Noncontact monitoring of blood oxygen saturation, using camera

Abstract

A device for non-contact measurement of blood oxygen saturation (SpO2) in a mammalian subject including a camera and one or more arrays of LEDs each having a first set of LEDs emitting near infrared (NIR), and the second set of LEDs emitting orange light located in an optical path adapted to transmit reflected light from a subject to the camera. A controller transmits a camera trigger to the camera, and is further coupled to transmit control signals to the one or more arrays of LEDs. A processor receives photoplethysmography (PPG) data signal values from the camera. The PPG data signal values are present in the reflected light and include pulsatile and non-pulsatile components. The processor determines SpO2 values from the PPG data signal values from the measured ratios of pulsatile to non-pulsatile components of the PPG signals.

Description

TECHNICAL FIELD

The present invention relates to non-contact measurement of blood oxygen saturation (SpO₂) in a mammalian subject, and, more particularly to a device having a camera, one or more arrays of LEDs emitting near infrared (NIR) and orange light, and a controller.

BACKGROUND

Oxygen saturation, along with heart rate, breathing rate, blood pressure, and body temperature, is a vital physiological parameter. It is a relative measure of the oxygen amount dissolved or carried in a given medium, such as blood. It indicates whether a person has sufficient supply of oxygen and reflects the health level of the cardiorespiratory system. Continuous monitoring of oxygen saturation level is important in detecting hypoxemia under many medical situations, including anesthesia, sleep apnea, and parturition. It is employed in intensive care, operating room, emergency care, neonatal care, sleep study, and veterinary care [1].

Mixed venous oxygen saturation (SvO₂), tissue oxygen saturation (StO₂), and arterial oxygen saturation (SaO₂) are a few major methods used for the determination of oxygen saturation levels in human body. SvO₂ is a measurement of the oxygen remaining in the blood after passing through the capillary bed, which indicates moment-to-moment variation between oxygen supply and demand [2]. It can be monitored using fiber optics catheters. StO₂ provides an assessment of tissue perfusion and it can be measured by near infrared (NIR) spectroscopy. SaO₂ is a measurement of oxygen saturation in the arteries. An estimation of SaOd₂ at peripheral capillary is called SpO₂, which is the primary focus of this paper. Unlike traditional SaO₂ measurement, which is normally conducted invasively via a blood test with a blood gas analyzer, SpO₂ can be measured by noninvasive methods. Monitoring SpO₂ provides a quick and convenient assessment of user's oxygenation status. The most widely used device for SpO₂ monitoring is pulse oximeter, which is often attached to the finger for measurement purpose. The hardware implementation of pulse oximetry includes two main components: 1) an illumination source usually composed of a dual-wavelength LED, and 2) a photodetector—typically a photodiode. SpO₂ values typically range from 95% to 100% in healthy individuals. Continuous low SpO₂ levels (<90%) may indicate an oxygen delivery problem [3].

Recent technological advances have enabled measurements of some of the physiological signals in noncontact ways [4]-[6]. Remote SpO₂ detection provides people with a method to measure oxygen saturation noninvasively under normal daily setting. Absence of physical contact between the subject and the device allows for a more comfortable and less stressful measurement condition. The inaccurate SpO₂ readings caused by varied pressure applied from finger to the contact sensor can also be avoided [7], besides preventing skin irritation that can occur in some individuals, especially infants, during extended monitoring periods. Noncontact pulse oximetry also provides a suitable SpO₂ measurement alternative for individuals with finger injuries, or for those with poor peripheral perfusion or dark pigmentation on fingers, for whom traditional pulse oximetry may otherwise lead to inaccurate measurements [8].

In recent years, researchers have attempted different SpO₂ measurement approaches using noncontact methods. For example, Humphreys et al. [9], [10] used a CMOS camera with LED arrays that emit two different wavelengths as the light source for noncontact pulse oximetry. Due to low frame rate and sensitivity to ambient light, the noise in the measured PPG signals was too large to obtain accurate SpO₂ values. Wieringa et al. [11] also used a CMOS camera, but with three different wavelengths to investigate the feasibility of an “SpO₂ camera.” However, no SpO₂ results were presented due to poor SNR of the PPG signals. Kong et al. [12] used two CCD cameras, each mounted with a narrow bandpass filter to capture PPG signals at two different wavelengths (520 and 660 nm) in ambient lighting condition. The test only covered a narrow SpO₂ range (97%-99%). For practical applications, such as clinical settings, it is necessary to be able to measure SpO₂ over a wider range (at least 80%-100%). Tarassenko et al. [13] and Bal et al. [14] used a camera to calculate SpO₂ based on the PPG information obtained from the RGB channels under ambient lighting condition. Other researchers have found that the PPG signals extracted from the red and blue channels were noisier than those extracted from the green channel [6], [15], [16]. Moreover, for digital cameras, each color channel (red, green, or blue) covers a band of optical spectrum [17] with a width of ˜100 nm, which is different from the traditional pulse oximetry method that uses monochromatic light sources with wavelengths selected to maximize the detection sensitivity of oxygenated and deoxygenated hemoglobin in blood. Tsai et al. [18], [19] used a CCD camera with red and infrared LEDs to take still images of hand and analyzed SpO₂ by looking into the reflective intensity of the shallow skin tissue. These authors compared the SpO₂ results against partial pressure of oxygen values (PaO₂), instead of the standard pulse oximetry. Although they showed correlation between the results obtained using the two methods, a direct demonstration of SpO₂ measurement is still lacking.

