精神
疾患 ，如重度 憂 鬱症（ Major Depressive Disorder, MDD）或雙相情感障礙
Bipolar Disorder, BD），影響了全世界大量的人，造成了嚴重的經濟和社會問題。
基於
EEG信號的機器學習方法在診斷各種精神疾病方面顯示出了顯著的效果。最
近的 MDD或 BD研究大多集中在 相干性 Coherence, COH）的分析上，這是區域間功
能連接的量化。然而，相干性受體積傳導效應的影響很大，它可能會產生虛假的連接。
最近，一些研究表明，基於相位滯後指數（ Phase Lag Index, PLI）的功能連接報告了皮
質區域之間更可靠的連接。本研究旨在研究 MDD/BD/HC的 PLI，並採用 PLI特徵來區
分不同的組別。在這項研究中， 腦電圖 是由 28名 BD2、 30名 MDD患者和 27名 HC受
試者在 張 眼休息（ Eye Opened Resting, EO）和閉眼休息 Eye Closed Resting, EC）狀態
下記錄的。注意，這裡的 BD2指的 是 BDII，即 BD的一個常見亞型。在 前處 理步驟之
ii
後，提取了各組在每個
後，提取了各組在每個休息休息狀態下的狀態下的PLI和其他常用的特徵，如和其他常用的特徵，如COH和頻和頻譜譜功率功率（（Spectral Power, SP）。）。然後然後使使用用費雪費雪準則準則(Fisher’s Criterion)對組間最具鑑別力的特徵進對組間最具鑑別力的特徵進行排序，並採用行排序，並採用序列序列前向選擇（前向選擇（Sequential Forward Selection, SFS））結合線性判別分析結合線性判別分析（（Linear Discriminant Analysis , LDA）分類器來確定最佳特徵。）分類器來確定最佳特徵。
本研究的主要結果可歸納為以下幾點
本研究的主要結果可歸納為以下幾點，，首先，首先，本研究的主要結果可歸納為以下幾點本研究的主要結果可歸納為以下幾點:首先首先，整體來說，在所有分類結果，使用，整體來說，在所有分類結果，使用 PLI 的分類率高於的分類率高於 COH 和和 SP。。BD2 和和 MDD 之間的分類可以提供之間的分類可以提供 94.55% (EO)和和 94.55% (EC)的準確性。三的準確性。三類別分類類別分類 (BD2 vs MDD vs HC)使用使用 PLI 可以提供可以提供 90.74% (EO)和和 100% (EC)的準確性。第二，將最佳的準確性。第二，將最佳 PLI 特特徵與決策值徵與決策值(DLDA)結合起來，組合成結合起來，組合成 BD2 - MDD 分類分類(BMC)指數和指數和 BD2 - HC 分類分類(BHC) 指數，供今後診斷應用。第三指數，供今後診斷應用。第三，，將訓練好之將訓練好之二別最佳模型與三類別最二別最佳模型與三類別最佳佳模型進行模型進行測試，驗證模型之泛用性測試，驗證模型之泛用性。。

Mental disorders, such as Major Depressive Disorder (MDD) or Bipolar Disorder (BD), affect a large number of people around the world, causing critical economic and social problems.
Machine learning methods based on EEG signals have shown remarkable results for the diagnosis of various mental disorders. Most of the recent MDD/BD research focuses on the analysis of coherence (COH), a quantification of between regions’ functional connectivity. However, coherence is highly affected by the volume conduction effect, which may create spurious connections. Recently, some studies have shown that phase lag index (PLI)-based functional connectivity reports a more reliable connection between cortical regions The present
study aims to investigate the PLI in MDD/BD/HC and employ PLI features to discriminate
between groups.
