this is for holding javascript data
Anisha Keshavan edited sectionAim_1__subsec.tex
about 8 years ago
Commit id: d83eead0280c7e84b86adcd0da599f112074a3fb
deletions | additions
diff --git a/sectionAim_1__subsec.tex b/sectionAim_1__subsec.tex
index 3978c4d..a15379d 100644
--- a/sectionAim_1__subsec.tex
+++ b/sectionAim_1__subsec.tex
...
\subsubsection*{Dynamic Functional Connectivity}
The analysis This aim will expand on the result of
Aim 1 assumes that temporal dynamics stay my preliminary data, which showed increased connectivity from the
same, meaning that precuneus to the
nodes sensorimotor strip was related to poor nine hole peg test performance. Studies of
traditionally defined networks such the precuneus and posterior cinculate cortex, referred to as the
default mode network are stationary. Recently, in \cite{Chang_2010}, researchers found that temporal dynamics are not stationary. In particular, they found posterior medial cortex, have shown that
the phase this region is anatomically and
coherence of the functional subdivided in humans and primates \cite{Margulies_2009}. The posterior cingulate
cortex (PCC, a hub of region is more connected to limbic structures, while the
default mode network) had variable connectivity with other nodes anterior, middle, and posterior sections of
not only the DMN, but other networks such as the
salience precuneus correspond to sensorimotor, cognitive/associative, and
attention networks. Given this finding, visual regions, respectively. While on average it
is possible that a disease pathology like MS may
change seem like the entire PMC structure is functionally connected to the medial prefrontal gyrus (of the default mode network), capitalizing on the
normal network dynamics functional boundaries of the PMC may result in
short periods a larger predictive power of
time, and that dynamic functional
connectivity may MRI, and could be
more sensitive to these small changes. For Alzheimers disease, researchers used indicatative of functional
dynamics to build a classifier to differentiate between patients with mild cognitive impairment or normal controls \cite{Wee_2015}. adaptation, in MS.
The analysis of Aim 1 assumes that temporal dynamics stay the same, meaning that the nodes of traditionally defined networks such as the default mode network are stationary. Recently, in \cite{Chang_2010}, researchers found that temporal dynamics are not stationary. In particular, they found that the phase and coherence of the posterior cingulate cortex (PCC, a hub of the default mode network) had variable connectivity with other nodes of not only the DMN, but other networks such as the salience and attention networks. Given this finding, it is possible that a disease pathology like MS may change the normal network dynamics in short periods of time, and that dynamic functional connectivity may be more sensitive to these small changes. For Alzheimers disease, researchers used functional dynamics to build a classifier to differentiate between patients with mild cognitive impairment or normal controls \cite{Wee_2015}.
The analysis consists of a sliding window approach outlined in \cite{Chang_2010}, where a graph is estimated from a section of the timeseries, and that section slides across the whole timeseries. In
\cite{Wee_2015}, \cite{Yang_2014}, researchers ran hierarchical clustering on the set of connectomes from the
method was modified sliding window analysis, in order to
enforce temporal smoothness (meaning that network connectivity strengths classify the connectomes into 5 states across all healthy control subjects. For each time-window, the four PMC regions could be classified into 1 of the 5 states. Yang and
topology cannot change drastically between adjacent sliding windows), which is accomplished using a fused multiple group LASSO. From colleagues found that the
graph, network measures, such as anterior, middle, posterior, and ventral (posterior cingulate) regions spent more time in the
clustering coefficient sensorimotor, associative/cognitive, visual and
small-worldness properties, are estimated\cite{Wee_2015}. The new resulting timeseries (clustering coefficient over time, or small-worldness over time) will then be used as features for a random forest classifier, limbic states, respectfully, which
will attempt aligned well with previous studies on the functional subdivisions of the PMC. However, they found that these regions did not consistently connect to
distinguish between their respective state throughout the
High and Low NHPT groups, and healthy controls. entire timeseries, but would switch states, spending a nearly even amount of time in each of the remaining 4 states. They also found this result to be scan-rescan reproducible.
I
prefer random forests to SVMs because there are less parameters hypothesize that the amount of time spent in the sensorimotor state by each of the four PMC regions will be elevated in MS patients in comparison to
go into healthy controls, and that the time spend in the sensorimotor state of the high disability group will exceed that of the
model. low disability group. A regularized least squares analysis will be run using age, gender, and lesion load as regressors, along with a 20-feature vector with the amount of time that each of the 4 PMC nodes spend in each of the 5 states (one of which is the sensorimotor state). The
goal is to show that the coefficients on the amount of time spent in the sensorimotor state is the highest, meaning that it contributes most to the prediction of patient group. This dynamic function connectivity analysis
already needs the following parameters to tune: 1) the window size, 2) the step size and 3) the regularization
parameters parameter for the
LASSO optimization. regularized least squares. Parameter tuning is done in a nested leave-one-out cross-validation, where the inner cross-validation fold is used to select the parameters and the results from outer fold are reported as the effect size.