| Abstract |
Objective: Aim of the present study was the investigation of functional connectivity
during working memory in schizophrenic subjects by means of time series analysis
(linear and non-linear) and theoretical graph measures. This was attempted through
analysis of EEG recordings during a working memory task, (the 2-back level of the
n-back task), as well as during two control tasks. The specific questions addressed
were: a) Does cortical connectivity, in schizophrenics with relatively preserved
working memory functions, express small-world properties up to the extent that
control subjects do? b) Are differences in SWN properties due to schizophrenia or to
differences in working memory performance and working memory system
efficiency?
Materials-Methods: we analyzed EEG data from a group of 20 schizophrenic
subjects with relatively preserved working memory performance, a group of 20
control subjects matched to the schizophrenics for age, sex and years of education
and a group of 20 controls with maximal performance on the working memory task
used.
The questionnaires and scales administered were the following:
Schizophrenic subjects: a) Scale for the Assessment of Negative Symptoms, SANS,
b) Scale for the Assessment of Positive Symptoms, SAPS.
Controls: a) Mini International Neuropsychiatric Interview
All: a)General and demographic information sheet, b) Edinburgh Handedness
Inventory, c) verbal scales of Weschler Adult Intelligence Scale (WAIS), d) Digit
Symbol subtest of WAIS practical scale, e) Stroop test and f) a modified version of
the n-back task. Either coherence, or synchronization likelihood was computed for
every pair formed by the 28 scalp electrodes (28x28 matrix) and a corresponding
adjacency matrix was extracted for a spectrum of pre specified threshold values.
This type of analysis allowed us to define the values of mean degree k for which the
groups differed the most. We, then, computed the characteristic graph measures
(clustering coefficient C and characteristic path length L) of the corresponding
graphs for fixed degree k, separately for each study group. The last step was to
compare the characteristic graph values to those computed for 50 random graphs
169
with the same degree distribution. We initially compared the two control groups.
Using matrices with synchronization likelihood values, we compared the group with
the weaker SWN properties to the schizophrenia group. At last we compared the
characteristic graph measures of schizophrenics and controls, for fixed degree k
values, while using matrices with coherence values.
For the statistical analysis of data we used repeated measures ANOVA (for
comparisons of the characteristic graph measures), t-tests for the demographic
measures comparisons and Mann-Whitney tests for neuropsychological measures
comparisons. P&λλτ 0.05 was considered significant.
Results: Schizophrenic patients had significantly different verbal IQ scores and Digit
symbol scores from matched controls (p=0.004 and p=0.013 respectively). Expert
controls had significantly different scores from the other two groups in every
neuropsychological test used, as expected. During the 2-back task, patients did not
differ from matched controls regarding performance and reaction times, while expert
controls differed from the other two groups regarding both parameters. Comparing
the two control groups we found that the matched control group showed stronger
SWN properties than the expert group, during the 2-back working memory task and
not during rest, regarding θ, α1, α2, β and γ1 bands. Comparisons of the
schizophrenia group and the expert group revealed SWN properties only for the
control group and during working memory in α1, α2, β and γ1 frequency bands. The
schizophrenia group showed SWN properties only in the θ band, while none of the
groups showed such properties in γ2 band. In the third paper published we
compared characteristic graph measures between the schizophrenia group and the
expert group. We found mean Clustering coefficient (mean C) significantly different
between groups in the a1 band during control condition (identification of pre
specified target) and not during working memory. Additionally, in the β and the γ1
frequency bands the control group optimized C values, from control to working
memory condition, while the patient group failed to do so. Simple main effects of
group were significant during working memory in the same frequency bands.
Regarding characteristic path length (L) differences were revealed in the a2
frequency band. Again controls optimized L values from control to working memory
condition, while the patient group failed to do so. Simple main effects of group were
170
significant only during control condition with the schizophrenia group expressing
more optimal organization (shorter path length) during the control condition, a
situation reversed during working memory.
Conclusions: The present study is the first ever published using graph theory
measures extracted from scalp EEG recordings to study the organization of cortical
networks in schizophrenia. The main conclusions are the following: a) the control
group with the significantly lower working memory performance and lower verbal IQ
scores showed more optimal organization of cortical networks (revealed by stronger
SWN properties) during a working memory task than a group of high performers,
with significantly greater verbal IQ scores. This finding supports the neural efficiency
hypothesis. b) During working memory, healthy subjects exhibit small-world
properties while such properties are not present in a schizophrenia group which
succeeded in the working memory task tested. This disruption of the optimal spatial
pattern of cortical functional connectivity in schizophrenia is not en epiphenomenon
of relative difficulties in coping with working memory procedures due to general
intelligence level.
|
Εκτίμηση των διαφορών στην εγκεφαλική λειτουργία των σχιζοφρενών σε σχέση με υγιείς μάρτυρες με χρήση συνδυασμού μεθόδων γραμμικής και μη γραμμικής ανάλυσης κατα τη διάρκεια γνωσιακής πρόκλησης
