Search published articles
Showing 1 results for Tabanfar
Z. Tabanfar , S.m. Firoozabadi , Z. Khodakarami, Z. Shankayi ,
Volume 9, Issue 4 (Fall 2018)
Abstract
Aims: Electroencephalogram (EEG) is an important clinical test for the diagnosis of many brain diseases. The aim of this study was the analysis of electroencephalogram data during rest in patients with brain tumor.
Materials and Methods: In the present analytic observational study, EEG data of 44 patients with brain tumor (tumoral group) and 31 healthy subjects (healthy group) during rest were used. After preprocessing, the linear temporal features, linear spectral features of different frequency bands, and non-linear features of fractal dimension and entropy were extracted. Then, the distinction between healthy and tumoral groups based on extracted features was investigated, using the Davis-Bouldin statistic method, linear discriminant analysis (LDA) and nonlinear K-Nearest Neighbor (KNN) classification.
Findings: There was no significant difference between the the fractal kutz dimension and the waveform length of the two healthy and tumoral groups. Among other features, the sample entropy with a significant reduction in the tumoral group made the most distinction between the two groups (0.69 for the healthy group and 0.53 for the tumoral group). The highest classification accuracy of the two groups was 84%, using the sample entropy and KNN classification.
Conclusion: EEG signals have the potential to distinct the patients with brain tumor and healthy subjects. Nonlinear entropy features with more adaptation to the nonlinear nature of the brain shows a higher accuracy in the representation of the tumoral group. The less entropy of the tumoral group indicates less complexity in the brain processing of this group than the healthy group.