The present invention, in contrast to the above discussed methods, is a new noncontact method which is based on video recording of a subject's facial area to measure SpO₂. To the best of the inventors' knowledge, this is the first demonstration of a low-cost video-based method with high temporal resolution and signal-to-noise ratio to accurately monitor wide range of SpO₂ without any physical contact between the subject and the device. The contributions of this study include: 1) development of a hardware system with video capture and illumination synchronization control, 2) identification of optimized light sources to achieve accurate PPG and SpO₂ detection when using noncontact method, 3) validation of method over a wide clinically relevant range of SpO₂ via a pilot study of subjects, and 4) addition of SpO₂ tracking to our previously reported noncontact physiological monitoring platform, which can detect heart rate, breathing pattern, and pulse transit time [4].

BRIEF SUMMARY OF THE DISCLOSURE

This summary is provided to introduce, in a simplified form, a selection of concepts that are further described below in the Detailed Description. This summary is not intended to identify key features of the claimed subject matter, nor is it intended to be used as an aid in determining the scope of the claimed subject matter.

Disclosed herein is device for non-contact measurement of blood oxygen saturation (SpO₂) in a mammalian subject including a camera and one or more arrays of LEDs each having a first set of LEDs emitting near infrared (NIR), and the second set of LEDs emitting orange light located in an optical path adapted to transmit reflected light from a subject to the camera. A controller transmits a camera trigger to the camera, and is further coupled to transmit control signals to the one or more arrays of LEDs. A processor receives photoplethysmography (PPG) data signal values from the camera. The PPG data signal values are present in the reflected light and include pulsatile and non-pulsatile components. The processor determines SpO₂ values from the PPG data signal values from the measured ratios of pulsatile to non-pulsatile components of the PPG signals.

BRIEF DESCRIPTION OF THE DRAWINGS

While the novel features of the invention are set forth with particularity in the appended claims, the invention, both as to organization and content, will be better understood and appreciated, along with other objects and features thereof, from the following detailed description taken in conjunction with the drawings, in which:

FIG. 1 shows an example of absorption spectra of HbO₂ and Hb.

FIG. 2A shows PPG signals measured simultaneously using the camera with red (λ=660 nm) LEDs.

FIG. 2B shows PPG signals measured simultaneously using the camera with orange (λ=611 nm) LEDs.

FIG. 3 shows a relationship between SpO₂ and RR (with wavelengths at 610 and 880 nm) based on equation (6) using extinction coefficients from [20].

FIG. 4. shows simulated plots of normalized RR versus SpO₂ based on four different wavelength combinations.

FIG. 5 shows an example of an experimental setup and control signals.

FIG. 6 shows an example of an image sequence acquisition with two different wavelengths.

FIG. 7 shows an example of an AC value obtained from PPG signal.

FIG. 8A shows an example of the average intensity of PPG signals obtained at 611 nm.

FIG. 8B shows an example of the average intensity of PPG signals obtained at 880 nm.

FIG. 8C shows an example of FFT spectra for 611 nm.

FIG. 8D shows an example of FFT spectra for 880 nm.

FIG. 9 shows an example of image intensity changes due to SpO₂ variation. Each PPG signal has been smoothed by a 100-point moving average filter.

FIG. 10A and FIG. 10B show examples of SpO₂ measured using a pulse oximeter (reference method) and using the new method disclosed herein.

FIG. 11 shows an example of correlation between the lowest SpO₂ values obtained from the presented noncontact and reference contact methods.

FIG. 12 shows an example of a Bland-Altman plot showing the difference between the SpO₂ values measured using the presented noncontact method and the commercial contact pulse oximetry versus the average values of the two methods.

In the drawings, identical reference numbers identify similar elements or components. The sizes and relative positions of elements in the drawings are not necessarily drawn to scale. For example, the shapes of various elements and angles are not drawn to scale, and some of these elements are arbitrarily enlarged and positioned to improve drawing legibility. Further, the particular shapes of the elements as drawn, are not intended to convey any information regarding the actual shape of the particular elements, and have been solely selected for ease of recognition in the drawings.

DETAILED DESCRIPTION

The following disclosure describes a device for non-contact measurement of blood oxygen saturation (SpO₂). Several features of methods and systems in accordance with example embodiments are set forth and described in the figures. It will be appreciated that methods and systems in accordance with other example embodiments can include additional procedures or features different than those shown in the figures. Example embodiments are described herein with respect to device for non-contact measurement of blood oxygen saturation (SpO₂) including a camera and LED arrays. However, it will be understood that these examples are for the purpose of illustrating the principles, and that the invention is not so limited.

Unless the context requires otherwise, throughout the specification and claims which follow, the word “comprise” and variations thereof, such as, “comprises” and “comprising” are to be construed in an open, inclusive sense that is as “including, but not limited to.”

