Page 294 - MEGIN Book Of Abstracts - 2023

ABSTRACT Frontotemporal dysconnectivity is a key Feature optimization method for machine
pathology in schizophrenia. The specific nature of this learning-based diagnosis of schizophrenia using
dysconnectivity is unknown, but animal models imply magnetoencephalography (2020)
dysfunctional theta phase coupling between hip-
pocampus and medial prefrontal cortex (mPFC). We Kim, Jieun; Kim, Min-Young; Kwon, Hyukchan; Kim, Ji-
tested this hypothesis by examining neural dynamics Woong; Im, Woo-Young; Lee, Sang Min; Kim, Kiwoong;
in 18 participants with a schizophrenia diagnosis, both Kim, Seung Jun
medicated and unmedicated; and 26 age, sex and IQ
matched control subjects. All participants completed Advanced Instrumentation Institute, Korea Research Institute
two tasks known to elicit hippocampal-prefrontal of Standards and Science, Daejeon, Republic of Korea;
theta coupling: a spatial memory task (during magne- Department of Medical Physics, University of Science and
toencephalography) and a memory integration task. Technology, Daejeon, Republic of Korea; Department of
In addition, an overlapping group of 33 schizophrenia Psychiatry, Konyang University College of Medicine, Konyang
and 29 control subjects underwent PET to measure the University Hospital, Daejeon, Republic of Korea; Department
availability of GABAARs expressing the α5 subunit (con- of Psychiatry, Kyung Hee University School of Medicine, Kyung
centrated on hippocampal somatostatin interneurons). Hee University Hospital, Seoul, Republic of Korea; Advanced
We demonstrate-in the spatial memory task, during Instrumentation Institute, Korea Research Institute of Stan-
memory recall-that theta power increases in left medial dards and Science, Daejeon, Republic of Korea; Department
temporal lobe (mTL) are impaired in schizophrenia, as is of Medical Physics, University of Science and Technology,
theta phase coupling between mPFC and mTL. Impor- Daejeon, Republic of Korea. Electronic address: kwkim@kriss.
tantly, the latter cannot be explained by theta power re.kr; Department of Psychiatry, Konyang University College
changes, head movement, antipsychotics, cannabis of Medicine, Konyang University Hospital, Daejeon, Republic
use, or IQ, and is not found in other frequency bands. of Korea. Electronic address: cortex@konyang.ac.kr
Moreover, mPFC-mTL theta coupling correlated strong-
ly with performance in controls, but not in subjects BACKGROUND When many features and a small num-
with schizophrenia, who were mildly impaired at the ber of clinical data exist, previous studies have used
spatial memory task and no better than chance on the a few top-ranked features from the Fisher's discrimi-
memory integration task. Finally, mTL regions showing nant ratio (FDR) for feature selection. However, there
reduced phase coupling in schizophrenia magnetoen- are many similarities between selected features. New
cephalography participants overlapped substantially method: To reduce the redundant features, we ap-
with areas of diminished α5-GABAAR availability in the plied a technique employing FDR in conjunction with
wider schizophrenia PET sample. These results indicate feature correlation. We performed an attention network
that mPFC-mTL dysconnectivity in schizophrenia is test on schizophrenic patients and normal subjects
due to a loss of theta phase coupling, and imply α5- with a 152-channel magnetoencephalograph. P300m
GABAARs (and the cells that express them) have a role amplitudes of event-related fields (ERFs) were used as
in this process. features at the sensor level and P300m amplitudes of
ERFs for 500 nodes on the cortex surface were used as
Keywords: hippocampus, prefrontal cortex, schizophrenia, features at the source level. Features were ranked using
spatial memory, theta FDR criterion and cross-correlation measure, and then
the highest ranked 10 features were selected and an
Brain: a journal of neurology (2020), Vol. 143, No. 4 exhaustive search was used to find combination having
(32236540) (27 citations) the maximum accuracy.

RESULTS At the sensor level, we found a single chan-
nel of the occipital region that distinguished the two
groups with an accuracy of 89.7 %. At source level, we

 ontents  Index 273
C