Center for Machine Learning and Intelligent Systems
About  Citation Policy  Donate a Data Set  Contact


Repository Web            Google
View ALL Data Sets

Multi-view Brain Networks Data Set
Download: Data Folder, Data Set Description

Abstract: Multi-layer brain network datasets derived from the resting-state electroencephalography (EEG) data.

Data Set Characteristics:  

Multivariate

Number of Instances:

70

Area:

Life

Attribute Characteristics:

Integer

Number of Attributes:

70

Date Donated

2020-08-06

Associated Tasks:

Classification, Clustering

Missing Values?

N/A

Number of Web Hits:

4710


Source:

Chang-Dong Wang, Sun Yat-sen University, changdongwang '@' hotmail.com


Data Set Information:

For constructing the multi-layer brain network datasets, we collect the resting-state electroencephalography (EEG) data from Department of Otolaryngology of Sun Yat-sen Memorial Hospital, Sun Yat-sen University. Three types of subjects participate the experiments, namely 51 deafness patients, 54 tinnitus patients and 42 normal controls. Since there may exist significant differences of brain network among the three types of subjects, a separate multi-layer brain network is constructed for each type of subjects. The EEG data are collected by the EEG analyzer with 128 scalp electrodes from Electrical Geodesics, Inc. The standard data acquisition and preprocessing procedure is applied, based on which 128 electrode EEG data can be obtained. On the datasets, only the 70 electrodes belonging to the ten regions of interest (ROI) are used as shown in the graphic file. Notice that the electrodes C17 and A23 in regions CF and CO are two reference electrodes that are not used in our study. These 70 electrodes are further divided into two classes corresponding to the upper and lower parts. That is, the 34 electrodes in the upper ROIs, namely CF, LAL, LAM, RAM and RAL are grouped into one class, while the 36 electrodes in the lower ROIs, namely CO, LOT, LPM, RPM and ROT are grouped into another class. Based on the class labels, the ground-truth community labels of the 70 nodes corresponding to the 70 electrodes can be obtained. After obtaining 70 electrodes, the STUDY module of EEGLAB2 is used to extract the features of the preprocessed data, i.e., extracting the power values of different frequency bands. In our study, 9 different frequency bands are used, including Delta (1-4Hz), Theta (4-8Hz), Alpha1 (8-10Hz), Alpha2 (10-12Hz), Beta1 (13-18Hz), Beta2 (18-21Hz), Beta3 (21-30Hz), Gamma1 (30.5-45Hz) and Gamma2 (55-70Hz). In each frequency band, the Pearson correlation coefficient is calculated between each pair of electrodes, based on which the interconnection between electrodes can be constructed. That is, for any two electrodes, the Pearson correlation coefficient between their EEG data is calculated for each subject. Then the average value over the 51 deafness patients (resp. 54 tinnitus patients and 42 normal controls) is calculated. If the average value is no smaller than 0.3, the two electrodes are interconnected. In this way, a 9-layer network is constructed for the deafness patients (resp. tinnitus patients and normal controls), where each layer corresponds to each frequency band and is named accordingly. For simplicity, the 9-layer network for the deafness patients (resp. tinnitus patients and normal controls) is called DBrain (resp. TBrain and NBrain).


Attribute Information:

See the relevant information.


Relevant Papers:

Ling Huang, Chang-Dong Wang and Hong-Yang Chao. HM-Modularity: A Harmonic Motif Modularity Approach for Multi-layer Network Community Detection. IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING 2019.



Citation Request:

Ling Huang, Chang-Dong Wang and Hong-Yang Chao. HM-Modularity: A Harmonic Motif Modularity Approach for Multi-layer Network Community Detection. IEEE TRANSACTIONS ON KNOWLEDGE AND DATA ENGINEERING 2019.


Supported By:

 In Collaboration With:

About  ||  Citation Policy  ||  Donation Policy  ||  Contact  ||  CML