Guang-Wen Lu | PubMed Health

PubMed2005

[Nonlinear analysis of epileptic multichannel EEG].

OBJECTIVE: To study the difference between the nonlinear eigenvalue of epileptic patients and normal subjects for accurate identification of epileptic EEG signal. METHODS: Multivariable phase space reconstruction technique was employed to analyze the EEG of epileptic patients for estimating the correlation dimension(D(2)), which was one of the most often used parameters in quantitative discretion of complicated nonlinear dynamics system. RESULT AND CONCLUSION: The correlation dimension is an important eigenvalue for epileptic EEG identification.

Electroencephalography↗

PubMed2002

[Evaluation method of CT number linearity].

OBJECTIVE: To explore the method for computed tomography (CT) number linearity evaluation. METHODS: The model CTP401 of Catphan500 phantom was adopted to test 43 sets of CT scanners under the same scanning conditions. 1- and 2-order curve fitting of the testing results of each scanner were performed, with the bending of the 2-order fitted curve defined as the degree of CT number non-linearity and the slope of the linear fitting as contrast scale. The 2-norms of the fitting coefficients were calculated for the purpose of fitting quality evaluation. MATLAB 6.1 was employed for programming. RESULTS: The 2-norm of the fitting coefficients of the 2-order fitting curve was approximately 1/10 of that of corresponding linear fitting, indicating that 2-order fitting accorded better to the parent population distribution pattern. The degree of the non-linearity was manifested by the bending of the 2-order fitting curve that increased monotonously with the 2-norm of the corresponding linear fitting. CONCLUSION: It is statistically valid to define CT number non-linearity and contrast scale on the basis of 1 and 2-order fittings of the testing results, which well serves the purpose of quantitative analysis.

Algorithms↗