Page 61 - MEGIN Book Of Abstracts - 2023

be quite distinct. We aimed to find out the difference in ment and Learning Science, Key Laboratory of Ministry
the neural mechanism of impaired working memory in of Education, Nanjing, China; Medical College of Nanjing
patients with bipolar and unipolar disorder. University, Nanjing, China

METHOD According to diagnostic criteria of bipolar OBJECTIVES In clinical practice, bipolar depres-
II disorder of the Diagnostic and Statistical Manual sion (BD) and unipolar depression (UD) appear to
of Mental Disorders, Fifth Edition (DSM-5) and as- have similar symptoms, causing BD being frequently
sessments, 13 bipolar II depression (BP II), 8 unipolar misdiagnosed as UD, leading to improper treatment
depression (UD) patients and 15 healthy controls (HC) decision and outcome. Therefore, it is in urgent need
were recruited in the study. We used 2-back tasks of distinguishing BD from UD based on clinical objec-
and magnetic source imaging (MSI) to test working tive biomarkers as early as possible. Here, we aimed to
memory functions and get the brain reactions of the integrate brain neuroimaging data and an advanced
participants. machine learning technique to predict different types
of mood disorder patients at the individual level.
RESULTS Compared with HC, only spatial working
memory tasks accuracy was significantly worse in both METHODS Eyes closed resting-state magnetoencepha-
UD and BP II (p = 0.001). Pearson correlation showed lography (MEG) data were collected from 23 BD, 30 UD,
that the stronger the FCs' strength of MFG-IPL and IPL- and 31 healthy controls (HC). Individual power spectra
preSMA, the higher accuracy of SWM task within left were estimated by Fourier transform, and statistic spec-
FPN in patients with UD (r = 0.860, p = 0.006; r = 0.752, tral differences were assessed via a cluster permutation
p = 0.031). However, the FC strength of IFG-IPL was neg- test. A support vector machine classifier was further
atively correlated with the accuracy of SWM task within applied to predict different mood disorder types based
left FPN in patients with BP II (r = - 0.591, p = 0.033). on discriminative oscillatory power.

CONCLUSIONS Our study showed that the spatial RESULTS Both BD and UD showed decreased frontal-
working memory of patients with whether UD or BP central gamma/beta ratios comparing to HC, in which
II was impaired. The patterns of FCs within these two gamma power (30-75 Hz) was decreased in BD while
groups of patients were different when performing beta power (14-30 Hz) was increased in UD vs HC. The
working memory tasks. support vector machine model obtained significant
high classification accuracies distinguishing three
Keywords: Depression, Frontoparietal network, Magneto- groups based on mean gamma and beta power (BD:
encephalography, Working memory 79.9%, UD: 81.1%, HC: 76.3%, P < .01).

BMC psychiatry (2021), Vol. 21, No. 1 (34781922) (2 CONCLUSIONS In combination with resting-state MEG
citations) data and machine learning technique, it is possible
to make an individual and objective prediction for
mode disorder types, which in turn has implications for
Magnetoencephalography resting-state spectral diagnosis precision and treatment decision of mood
fingerprints distinguish bipolar depression and disorder patients.
unipolar depression (2020)
Keywords: MEG, bipolar depression, resting state, support
Jiang, Haiteng; Dai, Zhongpeng; Lu, Qing; Yao, Zhijian vector machine, unipolar depression

Department of Psychiatry, the Affiliated Brain Hospital of Bipolar disorders (2020), Vol. 22, No. 6 (31729112) (9
Nanjing Medical University, Nanjing, China; Child Develop- citations)

 ontents  Index 40
C