iv
In this study, the EEG data were recorded from 28 BD2, 30 MDD patients, and 27 HC subjects during Eye-opened resting (EO) and eyes-closed resting (EC) states. Noted that the term BD2 here refers to BDII, a common subtype of BD. After the pre-processing steps, the PLI and other commonly-used features, such as COH and spectral power (SP), were extracted for each group in each resting state. The Fisher criterion was then used to rank the most between-group discriminative features, and the sequential forward selection (SFS) strategy combined with the Linear Discriminant Analysis (LDA) classifier was employed to determine the best feature subsets. The main results of the present study can be summarized as follow: Firstly, the classification rate using PLI was 94.55% and 94.55% when binary classification was performed in EO and EC states (BD2 vs MDD, BD2 vs HC). Secondly, for ternary classification in EO and EC states (BD2 vs MDD vs HC), the classification rate using PLI with One against One (OAO) LDA classifier is as high as 90.74% and 100%.Thirdly, for the binary classification of resting states (BD2 vs MDD, BD2 vs HC), the decision value (DLDA) and the best characteristics of PLI output from the LDA classifier can be linearly combined into the BD2 - MDD Classification (BMC) index and the BD2 - HC Classification (BHC) index, which is an objective method to help physicians diagnose a patient's condition. Thirdly, test the trained two-category optimal model and the three-category optimal model to verify the versatility of the model.

[1] E. Vieta, M. Berk, T. G. Schulze, A. F. Carvalho, T. Suppes, J. R. Calabrese, K. Gao, K. W. Miskowiak, and I. Grande, “Bipolar disorders,” Nature Reviews Disease Primers, vol. 4, no. 1, 2018.
[2] 基層醫學, “雙極性疾患,” 台灣家庭醫學醫學會, August,2013.
[3] S. Y. M. Tsai, C. C. Chen, C. J. Kuo, J. C. Lee, H. C. Lee, and S. M. Strakowski, “15-year outcome of treated bipolar disorder,” Journal of Affective Disorders, vol. 63, no. 1-3, pp. 215–220, 2001.
[4] P. B. Mitchell and G. S. Malhi, “Bipolar depression: Phenomenological overview and clinical characteristics,” Bipolar Disorders, vol. 6, no. 6, pp. 530–539, 2004.
[5] J. Angst, A. Gamma, F. Benazzi, V. Ajdacic, D. Eich, and W. Rössler, “Toward a re-definition of subthreshold bipolarity: Epidemiology and proposed criteria for Bipolar-II, minor bipolar disorders and hypomania,” Journal of Affective Disorders, vol. 73, no. 1-2, pp. 133–146, 2003.
[6] S. N. Ghaemi, E. Boiman, and F. K. Goodwin, “Insight and outcome in bipolar, unipolar, and anxiety disorders,” Comprehensive Psychiatry, vol. 41, no. 3, pp. 167–171, 2000.
[7] G. M. MacQueen and L. T. Young, “Bipolar II disorder: Symptoms, course, and response to treatment,” Psychiatric Services, vol. 52, no. 3, pp. 358–361, 2001.
[8] M. D. Stensland, J. F. Schultz, and J. R. Frytak, “Diagnosis of unipolar depression following initial identification of bipolar disorder,” The Journal of Clinical Psychiatry, vol. 69, no. 5, pp. 749–758, 2008.
[9] “Diagnosis of bipolar disorder in primary care,” Primary Health Care, vol. 21, no. 2, pp. 13–13, 2011.
[10] S. Pallaskorpi, K. Suominen, M. Ketokivi, O. Mantere, P. Arvilommi, H. Valtonen, S. Leppämäki, and E. Isometsä, “Five-year outcome of bipolar I and II disorders: Findings of the JORVI bipolar study,” Bipolar Disorders, vol. 17, no. 4, pp. 363–374, 2015.
[11] J. Angst, “The bipolar spectrum,” British Journal of Psychiatry, vol. 190, no. 3, pp. 189–191, 2007.
[12] M. J. Taylor and J. R. Geddes, “Evidence-based pharmacotherapy of bipolar disorder,” Essential Evidence-Based Psychopharmacology, pp. 39–52, 2012.
[13] L. Tondo, C. Visioli, A. Preti, and R. J. Baldessarini, “Bipolar disorders following initial depression: Modeling predictive clinical factors,” Journal of Affective Disorders, vol. 167, pp. 44–49, 2014.ognitive impairment,” The lancet, 2006.
