Surgery – Diagnostic testing – Measuring or detecting nonradioactive constituent of body...
Reexamination Certificate
2002-05-29
2003-11-25
Winakur, Eric F. (Department: 3736)
Surgery
Diagnostic testing
Measuring or detecting nonradioactive constituent of body...
C600S323000, C702S019000
Reexamination Certificate
active
06654623
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention relates to the suppression of superimposed interferences in a measurement signal with a substantially periodic useful signal.
The measurement of signals may roughly be subdivided in general into a) the recognition of individual, more or less singular events and b) the monitoring of more or less frequently repetitive, substantially periodic signals. In either case, superimposed interferences limit the reliability of the measurement, and the object is to avoid, to suppress, or to filter out these interferences.
For the purposes of the present document, periodic signals should be understood to be those signals for which the useful signal has at least a periodic component, at least over a given domain of time, the frequency of which may indeed be time-dependent.
The recognition of the useful signal and the suppression of interferences is essential in particular in the field of medical patient monitoring, because interferences may lead to a false interpretation of the measurement values, or may render the measurement wholly useless.
A measurement which was found to be particularly sensitive to interfering influences is the pulsoximetrical determination of the oxygen content of the blood, because pulsoximetry is much more strongly influenced by movement artifacts than by the pulse signal determining the blood oxygen. Pulsoximetry involves the non-invasive, continuous determination of the oxygen content of the blood (oximetry), based on an analysis of the photospectrometrically measured pulse. It is necessary for this that a pulse curve (plethysmogram) is available at several wavelengths. In practice, almost all appliances operate with only two wavelengths, which renders it possible to achieve inexpensive, compact solutions. The principle of photometry is based on the fact that the quantity of absorbed light is determined by the degree of absorption of a substance and by the wavelength. Pulsoximeters utilize this effect in that the arterial blood volume, and exclusively the arterial blood volume, pulsates in the rhythm of the heartbeat. If a conclusion is to be drawn as to the value of the oxygen saturation from the obtained measurement data, a ratio of values is derived from the measurement data, which ratio then represents the oxygen saturation value. The basics and application possibilities of pulsoximetry are generally known and have frequently been described, in particular in EP-A-262778 (with a good summary of the theory), U.S. Pat. No. 4,167,331, or by Kästle et al. in “A New Family of Sensors for Pulsoximetry”, Hewlett-Packard Journal, vol. 48, no. 1, pp. 39-53, Feb. 1997
Methods proposed for recognizing and suppressing artifacts in pulsoximetrical measurements are in particular processes in the time domain, adaptive filters, spectral analyses, and methods in the time-frequency domain.
Among the methods in the time domain is the peak method, in which the basic signal is subdivided into the individual pulses, and the ratio value is determined from the extreme values of a pulse (cf. EP-A-102816 or U.S. Pat. No. 5,349,519). The essence of this artifact suppression is the comparison of properties such as amplitude, time between maximum and minimum, length, etc., of an identified pulse with those of a reference pulse which was derived from preceding pulses. A further method in the time domain is the ECG synchronization as described in U.S. Pat. No. 4,802,486. A temporal reference to the peripheral pulse is obtained here, derived from the R-spike in the ECG. In the split-wave method (cf. U.S. Pat. No. 5,386,026), the basic signal is scanned at equidistant intervals, independently of the pulse. Two scanning points are combined each time so as to obtain the ratio. Pulse-independent, continuous SpO
2
values are thus created. A further method in the time domain for interference suppression is described in EP-A-870465. Here the basic signals are reduced to their AC components through constant average value subtraction. Interferences superimposed in the same direction on both basic signals can be eliminated through subtraction of the AC components.
A complex adaptive filter is described in WO 96/12435 and operates in accordance with the principle of echo suppression, known from telephony. A filter is controlled such that an image of the interference component is generated, which is subsequently subtracted from the interfered signal.
Spectral analyses for artifact recognition were published in a research paper by Rusch et al. in “Signal Processing Method for Pulsoximetry”, Comput. Biol. Med., vol. 26, no. 2, pp. 143-159, 1996, in which it is investigated whether it is possible to find the pulse frequency and amplitudes in a simpler manner in the frequency range. Several adjustments of the Fast Fourier transform (FFT) and the Discrete Cosine Transform (DCT) were compared. The algorithm proposed, however, has particular weaknesses in the suppression of movement artifacts.
Further methods of recognizing and suppressing artifacts are described in WO 97/00041, with the proposal to eliminate artifacts by simple mathematics, for which it is assumed that movement interferences lead to the same logarithmic changes in all wavelengths. In U.S. Pat. No. 5,588,429, the fractal dimension of the basic signals is determined, the fundamental idea being that the fractal dimension of a non-interfered signal is small, whereas that of an interfered signal is great. According to U.S. Pat. No. 5,355,882, only the instantaneous DC values are used within interfered intervals, whereas the AC values originate from moments before the interference.
Methods in the time-frequency domain for interference suppression with the use of so-called wavelet transforms are described inter alia in JP-A-10216096 (for living-body signals), U.S. Pat. No. 5,778,881 (for ECG applications), or EP-A-816863 (for radar applications).
U.S. Pat. No. 5,619,998 and U.S. Pat. No. 5,497,777 describe noise filter methods in the time-frequency domain for ultrasound imaging systems. The imaging signals are subdivided into overlapping sub-intervals of equal length. Each of the sub-intervals is transformed by means of a discrete wavelet technique. It is identified for each transformed sub-interval whether the wavelet transform coefficients relate to interferences or to the useful signal. The identification takes place here through the use of adaptive, non-linear threshold value formation. The wavelet coefficients which were selected as relating to the useful signal are retained, and those wavelets which were selected as belonging to interferences are erased. The remaining useful signal wavelet coefficients are transformed back in an inverse discrete wavelet transform.
Coifman et al. in “Experiments with Adapted Wavelet De-Noising for Medical Signals and Images”, published by Metin Akay in “Time Frequency and Wavelets in Biomedical Signal Processing”, IEEE, ISBN 0-7803-1147-7, 1997, pp. 323 ff., also describes an algorithm against interference in the time-frequency domain for medical signals and images. A one-dimensional signal such as, for example, a sound file, is subdivided into windows of a desired length. A wavelet packet transform with a number of filters is attempted in each window, the transform with the lowest entropy is retained as the best basis, and the coefficients are sorted in the order of falling amplitude. The coefficients having an amplitude smaller than a given energy threshold value are eliminated in each window, and a cost function for the coefficients (i.e. how many wavelet packet coefficients does it “cost” to achieve the energy for which all values >>0 are counted) is evaluated repeatedly until the cost is greater than a given cost threshold value. The coefficients not considered (too small) are erased, and a new signal is reconstructed from the remaining coefficients.
It was found to be particularly unfavorable in the two methods in the time-frequency domain mentioned last that the useful signal may also become modified and distorted in the case of an erroneous assignment of t
Vodopia John
Winakur Eric F.
LandOfFree
Interference suppression for measuring signals with periodic... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Interference suppression for measuring signals with periodic..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Interference suppression for measuring signals with periodic... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3178993