$225 DPS3.05 SIMULATION T E C H N I Q U E S MYOGRAPHY.
IN
ELECTRO-
S.D. Nandedkar (Durham, NC, USA) The motor unit action potential (MUAP) recorded in clinical electromyography (EMG) is the spatial and temporal summation of the action potentials (APs) from all muscle fibers in a motor unit (MU). An important determinant of MUAP waveform characteristics is the size of the recording electrode. We have used computer simulations to study the effects of electrode size and changes in MU anatomy on the recorded MUAP. A modified line source model of single muscle fiber action potentials was used to simulate MUAPs as recorded by single fiber (SF) EMG, concentric needle (CN) EMG and macro link electrodes. Results indicate that SFEMG recordings from a normal MU contain mainly the APs of the closest one to three nmscle fibers of the MU. The amplitude, area and duration, respectively, of the simulated CNEMG MUAPs are determined mainly by the number and size of muscle fibers within a semicircular territory of 0.5, 1.5 and 2.5 ram, around the tip of the electrode. The amplitude and area of simulated Macro EMG MUAPs increase in parallel with the number of muscle fibers in the MU. In simulations of pathologic MUs, we found three patterns of change in SFEMG and Macro EMG that reflect different types of MU reorganization. Patterns similar to these are seen in certain nerve and muscle diseases. Using simulations we have thus been able to demonstrate the effects of various anatomical characteristics of normal and pathologic MUs on the MUAPs recorded using different types of EMG electrodes.
MODELS. DPW 1.01 MULTIPLE A N D PARTIAL C O H E R E N C E S AS A T O O L IN THE MULTI-CHANNEL EEG ANALYSIS. P. Franaszczuk, Katarzvna J. Blinowska and M. Kowalczyk
demonstrated the difference between partial coherences and pair coherences. Pair coherences are calculated from the values of two signals in the set. Partial coherences are found from coefficients describing the whole set of signals and they give the coherence between two signals after removing the influence of all the other signals. From the pair coherences misleading conclusions can be drawn when their mutual relationship follows from dependence on the third channel. Results of analysis of the synchronization of ECoG will be presented.
DPW 1.02 INVESTIGATION OF SEIZURE PROPAGATION BY MEANS OF S E G M E N T A T I O N OF T H E MULTICHANNEL EEG. P. Penczek, W. Grochulski and M. Kowalczyk (Warsaw, Poland) A typical epileptic EEG recording consists of a rich variety of peak-like structures, whose shapes and the frequency of occurrence change considerably in time. The overall picture is still more complicated in the multichannel case. If we consider an EEG signal as stationary in a small time interval, use can be made of the multichannel autoregressive model to describe the signal. Since the characteristics of the process are changing in time, an adaptive model is called for such as Kalman filters. In the work presented, the classic single-channel Kalman filter was extended to the multichannel case. In this novel approach, the coefficients of the filter are recalculated for consecutive blocks of samples (several tens of samples per block) to speed up analysis. An automatic segmentation was performed by setting an arbitary level of discrimination for the set of AR coefficients, with the segment ends coinciding in all channels. The segments found were classified into groups of similar spectral characteristics by means of a clustering routine. Phase spectra and directed pair coherences calculated for each group allowed a detailed analysis of seizure propagation. In our case a four-channel epileptic EEG recorded in the hippocampus, thalamus, reticular formation and cortex - was analyzed and the time relations between the channels investigated.
(Warsaw, Poland) We have developed a method for analysis of multichannel EEG based on the autoregressive parametric model (AR), which makes possible the straightforward estimation of multiple and partial coherences. In the AR model, EEG signals are treated as a stochastic time series containing components connected with brain rhythms, and a random noise. The coefficients of the model can be found by minimizing the noise component. From the matrix of these coefficients, coherences and phase spectra can be easily estimated. We have applied this method to the 4-channel ECoG in case of spontaneous activity, after a pain stimulus and under the influence of analgesic drugs. From the muhiple coherences we obtained information about goodness of fit and completeness of the evaluated signal set. We have
DPWI.03 BEST ORDER VARIABILITY GRESSIVE EEG MODELLING.
OF AUTORE-
F. Vaz, P. Guedes De Oliveira and J.C. Principe (Aveiro, Portugal) Autoregressive modelling have been widely applied in EEG studies. Due to computation constraints different criteria to define the smallest 'best" order have been utilized. Some authors always use the same order (based on prior tests) while others choose the model order based on a measure of the rates of change of the error. In this work, the Akaike information criterion was used in 2