A filter to suppress ECG baseline wander and preserve ST-segment accuracy in a real-time environment.

Accurate monitoring of the ST-segment displacements in real-time environments can be distorted by the nonlinear phase response of a baseline filter such as the single-pole, high-pass (0.5 Hz) filter that is standard in the industry today. The authors have previously constructed a four-pole null phase (1.0 Hz) filter that is nearly ideal in suppressing baseline wander while preserving ST-segment accuracy; however, this foreward/backward filter requires capture of a large ECG segment before filtering, thereby producing a delay that is unacceptable in a real-time environment. As a practical compromise, a two-pole, phase-compensated (1.0 Hz) filter was constructed while introducing a small time delay (160 ms). It performs much better than the "standard filter" and almost as well as the "ideal" filter in several tests, namely (1) suppression of baseline wander in a series of ECGs, (2) suppression of artificial baseline, (3) response to a triangular impulse wave (American Heart Association test), and (4) J-point displacement in several ECGs.

Suppression of baseline wander in the ECG using a bilinearly transformed, null-phase filter.

The purpose of this study was to design and test a bilinearly transformed, null-phase (BLT/NP) filter for removing baseline wander and to compare it with the cubic spline for performance. For this purpose, the ECG data were filtered to remove high-frequency noise and low-frequency baseline wander to form a set of "clean" ECGs. Artificial low-frequency noise mimicking typical baseline wander was constructed from sine and cosine waves at 0.20 and 0.45 Hz and with amplitudes of 400 and 300 microV, respectively, and added to the "clean" ECGs to form the "test" ECGs. The BLT/NP filter and the cubic spline method each were applied to a "test" ECG to form a "restored" ECG. The measure of performance was the root mean square difference (RMSD) between the "restored" ECG and the initial "clean" ECG. RMSD values showed that on the average the BLT/NP filter performed as well as the cubic spline method and has the advantage that accurate determination of the QRS onset is not required.

Studies of Trypanosoma cruzi clones in inbred mice. III. Histopathological and electrocardiographical responses to chronic infection.

Histopathological and electrocardiographical (ECG) changes occur in the heart of C3H/HeN and C57BL/6 mice infected for 1 year with Trypanosoma cruzi clones Sylvio-X10/4 (X10/4), Miranda/78 (M/78), or Miranda/80 (M/80). Heart parasitism and a variable degree of inflammation occurred following infection with clones X10/4 or M/78 but not with M/80. Clone X10/4 caused more extensive myocardial inflammation and fibrosis than clone M/78. Myocardial fibrosis was more extensive in C3H than in C57 mice infected with clone X10/4. The normal ECG pattern of C3H mice is distinctly different from C57 mice. The PR intervals of mice infected with clone X10/4 greater than M/78 greater than M/80 approximately equal to controls. ECG abnormalities occurred more frequently in mice infected with clone X10/4 than in controls or mice infected with either M/78 or M/80 regardless of strain or sex and were generally more severe in C57 than in C3H infected with X10/4. First degree atrioventricular block occurred more frequently in C3H mice infected with clone X10/4 or M/78 and C57 mice infected with X10/4 than in all other groups. Complete atrioventricular dissociation occurred frequently in C57 mice infected with X10/4 and rarely in other mice. These results demonstrate that the myocardial response of mice to T. cruzi infection, both histological and electrophysiological, is modulated by both the mouse strain and the parasite isolate used.

Fatigue trends in and the diagnosis of myasthenia gravis by frequency analysis of EMG interference patterns.

Twenty patients with myasthenia gravis were studied. Needle interference patterns at maximal isometric contractions were recorded from the biceps brachii muscles. Each recording lasted for 30 seconds and induced some fatigue. The EMG signals were transformed into power spectra and were analyzed for differences between control and myasthenic fatigue trends and were tested for the power of the frequency variables to classify unknown subjects. Both groups showed a similar averaged spectra for the first 5 seconds. Thereafter, the controls manifested continuous increase in power, and a power peak frequency shift, toward low frequencies. The myasthenics showed an initial increase throughout the frequency ranges; however, later, there was a marked decrease in power and their peak frequency shifted toward the lower frequencies. These fatigue trends differed significantly from one another. Discriminant analysis correctly classified 83% of the subjects. This technique may be helpful in the diagnosis of myasthenia gravis.

A new technique for measuring muscle fiber conduction velocities in full interference patterns.

The motor unit potential shape, mainly its duration and frequency spectra, and the EMG IP crispiness and its frequency spectra are affected by the muscle fiber conduction velocities (MFCVs). Present techniques are somewhat deficient in that they are not adaptable to measure MFCVs continuously and intramuscularly in the presence of interference patterns, and to do so without interfering with the ongoing muscular activity. In this study a cross-correlation with averaging correlograms technique is presented. An EMG needle electrode, with two recording surfaces 1 cm apart, continuously record two channels of EMG activity which is analog-to-digital converted. Contiguous segments of the signals are cross-correlated, the evolved correlograms are averaged together, averaging-out the time-unlocked noise, and averaging-in a peak that represent the average time it takes the EMG signal to propagate from one recording surface to the other. From the distance between these two recording surfaces and the above calculated propagation time the MFCVs can be computed and monitored intramuscularly either in weak or in strong, in isometric or isotonic contractions. But for the fact that a needle is introduced, there is no interference with the muscle electrical activity. It is expected that this technique may add to EMG diagnosis of neuromuscular disorders, will be used to monitor muscular fatigue and applied in normalizing EMG spectra, conditioning them for a better use in diagnostic electromyography.

A hybrid compartmental model for the alligator Purkinje cell. I: Preferred somatopetal conduction of dendritic spikes and soma-axon interaction.

A compartmental hardware model of an alligator Purkinje cell is described, consisting of a branched dendritic tree with four zones of spike generation and electrically excitable soma and initial-segment regions. Passive properties of the model compartments are represented by a cable analog circuit. Simulated action potentials, generated by a combination of depolarizing and hyperpolarizing conductance changes, are triggered in active compartments when the simulated membrane potential passes through preset thresholds. These were set at values corresponding to 28mV depolarization in the dendrites, 22 mV in the soma, and 7 mV in the initial-segment compartment. Synaptic inputs consisting of brief (0.35 msec) rectangular conductance changes give rise to exponentially decaying postsynaptic potentials in the input compartment which are electrotonically spread to other compartments. Orthodromic activation of the model neuron by computer-generated random pulse trains generates a simple spike discharge in the initial-segment compartment without evoking complex spikes. Synchronized excitatory input to the same compartments, however, does evoke a complex spike response in the soma and initial segment, coupled with dendritic spikes. Following antidromic activation of the model neuron, dendritic spikes are not generated, demonstrating a tendency in the dendritic tree for preferential conduction of spikes toward the soma. Investigation of some of the factors underlying this tendency suggests that variations in voltage attenuation due to dendritic geometry, convergence of electrotonically spread dendritic spikes, and the relative durations of dendritic and somatic action potentials may contribute to it. The presence of a threshold gradient in the dendritic tree, proposed by Llinás and his coworkers, was not found to be necessary to explain this tendency toward somatopetal conduction, although it cannot be excluded by the model. Examination of the role of the conically shaped initial-segment region suggests that this zone may provide a low-pass filter for signals conducted electrotonically from the axon to the soma, blocking repolarization of the soma during the complex spike burst generated in the axon.