Pattern Classification of Spontaneous Magnetoencephalographic Activity between Unipolar and Bipolar Disorders
碩士 === 國立陽明大學 === 生物醫學資訊研究所 === 99 === Background: Currently, diagnosis of bipolar disorder relies on the presence of manic or hypomanic episode of individual’s current status or during past history, according to the guidebook of Diagnostic and Statistical Manual of Mental Disorders, fourth edition...
Main Authors: | , |
---|---|
Other Authors: | |
Format: | Others |
Language: | zh-TW |
Published: |
2011
|
Online Access: | http://ndltd.ncl.edu.tw/handle/19488805664595927551 |
id |
ndltd-TW-099YM005114007 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-TW-099YM0051140072015-10-13T20:37:07Z http://ndltd.ncl.edu.tw/handle/19488805664595927551 Pattern Classification of Spontaneous Magnetoencephalographic Activity between Unipolar and Bipolar Disorders 利用靜息態腦磁波區分憂鬱症與躁鬱症之研究 Tzu-Han Hu 胡咨含 碩士 國立陽明大學 生物醫學資訊研究所 99 Background: Currently, diagnosis of bipolar disorder relies on the presence of manic or hypomanic episode of individual’s current status or during past history, according to the guidebook of Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV). However, underreporting elevated mood in the past history or being depressive episode without manic/hypomanic episodes previously challenges the reliable recognition of bipolar patients. Further clinical assessment instruments for accurate diagnosis of bipolar disorder would be needed. Methods: In this study, we recruited 72 normal controls (NC), 42 Major depressive disorder (MDD) patients, and 54 Bipolar disorder (BD) patients from outpatients. The MDD patients were divided into two groups: one was normal-mild MDD (MDD_nm), and the other was moderate-severe MDD (MDD_ms). The BD patients were distinguished into bipolar I disorder (BD I) and bipolar II disorder (BD II). All participants recorded 3 minutes magnetoencephalography in the rest (MEG) with eye open. The artifact-free MEG signals were processed by amplitude spectral density with five frequency bands including delta (2-4 Hz), theta (4-8 Hz), alpha (8-13 Hz), beta (13-30 Hz), and gamma (30-50 Hz) and divided into 10 brain regions. The features for pattern classification consisted of band power, hemispheric asymmetry, spectral ratio and coherence, the totally feature numbers were 1285. The linear discriminant analysis (LDA) was applied to select discriminant features which were used in Bayesian classifier. This study included two systems. In the first system, this study used a hierarchical approach to classifier. The first large-scale divided into NC, MDD and BD and obtained the important features which were utilized to subdivided into MDD_nm and MDD_ms and BD subdivided into BD I and BD II. In the second system, this study used totally features to classifier two-group involved (1) MDD and BD; (2) MDD_nm and MDD_ms; (3) BD I and BD II. Results: The results showed the accuracy was high by used multivariate of linear discriminant analysis and Bayesian classification. In the 3-group classification of the hierarchical system, we used linear discriminant analysis selecting a minimal number of 145 features that have high discrimination abilities and 100% accuracy. The central-parietal region and temporal-parietal region were the main feature areas in the classification analysis. The hierarchical classification, the accuracy was 100% for the MDD_nm and MDD_ms, BD I and BD II classification of the analysis with feature numbers of 37 and 47 respectively. The second system, the accuracy also was 100% for the MDD and BD, MDD_nm and MDD_ms, BD I and BD II classification with feature numbers of 80, 30 and 45, respectively. The discriminant region between MDD and BD was the anterior temporal region, between MDD_nm and MDD_ms was the frontal region and between BD I and BD II was the central-parietal region. We also found that band powers and spectral ratios were important features for differentiating the mood disorder. Conclusions: Significant classification of patients in mood disorder was achieved with Bayesian analysis of MEG resting oscillations. The obtained differentiating regions and brain oscillations were consistent with previous studies. This study showed that small numbers of sensor can be reached a good classification results. The pattern classification may offer valuable biomarkers in the mood disorder for aided clinicians’ diagnosis. Li-Fen Chen 陳麗芬 2011 學位論文 ; thesis 93 zh-TW |
collection |
NDLTD |
language |
zh-TW |
format |
Others
|
sources |
NDLTD |
description |
