Hun H. Sun

An advanced signal processing technique for impedance cardiography

A new design using the latest technique in signal processing, the time-frequency analysis method, was developed to process impedance cardiography signals. This technique, when used to determine the relevant calculation parameters, was found to be more accurate than conventional methods. It was shown to be advantageous in reducing ventilation artifacts and motion noise, resulting in greater accuracy. Its cardiac output values had a much better correlation coefficient when compared in the clinical setting to the standard thermodilution technique than did the values from conventional impedance cardiography devices.<<ETX>>

Non-invasive gastric motility monitor: fast electrogastrogram (fEGG)

We propose a new analysis method to extract the motility information from the electrogastrogram signal that has been recorded at a higher sampler rate than the conventional approaches. This technique utilizes a fourth order Butterworth bandpass filter in extracting the 50-80 cycles per minute (cpm) activity that was previously noted to represent the spike activity range of the cutaneous signals of dogs. Receiver operating characteristics (ROC) analyses have been applied to the processed data to compare the detection performance of our fEGG technique to the conventional approaches that use the slow wave as the reference. The areas under the ROC curves comparing the changes from postprandial stage to fed stage for the fEGG study was found to be 0.961 while for the slow wave it was 0.686. We offer our method as a complementary one to the existing methods.

A study on methods for impedance cardiography

Impedance Cardiography (ICG) has been developed to measure non-invasively cardiac functions, such as stroke volume (SV) and ventricular ejection time (VET) by the use of maximum change of the first derivative of impedance wave /spl Delta/Z(dZ/dt). A series of conventional signal processing methods were first described briefly, then, a new time-frequency analysis method was introduced in detail. Their advantages and disadvantages were studied and compared. The results showed that the latest method has more advantages in reducing ventilation artifacts and motion noise, resulting in greater accuracy. Moreover, the time-frequency analysis method can extract detail information on the transient impedance signal, which is more precise for the practical case. This method may also be powerful in revealing the insight of signal because of its ability to disclose the intimate relationship between time and frequency by the three dimensional display.

Wavelet decomposition method on EEG analysis of G-LOC phenomena

Acceleration (+Gz) induced loss of consciousness (G-LOC) during high +Gz flight maneuvers continues to be a hazard for pilots of high performance aircraft. In the centrifuge studies, G-LOC detection of pilots flying under high +Gz forces is usually made by an observer outside of the gondola and therefore depends upon the reaction time and the alertness of the individual who does the monitoring. The authors propose to use the discrete wavelet transform together with the application of 1/f power distribution theory, to analyze the EEG signals of the pilot during high +Gz simulations. Analyzed by these algorithms, the EEG signal of pilot during G-LOC undergoes significant changes compared to the normal condition, and the authors further propose to classify the conditions of pilot under high +Gz into a set of states. This type of monitoring system may give the observer a cleat indication on the condition of the pilot without human error and in minimum reaction time.

Sinusoidal time-frequency wavelet family and its application in electrogastrographic signal analysis

This paper introduces a sinusoidal time-frequency wavelet family, with both real sine and complex exponential oscillating patterns, of various window functions. It provides a regular time-frequency analysis rather than a time-scale analysis for wavelet transform time-frequency localization of both stationary and non-stationary signals. It offers a convenient trade-off between time resolution at low frequency and frequency resolution at high frequency for any given time-frequency analysis application. Four different window functions are studied for the sinusoidal time-frequency wavelet family and it has been shown that both the real sine wavelets and the complex exponential wavelets using these four windows satisfy the wavelet admissibility condition. The time-frequency resolution of the corresponding wavelet time-frequency localization analysis has also been investigated. It has been show that the rising-cosine window sinusoidal wavelet provides close to optimal time-frequency resolution. It has also been shown that the modulated Gaussian wavelet becomes a special case of the Gaussian window complex exponential wavelet. We briefly describe our sinusoidal wavelet family and apply it to the time-frequency analysis of electrogastrographic (EGG) signals to simultaneously localize high frequency spike activities and track low frequency slow waves.

Two compartment fusion system designed for physiological state monitoring

A two-compartment fusion system designed to reduce high rates of false alarm (FAR) in single channel monitoring systems was tested with physiological data from pilots exposed to high +Gz forces on a human centrifuge. The first compartment expands input signals into the time-frequency domain, where transient changes are captured by wavelet coefficients in frequency ranges of interest. The second compartment optimally combines local decisions of various statistics using a unifying operation rule regardless of individual subject physiology and channel features. Three channels were used to measure respiration, blood pressure, and electroencephalogram under various high performance aircraft maneuver profiles: rapid onset run (ROR) to a fixed plateau, gradual onset run (GOR) at 0.1 Gz per second onset, and simulated aerial combat (SACM) profiles. Pilots sometimes perform anti-G straining maneuvers (AGSM) against the blood pressure drop at head level for greater tolerance. Signals were simultaneously processed to decide the presence of such AGSM. Significant reductions of FAR when detecting AGSM by signal fusion were achieved in our experiment (10/spl sim/38% during ROR/GOR, 25/spl sim/35% during SACM, and 21/spl sim/36% overall), when compared to single channel monitoring. This implies that our approach is very promising and system performance can be enhanced even with poor quality signals.

Detection of pilot state during high-G by wavelet analysis of EEG signal

Large acceleration forces can seriously affect the physiological conditions of a pilot and cause loss of consciousness (G-LOC). There is a lack of a quantitative method to detect the state of a pilot during the high G condition so that a more sophisticated protective device can be designed. The authors propose a G-LOC detection and state determination method by use of wavelet analysis of the EEG signal together with the application of 1/f power theory.

Impedance cardiography in critical-care monitoring

An impedance cardiography (ICG) prototype system developed for the purpose of critical-care monitoring is discussed. This systems operations is completely technician-free, providing a continuous display with digital results and four-channel color waveforms. From 20 patients all with Swan-Ganz catheters in their hearts, 82 correlation studies are reported, showing encouraging results with respect to accuracy as well as on system reproducibility. An anatomically specified electrode configuration allowing precise and reproducible positioning of the electrodes is defined.<<ETX>>

Mathematical modeling of perfluorocarbon chemical (PFC) bio-distribution during liquid ventilation of lambs

Liquid ventilation, a technique through which an infants stiff and fragile lungs are gently expanded by a synthetic liquid, has made enormous clinical progress in recent years. However, during LV, PFC vapors are known to diffuse across the alveoli and be subsequently distributed to various tissues via the arterial system. Since PFC vapors are soluble in tissue lipids, small amount of PFC is thus expected to accumulate in the tissues of the ventilated subject. In this paper, a flow-limited mathematical model, formulated on the mass-balanced principle, has been devised to predict the temporal distribution of tissue PFC concentrations. The solution to the model is dependent on blood flow and partition coefficient, and the modeling results are in general within one standard deviation of the mean experimental tissue concentrations.

Time Domain Analysis Of The Fractal System For Electrode Polarization Phenomenon

It is well known that the Electrode Polarization Phenomenon is a fractal system of l/f type. We shall first derive the time domain representation of the fractal system as a set of linear time-variant differential equations. We shall then present the asymptotic solution of these differential equations from which we can obtain the the dynamic behavior of the system in terms of impulse response and step response. It can be shown that electrode polarization phenomenon is a highly stable, error tolerant with extremely slow and smooth decay having a half-life of approximately 5min, as often observed in the polarized interface phenomenon.