[14] American Psychiatric Association. DSM-5 Task Force., “Diagnostic and statistical manual of mental disorders: DSM-5. 5th ed,” American Psychiatric Association: Washington, D.C., 2013.
[15] B. C. Dickerson, and R. A., Sperling, “Functional abnormalities of the medial temporal lobe memory system in mild cognitive impairment and Alzheimer's disease: insights from functional MRI studies,” Neuropsychological, 2008.
[16] H. He, C. Hu, Z. Ren, L. Bai, F. Gao, and J. Lyu, “Trends in the incidence and dalys of bipolar disorder at global, regional, and national levels: Results from the global burden of disease study 2017,” Journal of Psychiatric Research, vol. 125, pp. 96–105, 2020.
[17] K. R. Merikangas, R. Jin, J.-P. He, R. C. Kessler, S. Lee, N. A. Sampson, M. C. Viana, L. H. Andrade, C. Hu, E. G. Karam, M. Ladea, M. E. Medina-Mora, Y. Ono, J. Posada-Villa, R. Sagar, J. E. Wells, and Z. Zarkov, “Prevalence and correlates of bipolar spectrum disorder in the World Mental Health Survey Initiative,” Archives of General Psychiatry, vol. 68, no. 3, p. 241, 2011.
[18] H. U. Wittchen, “The burden of mood disorders,” Science, vol. 338, no. 6103, pp. 15–15, 2012.
[19] L. E. O. Sher, “Manic-depressive illness: Bipolar disorders and recurrent depression,” American Journal of Psychiatry, vol. 165, no. 4, pp. 541–542, 2008.
[20] R. M. Hirschfeld, “Differential diagnosis of bipolar disorder and major depressive disorder,” Journal of Affective Disorders, vol. 169, 2014.
[21] G. M. Goodwin, “Bipolar depression and treatment with antidepressants,” British Journal of Psychiatry, vol. 200, no. 1, pp. 5–6, 2012.
[22] G. McLoughlin, S. Makeig, and M. T. Tsuang, “In search of biomarkers in psychiatry: EEG-based measures of brain function,” American Journal of Medical Genetics Part B: Neuropsychiatric Genetics, vol. 165, no. 2, pp. 111–121, 2013.
[23] S. S. Jeste, J. Frohlich, and S. K. Loo, “Electrophysiological biomarkers of diagnosis and outcome in neurodevelopmental disorders,” Current Opinion in Neurology, vol. 28, no. 2, pp. 110–116, 2015.
[24] S. Olbrich, R. van Dinteren, and M. Arns, “Personalized medicine: Review and perspectives of promising baseline EEG biomarkers in major depressive disorder and attention deficit hyperactivity disorder,” Neuropsychobiology, vol. 72, no. 3-4, pp. 229–240, 2015.
[25] D.-J. Kim, A. R. Bolbecker, J. Howell, O. Rass, O. Sporns, W. P. Hetrick, A. Breier, and B. F. O'Donnell, “Disturbed resting state EEG synchronization in bipolar disorder: A graph-theoretic analysis,” NeuroImage: Clinical, vol. 2, pp. 414–423, 2013.
[26] J. W. Y. Kam, A. R. Bolbecker, B. F. O'Donnell, W. P. Hetrick, and C. A. Brenner, “Resting state EEG power and coherence abnormalities in bipolar disorder and schizophrenia,” Journal of Psychiatric Research, vol. 47, no. 12, pp. 1893–1901, 2013.
[27] M. M. Khaleghi,M.R. Mohammadi, “Characteristics of alpha band frequency in adolescents with bipolar II disorder: A resting-state QEEG study,” Iranian Journal of Psychiatry, pp. 8-12, 2015
[28] N. Handayani, S. N. Khotimah, F. Haryanto, I. Arif, and W. P. Taruno, “Resting state EEG power, intra-hemisphere and inter-hemisphere coherence in bipolar disorder,” AIP Conference Proceedings, 2017.