碩士 === 國立陽明大學 === 生物醫學資訊研究所 === 99 === Background: Currently, diagnosis of bipolar disorder relies on the presence of manic or hypomanic episode of individual’s current status or during past history, according to the guidebook of Diagnostic and Statistical Manual of Mental Disorders, fourth edition (DSM-IV). However, underreporting elevated mood in the past history or being depressive episode without manic/hypomanic episodes previously challenges the reliable recognition of bipolar patients. Further clinical assessment instruments for accurate diagnosis of bipolar disorder would be needed. Methods: In this study, we recruited 72 normal controls (NC), 42 Major depressive disorder (MDD) patients, and 54 Bipolar disorder (BD) patients from outpatients. The MDD patients were divided into two groups: one was normal-mild MDD (MDD_nm), and the other was moderate-severe MDD (MDD_ms). The BD patients were distinguished into bipolar I disorder (BD I) and bipolar II disorder (BD II). All participants recorded 3 minutes magnetoencephalography in the rest (MEG) with eye open. The artifact-free MEG signals were processed by amplitude spectral density with five frequency bands including delta (2-4 Hz), theta (4-8 Hz), alpha (8-13 Hz), beta (13-30 Hz), and gamma (30-50 Hz) and divided into 10 brain regions. The features for pattern classification consisted of band power, hemispheric asymmetry, spectral ratio and coherence, the totally feature numbers were 1285. The linear discriminant analysis (LDA) was applied to select discriminant features which were used in Bayesian classifier. This study included two systems. In the first system, this study used a hierarchical approach to classifier. The first large-scale divided into NC, MDD and BD and obtained the important features which were utilized to subdivided into MDD_nm and MDD_ms and BD subdivided into BD I and BD II. In the second system, this study used totally features to classifier two-group involved (1) MDD and BD; (2) MDD_nm and MDD_ms; (3) BD I and BD II.
Results: The results showed the accuracy was high by used multivariate of linear discriminant analysis and Bayesian classification. In the 3-group classification of the hierarchical system, we used linear discriminant analysis selecting a minimal number of 145 features that have high discrimination abilities and 100% accuracy. The central-parietal region and temporal-parietal region were the main feature areas in the classification analysis. The hierarchical classification, the accuracy was 100% for the MDD_nm and MDD_ms, BD I and BD II classification of the analysis with feature numbers of 37 and 47 respectively. The second system, the accuracy also was 100% for the MDD and BD, MDD_nm and MDD_ms, BD I and BD II classification with feature numbers of 80, 30 and 45, respectively. The discriminant region between MDD and BD was the anterior temporal region, between MDD_nm and MDD_ms was the frontal region and between BD I and BD II was the central-parietal region. We also found that band powers and spectral ratios were important features for differentiating the mood disorder.
Conclusions: Significant classification of patients in mood disorder was achieved with Bayesian analysis of MEG resting oscillations. The obtained differentiating regions and brain oscillations were consistent with previous studies. This study showed that small numbers of sensor can be reached a good classification results. The pattern classification may offer valuable biomarkers in the mood disorder for aided clinicians’ diagnosis.
|
author2 |
Li-Fen Chen |
author_facet |
Li-Fen Chen Tzu-Han Hu 胡咨含 |
author |
Tzu-Han Hu 胡咨含 |
spellingShingle |
Tzu-Han Hu 胡咨含 Pattern Classification of Spontaneous Magnetoencephalographic Activity between Unipolar and Bipolar Disorders |
author_sort |
Tzu-Han Hu |
title |
Pattern Classification of Spontaneous Magnetoencephalographic Activity between Unipolar and Bipolar Disorders |
title_short |
Pattern Classification of Spontaneous Magnetoencephalographic Activity between Unipolar and Bipolar Disorders |
title_full |
Pattern Classification of Spontaneous Magnetoencephalographic Activity between Unipolar and Bipolar Disorders |
title_fullStr |
Pattern Classification of Spontaneous Magnetoencephalographic Activity between Unipolar and Bipolar Disorders |
title_full_unstemmed |
Pattern Classification of Spontaneous Magnetoencephalographic Activity between Unipolar and Bipolar Disorders |
title_sort |
pattern classification of spontaneous magnetoencephalographic activity between unipolar and bipolar disorders |
publishDate |
2011 |
url |
http://ndltd.ncl.edu.tw/handle/19488805664595927551 |
work_keys_str_mv |
AT tzuhanhu patternclassificationofspontaneousmagnetoencephalographicactivitybetweenunipolarandbipolardisorders AT húzīhán patternclassificationofspontaneousmagnetoencephalographicactivitybetweenunipolarandbipolardisorders AT tzuhanhu lìyòngjìngxītàinǎocíbōqūfēnyōuyùzhèngyǔzàoyùzhèngzhīyánjiū AT húzīhán lìyòngjìngxītàinǎocíbōqūfēnyōuyùzhèngyǔzàoyùzhèngzhīyánjiū |
_version_ |
1718049089705213952 |