from the image sequence and to represent it using a few
easily interpreted images. They want to classify each pixel of the
acquired slice as belonging to one of three possible tissue types
(non-tumour, benign tumour and malign tumour). Typically these data
are affected by random degradation due to measurement noise and
deterministic degradation due to patient motion during image
acquisition. Both types of degradation have to be taken into account
to obtain accurate classification results. Both a non-parametric and
a parametric approach are developed to reconstruct the 'true' images
from the acquired sequence.
In the first case Bayesian estimation of the true image intensity for
each pixel and each time is performed without using a spatio-temporal
model for the signal, and so relevant quantities (attributes for the
further classification step) describing image intensity variation are
computed independently for each pixel. Alternatively, in the
parametric approach a specific spatio-temporal model is adopted for
the signal. The parameters of the model are estimated, and the
reconstruction of the 'true' images and the classification is
performed based on these.
Different spatio-temporal prior models are adopted for the two
approaches and all the hyper-parameters are estimated as a part of the
procedure. In addition, a fully Bayesian approach for obtaining a
classification image from the sequence is presented together with a
method for estimating the normalizing constant of the prior.
Finally, classification results based on the non-parametric approach
and a comparison with standard classification techniques will be
given.
A software package that implements these methods will be demonstrated
throughout the talk.
----------------------------------------------------------------------------
The Pope building is number 31 on the main campus map, which can be found at
http://www.nottingham.ac.uk/univ/campus-maps/up-map.html
If travelling by car there is Visitors' Pay and Display parking near the Economics and Geography Building in the centre of campus.
Please contact Ian Dryden ([log in to unmask]) if you need
any further information, or see http://www.maths.nott.ac.uk/~ild/siap.html
|