[29] F. M. Howells, H. S. Temmingh, J. H. Hsieh, A. V. van Dijen, D. S. Baldwin, and D. J. Stein, “Electroencephalographic delta/alpha frequency activity differentiates psychotic disorders: A study of schizophrenia, bipolar disorder and methamphetamine-induced psychotic disorder,” Translational Psychiatry, vol. 8, no. 1, 2018.
[30] A. Khaleghi, M. R. Mohammadi, M. Moeini, H. Zarafshan, and M. Fadaei Fooladi, “Abnormalities of alpha activity in Frontocentral region of the brain as a biomarker to diagnose adolescents with bipolar disorder,” Clinical EEG and Neuroscience, vol. 50, no. 5, pp. 311–318, 2019.
[31] A. Damborská, C. Piguet, J.-M. Aubry, A. G. Dayer, C. M. Michel, and C. Berchio, “Altered electroencephalographic resting-state large-scale brain network dynamics in Euthymic Bipolar Disorder patients,” Frontiers in Psychiatry, vol. 10, 2019.
[32] M. Sunaga, Y. Takei, Y. Kato, M. Tagawa, T. Suto, N. Hironaga, N. Sakurai, and M. Fukuda, “The characteristics of power spectral density in bipolar disorder at the Resting State,” Clinical EEG and Neuroscience, 2021.
[33] J. Zhang, T. Liu, Z. Shi, S. Tan, D. Suo, C. Dai, L. Wang, J. Wu, S. Funahashi, and M. Liu, “Impaired self-referential cognitive processing in bipolar disorder: A functional connectivity analysis,” Frontiers in Aging Neuroscience, vol. 14, 2022.
[34] C. Tas, M. Cebi, O. Tan, G. Hızlı-Sayar, N. Tarhan, and E. C. Brown, “EEG power, Cordance and coherence differences between unipolar and bipolar depression,” Journal of Affective Disorders, vol. 172, pp. 184–190, 2015.
[35] T. T. Erguzel, C. Tas, and M. Cebi, “A wrapper-based approach for feature selection and classification of major depressive disorder–bipolar disorders,” Computers in Biology and Medicine, vol. 64, pp. 127–137, 2015.
[36] T. T. Erguzel, G. H. Sayar, and N. Tarhan, “Artificial Intelligence Approach to classify unipolar and bipolar depressive disorders,” Neural Computing and Applications, vol. 27, no. 6, pp. 1607–1616, 2015.
[37] C. J. Stam, G. Nolte, and A. Daffertshofer, “Phase lag index: Assessment of functional connectivity from multi-channel EEG and Meg with diminished bias from common sources,” Human Brain Mapping, vol. 28, no. 11, pp. 1178–1193, 2007.
[38] R. Sharp, “The hamilton rating scale for depression,” Occupational Medicine, vol. 65, no. 4, pp. 340–340, 2015.
[39] M. Lukasiewicz, S. Gerard, A. Besnard, B. Falissard, E. Perrin, H. Sapin, M. Tohen, C. Reed, and J.-M. Azorin, “Young mania rating scale: How to interpret the numbers? determination of a severity threshold and of the minimal clinically significant difference in the emblem cohort,” International Journal of Methods in Psychiatric Research, vol. 22, no. 1, pp. 46–58, 2013.
[40] G. Jackson-Koku, “Beck Depression Inventory,” Occupational Medicine, vol. 66, no. 2, pp. 174–175, 2016.
[41] L. C., Gaetz, W., and H, Zhu, “Estimation of time-varying coherence and its application in understanding brain functional connectivity,” EURASIP Journal on Advances in Signal Processing, 2010.
[42] M. S., Ratsch, G. ,J. Weston, B. Scholkopf, and K. R. Mullers, “Fisher discriminant analysis with kernels,” In Neural networks for signal processing IX: Proceedings of the 1999 IEEE signal processing society workshop, 1999.
[43] A. Justel, D. Peña, and R. Zamar, “A multivariate Kolmogorov-Smirnov test of goodness of fit,” Statistics & Probability Letters, vol. 35, no. 3, pp. 251–259, 1997.