Reference throughout this specification to “one example” or “an example embodiment,” “one embodiment,” “an embodiment” or combinations and/or variations of these terms means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the present disclosure. Thus, the appearances of the phrases “in one embodiment” or “in an embodiment” in various places throughout this specification are not necessarily all referring to the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.

DEFINITIONS

Generally, as used herein, the following terms have the following meanings when used within the context of microarray technology:

The articles “a” or “an” and the phrase “at least one” as used herein refers to one or more.

As used herein, “plurality” is understood to mean more than one. For example, a plurality refers to at least two, three, four, five, ten, 25, 50, 75, 100, 1,000, 10,000 or more.

As used in this specification, the terms “processor” and “computer processor” encompass a personal computer, a tablet computer, a smart phone, a microcontroller, a microprocessor, a field programmable object array (FPOA), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), a programmable logic array (PLA), or any other digital processing engine, device or equivalent capable of executing software code including related memory devices, transmission devices, pointing devices, input/output devices, displays and equivalents.

“Obtaining” is understood herein as manufacturing, purchasing, or otherwise coming into possession of.

“PPG” refers to photoplethysmography signals.

“ROI” refers to region of interest.

“RR” refers to RR, ratio of ratios, refers to the ratio of absorbance at two wavelengths (λ1 and λ2).

“SNR” refers to signal-to-noise ratio.

“SpO₂” refers to peripheral capillary oxygen saturation, an estimate of the amount of oxygen in the blood.

EXAMPLE EMBODIMENTS

Referring now to FIG. 1, an example of SpO₂ measurement using a dual-wavelength imaging system is shown. SpO₂ is the percentage of oxygenated hemoglobin at peripheral capillary and can be expressed by the following equation,

$\begin{array}{cc}{\mathrm{SpO}}_2=\frac{\left[{\mathrm{HbO}}_2\right]}{\left[{\mathrm{HbO}}_2\right]+\left[\mathrm{Hb}\right]}\times 100\%.& \left(1\right)\end{array}$ where [HbO₂], plot 10, is the concentration of oxygenated hemoglobin and [Hb], plot 20, is the concentration of deoxygenated hemoglobin.

Traditional pulse oximetry measures SpO₂ based on the differential absorption of light by HbO₂ and Hb at two wavelengths. Depending upon the optical absorption spectrum of HbO₂ and Hb shown, it is possible to select two wavelengths, λ₁ and λ₂, such that absorbance by HbO₂ is more at λ₂ than at λ₁, while the absorbance by Hb is more at λ₁ than at λ₂.

The Beer-Lambert law—widely used to determine a solution's configuration by optical transmittance measurement—states that light absorption by a substance in a solution is proportional to its concentration [21]. Pulse oximetry assumes that the pulsatile component (AC) of optical absorption originates from the pulsatile arterial blood, and the non-pulsatile component (DC) contains contributions from non-pulsatile arterial blood, venous blood, and other tissues. The pulsatile signals (AC) can be normalized by the non-pulsatile signals (DC) at λ1 and λ2, to give the pulsatile absorbance rates as follows,

$\begin{array}{cc}{R}_{{\lambda }_1}=\frac{{\mathrm{AC}}_{{\lambda }_1}}{{\mathrm{DC}}_{{\lambda }_1}}.& \left(2\right)\\ R_{{\lambda }_2}=\frac{{\mathrm{AC}}_{{\lambda }_2}}{{\mathrm{DC}}_{{\lambda }_2}}.& \left(3\right)\end{array}$

The ratio of absorbance at two wavelengths is defined as ratio of ratios, RR,

$\begin{array}{cc}\mathrm{RR}=\frac{R_{{\lambda }_1}}{R_{{\lambda }_2}}=\frac{{\mathrm{AC}}_{{\lambda }_1}\text{/}{\mathrm{DC}}_{{\lambda }_1}}{{\mathrm{AC}}_{{\lambda }_2}\text{/}{\mathrm{DC}}_{{\lambda }_2}}.& \left(4\right)\end{array}$

RR can be regarded as nearly linear with respect to SpO₂ [12], [14], [22],SpO₂=k×RR+b.  (5)where k and b are linear equation coefficients. Thus, SpO₂ value can be obtained by measuring RR. This dual-wavelength ratio method provides an easy way to determine SpO₂, and the result is independent of both light path length and concentration of blood constituents that absorb light.

The presently disclosed method of SpO₂ detection is based on optical principles similar to traditional pulse oximetry. A key difference is the ability to track SpO₂ change using a noncontact method based on the reflected light as disclosed herein.

As mentioned above, SpO₂ detection using the RR value requires employing at least two wavelengths. For accurate measurement, it is preferable that 1) the measured PPG signals have high SNR at both wavelengths, and 2) optical absorption associated with HbO₂ is opposite to that associated with Hb, and the differences between them are large at the two wavelengths, as shown in FIG. 1.

