Patent Application
20170035360
It is known to obtain biometric parameter data from a patient surface supporting a patient by use of one or more force sensors which record a force applied by the patient's body surface on the sensor. The patient support surface has a plurality of bladders where pressure data is measured to control the amount of air pressure applied to the bladder. See for example U.S. Pat. No. 7,515,059. Also see U.S. Pat. No. 7,699,784 where the signals output by one or more force sensors are analyzed, such as by a Fast Fourier Transform analysis.
In U.S. Pat. No. 8,413,273 a hospital bed chair is shown with an inflatable bladder wherein a weight of the patient can be measured by measuring the force imparted by the bladder to a load cell (force sensor).
In US Patent Publication 2014/0135635 one or more force transducers are provided for a patient support apparatus. A signal processing is applied to the signals output by the force transducers to determine patient parameters such as blood volume, heart rate, and respiratory rate information. Heart rate and respiratory rate information is derived from the blood volume pulse information. The signal processing to accomplish such extraction is shown in FIG. 5 of the '635 publication and includes the use of Fast Fourier Transform power spectrum analysis, bandpass filters, and power spectral density analysis.
It is an object to utilize a patient support surface to obtain patient biometric parameter data without the use of force sensors impacted by a patient support surface.
In a method or system for obtaining patient biometric parameter data from a patient support surface comprising at least one air pressure bladder inflated by air pressure through an air supply line from an air pump reservoir controlled by a bladder air controller receiving an air pressure signal from a pressure sensor in the air supply line, signal processing a variation of the air pressure signal from the pressure sensor to extract the patient biometric parameter data.
For the purposes of promoting an understanding of the principles of the invention, reference will now be made to preferred exemplary embodiments/best mode illustrated in the drawings and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of the invention is thereby intended, and such alterations and further modifications in the illustrated embodiments and such further applications of the principles of the invention as illustrated as would normally occur to one skilled in the art to which the invention relates are included herein.
Extraction of key biometric indicators passively from the patient environment is an area that is of increasing importance. The present exemplary embodiment addresses a method of extracting biometric indicators from a patient including but not limited to heart rate and breathing rate of a patient using hardware which may already be present in a hospital bed.
According to one exemplary embodiment, a novel method and apparatus is provided of extracting patient biometric data from one or more existing pressure sensors used to control pressure in at least one bladder of a patient support surface to allow better data on a patient's state of health with substantially no additional hardware and using primarily software. Data such as heart rate, breathing rate, or other body sounds can be harvested and analyzed using advance signal processing algorithms to construct a state vector for the patient.
The present exemplary embodiment uses pneumatic pressure sensors already employed to regulate the air pressure inside the patient support surface, colloquially called the mattress. These pressure sensors are typically MEM (micro electro-mechanical system) piezo load beam types of sensors, which have a wide bandwidth output (relative to the bandwidth of the pressure actually being monitored) allowing higher bandwidth signals which emanate from the patient and are impressed onto the surface to be sensed as a small signal perturbation on the much larger, relatively static pressure signal. One commercially available sensor has a 1 kHz sensor bandwidth for zero to full scale pressure readings. Frequency response for smaller amplitude signals is not specified, but is likely even higher. Still, a 1 kHz sensor bandwidth is enough to capture breathing and heart rates, as well as breath sound and cardiac sound capture for audio spectral analysis and anomaly detection. Other sounds produced by the human body may be able to be sensed and processed such as bowel sounds to facilitate early detection of problem conditions developing in patients. Patient motion can be inferred using characterization algorithms to spot pressure variations associated with patient movement.
In the exemplary embodiment the mechanical connection that the surface enclosure has to the volume of air used to support the patient is employed to couple these sounds in the form of pressure fluctuations to the MEMS pressure sensors, which represent these pressure waves as small-signal disturbances around a quiescent point which has variable signal to noise ratio, dependent upon the relative magnitude of the pressure signal which is quasi-static and the biometric signals which are much lower amplitude but higher frequency.
Biometric data capture is not limited to simply air pressure sensors. If other sensors (accels, magnetometers, gyroscopes, electric field sensors, temperature sensors, etc.) are included, then there are myriad biometric data that can be harvested and analyzed to get a better picture of the overall health of the patient.
As shown in
In one exemplary embodiment, the bladder air control system described above associated with the patient support surface 10 may be pre-existing and already installed, such as for a patient hospital bed. With the exemplary embodiment, the pressure sensors 23, 24, and 25 are also utilized to provide respective air pressure signals via electrical lines 28, 29, and 30 to a patient biometric parameter data extraction processor 31 which determines the patient biometric parameter data based on signal processing of a respective variation of the respective air pressure signals. This data extraction processor 31 is preferably connected to a respective display 32 and respective printer 33 for output of extracted patient biometric data for various patient biometric parameters. Furthermore, as previously described, preferably the pressure sensors, which may be pre-existing, have a 1 kHz sensor bandwidth which is sufficient to capture breathing and heart rates as well as breath sound and cardiac sound capture for audio spectral analysis and anomaly detection. As previously indicated these are known as MEMS pressure sensors (micro electro-mechanical system) and piezo load beam types of sensors.
Thus with the method and system of the exemplary embodiment, extraction of patient biometric data can be performed from existing surface pressure sensors to allow better data on a patient's state of health with perhaps no additional hardware and perhaps employing software only.
