| 摘要 |
Disclosed is a system and method to derive a diagnostic index indicative of a selected cerebral phenomena which obtains 19 unipolar EEG signals from regions of interest on both the left and right hemispheres of a subject's brain. The system uses high-gain, low-noise amplifiers to maximize the dynamic range for low energy wave components of the signals. Band-pass filtering is used to reduce noise and to avoid aliasing. The system applies commonly used digital signal processing (DSP) techniques to digitize, to low-pass filter (100 Hz), and to decimate the signals. Power spectral, bispectral, and higher-order spectral processing is then performed. In a preferred embodiment, the system divides the most recent 63 seconds of digitized EEG data from each lead into 60 4-second intervals, each with 3 seconds of overlap with the previous interval. For a selected set of derived leads, the system produces auto power spectrum, autobispectrum, and auto higher-order spectrum variables, by using either a Fast Fourier Transform (FFT) based approach or a parametric approach. Any pair of leads can be combined to compute cross power spectrum, cross bispectrum, and cross higher-order spectrum variables. The spectral values are separated into bins and a value representative of the bin is selected or computed and then each value is multiplied by a predetermined coefficient. The resulting products are summed to arrive at the diagnostic index.
|