Traditional contact-based pulse oximetry uses a dual-wavelength LED at red (λ=660 nm) and near infrared (NIR) (λ=940 nm) wavelengths as light source, and a photodiode as light detector. For transmission-mode pulse oximetry, the LED and photodiode are placed at either sides of the tissue (e.g., finger or earlobe), and for reflection-mode pulse oximetry, the LED and photodiode are positioned on the same side of the tissue. As shown in FIG. 1, the red light 10 at 660 nm is absorbed more by Hb than by HbO₂, while the NIR light 20 at 940 nm is absorbed more by HbO₂ than by Hb. The 660 and 940 nm wavelength combination produces high-quality data for the contact pulse oximetry, but it is not suitable for the presented noncontact method.

It was observed that the use of red LED at 660 nm for the non-contact method results in poor PPG signal. To find a suitable replacement for the 660-nm LED, we evaluated the noncontact PPG signals at various wavelengths, ranging from 470 to 940 nm, and observed that the best PPG signal is obtained when green light is used, which is consistent with the literature [6], [15], [16]. However, the optical absorption difference between HbO₂ and Hb at green is small (see FIG. 1); thereby, making green unsuitable for SpO₂ measurement. We ruled out blue because its optical absorption is similar to NIR with high HbO₂ absorbance and low Hb absorbance despite the fact that its use produces good-quality PPG signal [23]. We determined orange (λ=590 to 635 nm) to be the most suitable substitute for our application because its optical absorption property fulfils the specified criteria—high Hb and low HbO₂ absorbance—and due to the superior PPG signals (shown in FIG. 2A and FIG. 2B) measured via the noncontact method when using orange LED, as compared to red LED. FIG. 2A, shows PPG signals measured simultaneously using the camera with red (λ=660 nm) LEDs. The PPG signals 30 are plotted with intensity in angstrom units (a.u.) on the Y axis and time in seconds on the X axis. Referring now to FIG. 2B shows PPG signals measured simultaneously using the camera with orange (λ=611 nm) LEDs. The PPG signals 32 are plotted with intensity in angstrom units (a.u.) on the Y axis and time in seconds on the X axis.

We also examined the suitability of the 940-nm LED—used in conventional contact pulse oximetry—for our imaging system, and found the PPG signal obtained using it to be unsatisfactory due to low SNR. The primary reason for this was the low quantum efficiency of the CMOS imager at 940 nm. NIR at wavelength 880 nm was found to provide better quality PPG signal obtained using the camera sensor. Moreover, 880 and 940 nm have similar optical absorptions by HbO₂ and Hb. These two reasons prompted the use of 880-nm LED, instead of 940-nm LED, in conjunction with the 610-nm orange LED for the presented method.

Performance of the 610-nm orange and 880-nm NIR combination was examined with a simulation of the dependence of RR on SpO₂ from 70% to 100%. The simulation was based on the Beer-Lambert law and the assumption that absorption of light in blood is only related to HbO₂ and Hb, which lead to the following RR equation:

$\begin{array}{cc}\mathrm{RR}=\frac{s\times {\varepsilon }_{{\mathrm{HbO}}_2{\mathrm{\_\lambda }}_1}+\left(1-s\right)\times {\varepsilon }_{{\mathrm{Hb}\mathrm{\_\lambda }}_1}}{s\times {\varepsilon }_{{\mathrm{HbO}}_2{\mathrm{\_\lambda }}_2}+\left(1-s\right)\times {\varepsilon }_{{\mathrm{Hb}\mathrm{\_\lambda }}_2}}.& \left(6\right)\end{array}$ where s is the oxygen saturation (SpO₂), and ε_HbO2-λi and ε_{Hb_λi} are the extinction coefficients of HbO₂ and Hb at the two wavelengths [21].

Referring now to FIG. 3, a relationship between SpO₂ and RR (with wavelengths at 610 and 880 nm) based on equation (6) using extinction coefficients from [20] is shown. Line 34 is a linear fit for plot points 36. The simulation result indicates that SpO₂ is linearly proportional to RR (R²˜1) with a maximum error <0.6% over a broad range (70%-100%), and can be approximated by equation (5). The linear relationship is consistent with other studies [12], [14], [22]. Coefficient k in equation (5) can be estimated from the linear fit curve slope, which is −12.1

Referring now to FIG. 4, simulated plots of normalized RR versus SpO₂ based on four different wavelength combinations are shown. Similar simulations with different wavelength combinations were carried out to determine the best wavelength combination for accurate SpO₂ measurement. A plot of normalized RR on the y-axis is plotted against SpO₂ on the X axis where curve 40 represents Red (λ=660 nm) and NIR (λ=880 nm), where curve 42 represents Orange (λ=610 nm) and NIR (λ=880 nm), where curve 44 represents Green (λ=528 nm) and NIR (λ=880 nm). Backspace, and where curve 46 represents Blue (λ=470 nm) and NIR (λ=880 nm). The extinction coefficients used are from reference [20].

