|
|
|
|
Ana Rapoport: Effect of Computer Interface on Accuracy of
Neuroimaging Data Annotation
Abstract:
Tracing human brain
regions on structural magnetic resonance (MR) images is a key step in the
analysis of brain changes that accompany brain development, disease, and
aging. Tracing is repetitive and requires a high degree of precision in
order for the acquired information to be valid and yield significant
results. A standard mouse
cannot provide the accuracy, precision and comfort necessary to acquire
such results. Through this study we want to determine whether a Tablet PC
or a pen mouse will provide more precise traces as well as provide more
comfort for the tracer. The project entails the recruitment of twenty
subjects between the ages of 18 and 25 with an equal number of female and
male subjects. Each subject is trained to identify and trace the caudate
nucleus, a brain structure, subjects are tracing the caudate nucleus on a
set of 15 images currently. Once the task is completed, the subjects are
asked to fill out a questionnaire, which lets them numerically evaluate
their own impressions of their comfort and precision during tracing. Once
all the data is collected, statistical models will determine which device
provided us with the most consistent tracing from image to image and tracer
to tracer, as well as the most comfortable, intuitive experience as
indicated by the tracer.
|
|
|
|
Peter Harris: Structural
Effects of Hypertension on the Corpus Callosum
Abstract: Although both cardiovascular and
cognitive conditions are prevalent among elderly populations, the
relationship between blood flow dysfunction and the brain remains unclear.
The effects that acute cerebrovascular injuries, such as stroke or
white-matter hyperintensities, have on brain structure are well-documented.
Less is known about the long-term effects of hypertension, a prevalent and
chronic vascular problem, on brain structure and cognition. The goal of
this study is to document hypertension- related brain structure volume and
shape changes in an Icelandic elderly population. The subjects are very
homogenous in terms of demographic characteristics, hypertension treatment,
access to healthcare, and genetics. To elucidate volume change we will
manually trace the corpus callosum, a structure that connects the two
hemispheres and is the largest white matter tract in the brain. Shape
change will be assessed using advanced regional measures. The corpus
callosum will be examined because it is both easily recognizable and
traceable, facilitating accurate volume and shape computations. Although
previous studies have measured these types of change, there have been
numerous limitations, including: heterogeneity within the subject
population, poor regional measures, and a lack of longitudinal studies. Our
hypothesis is that hypertensive patients will have smaller callosal volumes
in general, with particular regions of the callosum being more affected
than others due to differential vascularization. Larger regional changes
should correlate with reduced blood flow or anatomical vascular limitations
in that region. Future studies may incorporate longitudinal designs or investigate
the mechanism by which hypertension causes structural change
|
|
|
|
Gautam Prasad: Automated Brain Region Delineation in Structural Magnetic
Resonance Images
Abstract:
Brain image
segmentation plays a major role in understanding brain changes associated
with aging and disease. Automatic brain region segmentation is necessary
because it provides more consistency and accuracy compared to manual
tracing of regions, which require a great deal of time and effort and are
plagued with irregularities and human error. This research focused
specifically on Magnetic Resonance (MR) images of the brain, although the
techniques can be generalized to many types of segmentation. Robust probabilistic models of
brain image appearance along lines normal to the boundary of the brain were
created to assess the goodness of fit of models of the brain boundary and how
to improve the positions of the models. The models were based on image
intensities in the brain periphery. Efficient computer programs estimate
the position of the brain boundary based on these models. The models and
fits were evaluated on brain MR images of elderly subjects and produced
promising results. Work was also done on incorporating model selection into
the program so that different regions of the brain could be fitted using a
model specific to that area. Further work will focus on creating a
probabilistic model showing the uncertainly of the delineation and to
quantify the quality of the image data.
|
|