Each of the functional blocks illustrated in
In other exemplary embodiments other methods of normalizing the data may be employed.
In block 35 for independent component analysis (ICA) transform/component selection, a blind source separation operation is performed using an independent component analysis (ICA) transform. ICA is one technique for separation of independent signals from a set of observations that are composed of linear mixtures of underlying source signals. The underlying signal of interest in this embodiment is the blood volume pulse information (BVP) that propagates through the body. During the cardiac cycle, increased flow through the body's blood vessels results in forces produced by the body on to objects in contact or in proximity to the body. As the BVP changes, the bladder pressure sensors record a mixture of the BVP signal with different weights. These observed signals are denoted by yi′(t), y2′(t) yn′(t) which are signals recorded at time t. In this embodiment the ICA model assumes that the observed signals are linear mixes of the source signals as shown in equation 2 below. In Equation 2 below y′(t) represents a matrix of observed signals, x(t) is an estimate of the underlying source signals and matrix A contains mixture coefficients.
y′(t)=Ax(t) Equation 2
The object of ICA in this embodiment is to determine a demixing matrix W shown in equation 3 below that is an approximation of the inverse of the original mixing matrix A whose output is an estimate of the matrix x(t) containing the underlying source signals. In one embodiment iterative methods are used to maximize or minimize a cost function that measures the non-Gaussianity of each source to uncover the independent sources. In this embodiment ICA analysis is based on the Joint Approximation Diagonalization of Eigenmatrices (JADE) algorithm.
{circumflex over (x)}y′(t)=Wy′(t) Equation 3
Blind source separation ICA analysis in operation block 35 is configured to separate from the pressure signals fluctuations caused predominantly by BVP. In one embodiment data received from the pressure sensors include operation of the support surface 10 including but not limited to percussion and vibration therapy and/or inflation and deflation of the surface 10 and/or operation of any motors on the surface and blind source separation ICA analysis is configured to separate these source signals. In one alternate embodiment identification of components of signals indicative of operation of the surface is aided by predetermined information identifying characteristics of operation of the surface 10. The signal of interest identified in operation block 35 undergoes an artifact suppression process in operation block 36 in this embodiment. The artifact suppression operation block in this embodiment includes interpolation and/or removal of data while in another embodiment data may be normalized after interpolation and/or removal. In yet another embodiment operation block 36 for artifact suppression may be omitted.
Once the ICA signals are generated from operation block 35, the function responsible for BVP is uncovered in discrete fourier transform (DFT) operation block 37 upon generation of power spectrums for the ICA signals. In one embodiment the power spectrum of the signal with the highest peak is selected for analysis. In another embodiment the signal with a peak in power spectrum in the range where BVP is known to exist is selected. This is done automatically in this embodiment. However in other embodiments a caregiver may select a signal of interest using the user interface display 32. In one embodiment, weight determined by the pressure sensors is used to determine if a patient is indeed on the support surface 10. If it is determined that a person is not on the support surface 10 the operation is terminated and a message is displayed to the caregiver in another embodiment.
In operation block (bandpass moving average or Kalman filter) 38 the signal of interest is smoothed. In this embodiment the signal of interest is smoothed using a five point moving average filter and bandpass filter in an area of interest, in this embodiment 0.7-4 Hz. In other embodiments any data manipulation algorithm may be used.
In operation block 39 power spectral density estimation analysis of IBI information is used to identify heat rate variability (HRV) information. In this embodiment a Lomb periodogram is used to analyze HRV.
Information from power spectral density estimation analysis is normalized in normalization operation block 40 in this embodiment, while in other embodiments this operation may be eliminated.
Normalized information from normalization operation block 40 is used in the biometric signal statistics parameter amalgamation operation block 41 which also receives biometric signal statistics from library 42.
Operation block 41 outputs the parameter amalgamation to biometric signal statistics parameter estimation operation block 43 which in turn outputs biometric parameter estimate vectors.
The artifact suppression operation block 36 also outputs to operation block 43 (discrete wavelet transform (DWT)) which in turn outputs to bandpass/moving average or Kalman filter operation block 44, power spectral density analysis block 45, and normalization block 46. The blocks 44, 45, and 46 were previously described in connection with operation blocks 38, 39, and 40. The normalization block 46 outputs to the biometric signal statistics parameter amalgamation block 41 and biometric signal statistics parameter estimation block 43 as previously described.
The artifact suppression block 36 also outputs to signal statistics extraction block 47 which in turn outputs to bandpass/moving average or Kalman filter block 48, power spectral density analysis block 49, and normalization block 50. Blocks 48, 49, and 50 are similar to the description previously provided for blocks 38, 39, and 40. Normalization block 50 outputs to the biometric signal statistics parameter amalgamation block 41 and biometric signal statistics parameter estimation block 43 previously described.
Although preferred exemplary embodiments are shown and described in detail in the drawings and in the preceding specification, they should be viewed as purely exemplary and not as limiting the invention. It is noted that only preferred exemplary embodiments are shown and described, and all variations and modifications that presently or in the future lie within the protective scope of the invention should be protected.
| Number | Date | Country | |
|---|---|---|---|
| 62201324 | Aug 2015 | US |