For easy comparison of results at different wavelength combinations, each plot was normalized by the corresponding RR value at 100% SpO₂. The steeper the curve, the more sensitive a combination is to SpO₂ change. The red/NIR combination, which is widely used in the traditional pulse oximetry, shows the steepest curve. When SpO₂ drops from 100% to 70%, RR changes by 319%, indicating that this combination is most sensitive to SpO₂ change. However, for the green/NIR and blue/NIR combinations, RR changes by only 14% and 5%, respectively, indicating their unsuitability to detect SpO₂ changes. The orange/NIR combination shows a change of RR by 190%. Although it is not as good as the red/NIR combination, it is the best choice for noncontact SpO₂ tracking when the SNRs of the PPG signals are considered.

Hardware Implementation

Referring now to FIG. 5, an example of an experimental setup and control signals is shown. The experimental setup includes a PixeLINK monochromatic camera 50, model number PL-B741 EU, with a Fujinon HF16HA1B 16-mm f/1.4 fixed focal lens was used to record the videos. The illumination system consisted of two identical LED arrays 52 placed symmetrically on the left and right sides of the camera. Each array included alternating rows of NIR (QED223, Fairchild Semiconductor) 54 and orange (SLI-570DT3F, Rohm Semiconductor) LEDs. A microcontroller (Texas Instruments MSP430F5348) 60 was connected to the camera by camera trigger 66, to the 611 nm LEDs by control line 62, and to the 880 nm LEDs by control line 64. The microcontroller 60 was used to generate LED signals 162, 164 and camera trigger signals 166 to switch the NIR and orange LEDs on and off alternatively so as to capture an image every 50 ms at the camera's trigger signal rising edge when either the NIR or the orange LEDs were on. The camera 50 was triggered 20 times/s, so for each wavelength, the corresponding frame rate was 10 frames/s. All videos were taken indoors without ambient lights to avoid noise caused by ambient lighting.

Experiment Design

Still referring to FIG. 5, subjects 70 were asked to sit still approximately 30 cm from the camera and LED arrays. As long as clear focus and proper size were guaranteed for the region of interest (ROI), the distance did not affect the signal much, which was partially due to ac/dc normalization. A blindfold was used for comfort and eye protection. Each experiment lasted 5 min. The normal SpO₂ range is 95%-100% in healthy people. To validate the presented method for low SpO₂ (<90%), the subjects were asked to hold their breath until they felt uncomfortable in order for their blood to reach low oxygen saturation level—a technique also used by other researchers [12], [22]. There is no known risk associated with holding breath for 0.5-1.5 min in healthy people. To produce a noticeable drop in SpO₂, the subjects must hold their breath for at least 30 s. Notwithstanding this equal time duration, the SpO₂ drop observed varied from one individual to another because of the varying lung capacity and hemoglobin oxygenation efficiency. It was noted that after holding breath for 1 min, SpO₂ dropped below 90% in some subjects while in others, the drop was much smaller.

The subjects were asked to breathe normally for the first 2 min, during which the SpO₂ value was stable due to the sufficient oxygen supply. After this initial 2-min period, each subject was asked to hold breath as long as possible to produce an SpO₂ drop. When the subject resumed breathing, SpO₂ recovered to the same level as that before holding breath, usually within a few.

SpO₂ measurements were also carried out simultaneously using a (not shown) commercial contact pulse oximeter (Homedics PX-100) for comparison. The commercial contact pulse oximeter provided a reading nearly every 10 s.

Data Processing

Now referring to FIG. 6, an example of an image sequence acquisition with two different wavelengths shown. A previous study [4] conducted at Arizona State University has shown that the area around the lips provides a suitable region for good PPG signal measurement. For this reason, an area of 160×120 pixels around the lips was selected as the ROI. After capturing the videos, the ROI was analyzed using the following procedure. The acquired images 70 were sorted into two groups, viz., NIR 77 and orange 75, based on the wavelength at which they were captured. In each group, the image intensity was averaged over all the ROI pixels in every frame to obtain the PPG signal at the corresponding wavelength. Each of the two PPG signals were divided equally into 10-s subsets to provide SpO₂ data with a time resolution similar to the commercial contact pulse oximeter that was used as a reference.

Now referring to FIG. 7, an example of an AC value obtained from PPG signal is shown. For each of the 10-s subsets, the ac and the dc components of the PPG signal were obtained using the average peak-to-peak and mean values, respectively. Fast Fourier transformation (FFT) can also be used to extract the PPG signal ac component [24]-[29], but it works well only when the heart rate remains constant, and yields inaccurate results when heart rate variability is high. The RR values were determined from equation (4) by using the measured ac and the dc PPG signal components. To extract SpO₂ from RR, equation (5) was used, where k was determined from the slope of the plot shown in FIG. 3, and the intercept b was determined from the baseline SpO₂ level and corresponding average RR value obtained during the initial 2 min of each test.

Validation of the Wavelength Selection

Now referring simultaneously to FIG. 8A-FIG. 8D examples of the average intensity of PPG signals and corresponding FFT spectra obtained at 611 nm are shown 880 nm in separate plots. The detection of noncontact PPG signals was validated at 611 nm and 880 nm. PPG signals were recorded at the two wavelengths simultaneously, both showing heart beating clearly as plotted FIG. 8A and FIG. 8B. AC/DC normalization was used for compensating the difference in image intensities at the two wavelengths [see equations (2), (3)]. Referring now more particularly to FIG. 8C and FIG. 8D, FFT spectra of the simultaneously recorded PPGs at the two wavelengths show pronounced peaks at 1.5 Hz, which correspond to the heart rate.

Now referring to FIG. 9, an example of image intensity changes due to SpO₂ variation it is shown. Plotted are PPG signals 90, 92 at 611 nm and 880 nm, respectively. Each PPG signal has been smoothed by a 100-point moving average filter. When SpO₂ drops, HbO₂ concentration decreases, and Hb concentration increases. In this case, we expect that more orange light and less NIR light will be absorbed. Consequently, the reflectance of orange light will drop and that of the NIR light will increase. When SpO₂ increases, opposite changes in the reflectance are expected. Plot 92 shows that the average intensity at 880 nm increased when the subject held breath (SpO₂ dropped) and decreased after the subject resumed breathing (SpO₂ increased). Plot 90 shows an opposite trend at 611 nm. These observations are consistent with the optical absorption properties of HbO₂ and Hb at 611 and 880 nm.

SpO2 Measurement

Now referring jointly to FIG. 10A and FIG. 10B, examples of SpO₂ measured using a pulse oximeter (reference method) and using the new method disclosed herein are shown. Two sets of measurements performed for validation purpose, wherein SpO₂ values were obtained and plotted against time using the presented method as plotted on curves 106, 108 and were compared against those obtained using the reference device as plotted on curves 102, 104 every 10 s over a 5-min measurement duration. The comparisons shown indicate that the SpO₂ measured using the noncontact method is consistent with that measured using the contact-based reference method. The stable SpO₂ value at 98% corresponds to the normal breathing period of 2 min and the evident reduction corresponds to the time period for which breath is held. The SpO₂ value restoration corresponds to the resumption of normal breathing. A delay (˜10 s) in the reading of the reference pulse oximeter was corrected for comparison with the noncontact SpO₂ detection.

Small-Scale Pilot Study

Referring now to FIG. 11, an example of correlation between the lowest SpO₂ values obtained from the presented noncontact and reference contact methods. Line 110 is a linear fit of the data points. To demonstrate the robustness of the disclosed noncontact method to monitor SpO₂, a small-scale pilot study was conducted and statistical analysis of the data was completed. Six subjects were enrolled in the Institutional Review Board study approved by Arizona State University (No. STUDY00002240). The subjects included different genders (three males, three females), ages (27.3±2.8 years old, mean±SD), and skin colors. Informed consents were obtained from all the subjects following an approved protocol. None of the subjects had any known respiratory disease. The test was repeated as described above on different subjects and the lowest SpO₂ values were compared as determined by using the presented method and reference pulse oximetry. Plot 111 is a plot of the lowest SpO₂ values from 43 tests and linear least square regression. A good linear correlation (R²=0.87) was found between the presented and reference methods over a wide range of oxygen saturation levels. Slope of the linear fitting curve is about 0.86, which is small than the ideal value of 1, with standard error of 0.05. The data are dispersed around the fitted linear curve 110, which may be attributed to subject movement, and light scattering effects.

Now referring to FIG. 12, an example of a Bland-Altman plot showing the difference between the SpO₂ values measured using the presented noncontact method and the commercial contact pulse oximetry versus the average values of the two methods is presented. The mean of the differences between the presented method and reference method is −0.07%. The interval for 95% limits of agreement between the two methods is from −2.65% to 2.51%, which is calculated by mean difference ±1.96×standard deviation of the differences. The root-mean square error is 1.3 and r is 0.936 (p<0.001). Since p<0.05 indicates a significant correlation between the two methods under comparison, we conclude that the observed correlation between our noncontact SpO₂ detection method and the traditional contact pulse oximetry is statistically significant. The lowest SpO₂ observed during testing for this study was 83%. Lower SpO₂ values were difficult to achieve by holding breath in healthy individuals. A person with SpO₂ lower than 80% is considered to be in a state of hypoxia. Our method was validated for SpO₂ values ranging from 83%-100%, which is the normal oxygen saturation level range in most healthy individuals. Thus, the presented method can be used for daily SpO₂ monitoring.

The invention has been described herein in considerable detail in order to comply with the Patent Statutes and to provide those skilled in the art with the information needed to apply the novel principles of the present invention, and to construct and use such exemplary and specialized components as are required. However, it is to be understood that the invention may be carried out by different equipment, and devices, and that various modifications, both as to the equipment details and operating procedures, may be accomplished without departing from the true spirit and scope of the present invention.

REFERENCES

The teachings of the following publications are incorporated herein in their entirety by this reference.

• [1]M. Tavakoli et al., “An ultra-low-power pulse oximeter implemented with an energy-efficient transimpedance amplifier,” IEEE Trans. Biomed. Circuits Syst., vol. 4, no. 1, pp. 27-38, January 2010.
• [2] Understanding Continuous Mixed Venous Oxygen Saturation (SvO2) Monitoring With the Swan-Ganz Oximetry TD System, Edwards Life-sciences LLC, Irvine, Calif., USA, 2002, p. 1.
• [3] What Does SpO₂ Mean? What is the Normal Blood Oxygen Level?” Withings, Issy-les-Moulineaux, France (2015). [Online]. Available: withings.zendesk.com.
• [4] D. Shao et al., “Noncontact monitoring breathing pattern, exhalation flow rate and pulse transit time,” IEEE Trans. Biomed. Eng., vol. 61, no. 11, pp. 2760-2767, November 2014.
• [5]M.-Z. Poh et al., “Advancements in noncontact, multiparameter physiological measurements using a webcam,” IEEE Trans. Biomed. Eng., vol. 58, no. 1, pp. 7-11, January 2011.
• [6] W. Verkruysse et al., “Remote plethysmographic imaging using ambient light,” Opt. Exp., vol. 16, no. 26, pp. 21434-21445, December 2008.
• [7] V. Kamat, “Pulse oximetry,” Indian J. Anaesth., vol. 46, no. 4, pp. 261-268, 2002.
• [8] P. E. Bickler et al., “Effects of skin pigmentation on pulse oximeter accuracy at low saturation,” Anesthesiology, vol. 102, no. 4, pp. 715-719, 2005.
• [9] K. Humphreys et al., “A CMOS camera-based pulse oximetry imaging system,” in Proc. IEEE Eng. Med. Biol. Soc. Conf., Shanghai, China, 2005, vol. 4, pp. 3494-3497.
• [10] K. Humphreys et al., “Noncontact simultaneous dual wavelength photoplethysmography A further step toward noncontact pulse oximetry,” Rev. Sci. Instrum., vol. 78, no. 4, p. 044304, 2007.
• [11] F. P. Wieringa et al., “Contactless multiple wavelength photoplethysmographic imaging: A first step toward ‘SpO₂ camera’ technology,” Ann. Biomed. Eng., vol. 33, no. 8, pp. 1034-1041, 2005.
• [12] L. Kong et al., “Non-contact detection of oxygen saturation based on visible light imaging device using ambient light,” Opt. Exp., vol. 21, no. 15, pp. 17464-17471, 2013.
• [13] L. Tarassenko et al., “Non-contact video-based vital sign monitoring using ambient light and auto-regressive models,” Physiol. Meas., vol. 35, no. 5, pp. 807-831, 2014.
• [14] U. Bal, “Non-contact estimation of heart rate and oxygen saturation using ambient light,” Biomed. Opt. Exp., vol. 6, no. 1, pp. 86-97, December 2014.
• [15] U. S. Freitas, “Remote camera-based pulse oximetry,” in Proc. 6th Int. Conf. eHealth, Telemed., Social Med., Barcelona, Spain, 2014, pp. 59-63.
• [16] J. Lee et al., “Comparison between red, green and blue light reflection photoplethysmography for heart rate monitoring during motion,” in Proc. IEEE Eng. Med. Biol. Soc. Conf., 2013, pp. 1724-1727.
• [17] J. Jiang et al., “What is the space of spectral sensitivity functions for digital color cameras?” in Proc. Workshop Appl. Comput. Vis., 2013, pp. 168-179.
• [18] H.-Yi Tsai et al., “A study on oxygen saturation images constructed from the skin tissue of human hand,” in Proc. IEEE Int. Instrum. Meas. Technol. Conf., Minneapolis, Minn., USA, May 2013, pp. 58-62.
• [19] H.-Y. Tsai et al., “A noncontact skin oxygen-saturation imaging system for measuring human tissue oxygen saturation,” IEEE Trans. Instrum. Meas., vol. 63, no. 11, pp. 2620-2631, November 2014.
• [20] S. Prahl, Optical Absorption of Hemoglobin, Oregon Medical Laser Center, Portland, Oreg., USA (1998). [Online]. Available: http://omlc.org/spectra/hemoglobin/summary.html
• [21] G. M. Azmal et al., “Continuous measurement of oxygen saturation level using photoplethysmography signal,” in Proc. IEEE Int. Conf. Biomed. Pharmaceutical Eng., Singapore, 2006, pp. 504-507.
• [22] C. G. Scully et al., “Physiological parameter monitoring from optical recordings with a mobile phone,” IEEE Trans. Biomed. Eng., vol. 59, no. 2, pp. 303-306, February 2012.
• [23] L. F. C. Martinez et al., “Optimal wavelength selection for noncontact reflection photoplethysmography,” in Proc. SPIE, 22nd Congr. Int. Commission Opt., vol. 8011, p. 801191 November 2011.
• [24] W. S. Johnston, “Development of a signal processing library for extraction of SpO₂, HR, HRV, and RR from photoplethysmographic waveforms,” M.S. thesis, Dept. Biomed. Eng., Worcester Polytechnic Inst., Worcester, Minn., USA, 2006.
• [25] T. L. Rusch et al., “Alternate pulse oximetry algorithms for SpO₂ computation,” in Proc. IEEE Eng. Med. Biol. Soc. Conf., Baltimore, Md., USA, November 1994, vol. 2, pp. 848-849.
• [26] J. E. Scharf et al., “Pulse oximetry through spectral analysis,” in Proc. 12th Southern Biomed. Eng. Conf., New Orleans, La., USA, April 1993, pp. 227-229.
• [27] J. E. Scharf and T. L. Rusch, “Optimization of portable pulse oximetry through Fourier analysis,” in Proc. 12th Southern Biomed. Eng. Conf., New Orleans, La., USA, April 1993, pp. 233-235.
• [28] T. L. Rusch et al., “Signal processing methods for pulse oximetry,” Comput. Biol. Med., vol. 26, no. 2, pp. 143-159, 1996.
• [29] J. M. Kim et al., “Signal processing using Fourier & wavelet transform for pulse oximetry,” in Proc. Conf. Lasers Electro-Opt., Chiba, Japan, 2001, vol. 2, pp. 310-311.
• [30] Y. Sun et al., “Motion-compensated noncontact imaging photoplethysmography to monitor cardiorespiratory status during exercise,” J. Biomed. Opt., vol. 16, no. 7, p. 077010, 2011.
• [31] L. Feng et al., “Motion-resistant remote imaging photoplethysmography based on the optical properties of skin,” IEEE Trans. Circuits Syst. Video Technol., vol. 25, no. 5, pp. 879-891, May 2015.
• [32] J. G. Webster, Design of Pulse Oximeters. London, U.K.: Institute of Physics

Publishing, 1997.

• [33] I. Fine and A. Weinreb, “Multiple scattering effect in transmission pulse oximetry,” Med. Biol. Eng. Comput., vol. 33, no. 5, pp. 709-712, 1995.
• [34] G. Zonios et al., “Pulse oximetry theory and calibration for low saturations,” IEEE Trans. Biomed. Eng., vol. 51, no. 5, pp. 818-822, May 2004.

Claims

1. A device for non-contact measurement of blood oxygen saturation (SpO2) in a human subject having a face, comprising: a camera;a microcontroller;an illumination system, coupled to receive control signals from the microcontroller, consisting of a first LED array placed proximate one side of the camera and a second LED array symmetrically placed proximate another side of the camera each of the first and second LED arrays having a first plurality of NIR LEDs emitting near infrared (NIR) light, and a second plurality of orange LEDs emitting orange light alternating with the first plurality of NIR LEDs, where the first and second LED arrays are located in optical paths adapted to transmit reflected light from the face to the camera;wherein the microcontroller is connected to the first and second LED arrays to generate control signals for alternately switching the NIR LEDs and orange LEDs on and off for both the first and second LED arrays;wherein the microcontroller is also coupled to transmit a camera trigger to the camera, where the camera generates video signal values sensed from the reflected light where the reflected light includes pulsatile components and non-pulsatile components;a processor for receiving the video signal values from the camera; andwherein the processor determines SpO2 values that are proportional to the ratios of the absorbance of the NIR LEDs and the orange LEDs as transmitted in the reflected light.

2. The device of claim 1, wherein the NIR LEDS comprise a plurality of LEDs having an emitted nominal wavelength of 880 nm, and the orange LEDs have an emitted wavelength in the range of 590 nm to 635 nm.

3. The device of claim 1, wherein the camera captures an image at the camera's trigger signal rising edge when either the NIR or the orange LEDs are illuminated.

4. The device of claim 3, wherein the camera comprises a monochromatic camera.

5. A method for non-contact measurement of blood oxygen saturation (SpO2) in a human subject having a face, comprising: locating an illumination system consisting of a first LED array and a second LED array placed on respective sides of a camera each of the first and second LED arrays having alternating sets of NIR LEDs emitting near infrared (NIR), and orange LEDs emitting orange light in an optical path adapted to transmit reflected light from the face to the camera;transmitting reflected light from the face to the camera by alternately illuminating the NIR LEDs and orange LEDs for both the first and second LED arrays;operating the camera to capture the reflected light to produce a plurality of video signal values;processing the video signal values from the camera, wherein the video signal values include pulsatile components and non-pulsatile components; anddetermining SpO2 values from the video signal values as proportional to the ratios of the absorbance of the NIR light and the orange light after they are emitted from the first and second LED arrays and then reflected from the face.

6. The method of claim 5, wherein illuminating the first and second arrays of LEDs comprises illuminating the NIR LEDS including a plurality of LEDs having a nominal emitted wavelength centered around 880 nm, and the orange LEDs including an emitted wavelength in the range of 590 nm to 635 nm.

7. The method of claim 5, wherein operating the camera comprises transmitting camera trigger signals to the camera.

8. The method of claim 7, wherein operating the camera further comprises capturing an image when the camera's trigger signal produces a rising edge when either the NIR or the orange LEDs are illuminated.

9. The method of claim 8, wherein the act of capturing an image comprises capturing a monochromatic image.

10. The method of claim 8 wherein the act of capturing an image comprises capturing reflected light from a region of interest including upper and lower lips of the face.

11. The method of claim 10 wherein the region of interest comprises an area of 160 pixels×120 pixels around the upper and lower lips.

12. The device of claim 3, wherein the camera captures an image at 10 frames/s.