`Pulse Oximeter Design
`
`Jianchu Yao, M.S. and Steve Warren, Ph.D.
`Department of Electrical & Computer Engineering, Kansas State University
`Manhattan, KS 66506, USA
`
`Abstract
`This paper addresses the design of a plug-and-play pulse oximeter and its application to a
`biomedical instrumentation laboratory and other core Electrical Engineering courses. The low-
`cost, microcontroller-based unit utilizes two light-emitting diodes as excitation sources, acquires
`reflectance data with a photodiode, and sends these raw photo-plethysmographic data to a
`personal computer via an RS-232 serial link. A LabVIEW interface running on the personal
`computer processes these raw data and stores the results to a file. The design of this pulse
`oximeter is unique in two ways: the excitation sources are driven just hard enough to always
`keep the photodiode active (meaning the sensor can be used in ambient light), and the hardware
`separates out the derivatives of the red and infrared photo-plethysmograms so that it can amplify
`the pulsatile component of each signal to fill the range of the analog-to-digital converter. Unlike
`commercial pulse oximeters whose packaging hides the hardware configuration from the
`students, the open, unpackaged design stimulates student interest and encourages dialogue with
`the developer; the in-house nature of the design appeals to students. Moreover, most pulse
`oximeters on the market are expensive and provide users with a front panel that displays only
`percent oxygen saturation and heart rate. This low-cost unit provides unfiltered pulsatile data,
`allowing students to investigate tradeoffs between different oxygen saturation calculation
`methods, test different filtering approaches (e.g., for motion artifact reduction), and extract other
`biomedical parameters (e.g., respiration rate and biometric indicators). Time-domain data from
`these units have been used in linear systems and scientific computing courses to teach filtering
`techniques, illustrate discrete Fourier transform applications, introduce time-frequency
`principles, and test data fitting algorithms.
`
`I. Introduction
`An optical pulse oximeter measures the intensity of light passing through heterogeneous tissue
`and uses variations in this light intensity (primarily resulting from the fractional volume variation
`of arterial blood) to calculate blood oxygen saturation. Due to its non-invasive nature, high
`precision in its operational range, and reasonable cost, optical pulse oximetry is widely adopted
`as a standard patient monitoring technique. Although its foundations date back more than fifty
`years,1 many facets of this technology still attract researchers. Current interest areas include
`motion artifact reduction,2, 3 power consumption optimization,4 low-perfusion measurements,5, 6
`and issues germane to various application environments (e.g., wearability for battlefield and
`home care monitors).7-9 It is important for biomedical engineering students to understand the
`principles of pulse oximetry, hardware/software design issues, and signal processing approaches.
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`APPLE 1023
`
`1
`
`
`
`Pulse oximeter design addresses engineering areas such as optical component selection,
`mechanical layout, circuit design, microprocessor control, digital communication, and signal
`processing. Therefore, a pulse oximeter not only serves as an excellent study vehicle that allows
`students to learn techniques such as photoplethysmographic signal processing; it also provides a
`platform where students can acquire hands-on experience in practical device design. In addition,
`the real-time data that a pulse oximeter offers gives instructors flexibility when assigning
`projects and homework to students of various educational levels (graduate and undergraduate)
`and backgrounds (e.g., electrical engineering or biology).
`
`Many commercial pulse oximeters display calculated parameters (i.e., percent oxygen saturation
`and heart rate) on their front panels, hiding the original unfiltered data from which these
`calculations were made. In this paper, we present an “in-house” pulse oximeter that provides raw
`sensor data for use in the classroom. The device is utilized in bioinstrumentation laboratory
`sessions, and its data provide real-world signals to other core Electrical Engineering courses.
`
`This paper first briefly describes the theory behind photoplethysmographic (PPG) pulse oximetry.
`It then presents the development of a pulse oximeter, emphasizing design features that enable its
`application to education. These features include (a) a stand-alone pulse oximeter module with a
`novel circuit design, an open form-factor, and multiple signal outputs, (b) a personal computer
`station with a flexible, user friendly LabVIEW interface and a variety of signal processing
`options, and (c) the production of raw data that can be used for parameter extraction exercises.
`The paper describes how this device and it features have been applied in classroom environments
`to stimulate student learning. Several examples are introduced in detail, including (a) a pulse
`oximetry laboratory/lecture pair for a bioinstrumentation course sequence, (b) data sources for
`course projects in Linear Systems (EECE 512) and Scientific Computing (EECE 840), and (c) a
`platform upon which undergraduate honors research students can build. This approach can be
`extended to other devices and classes.
`
`II. Theory – Principles of Pulse Oximetry
`PPG pulse oximetry relies on the fractional change in light absorption due to arterial pulsations.
`In a typical configuration, light at two different wavelengths illuminating one side of tissue (e.g.,
`a finger) will be detected on the same side (reflectance mode) or the opposing side (transmission
`mode) after traversing the vascular tissues between the source and the detector.10 When a
`fingertip is simplified as a hemispherical volume that is a homogenous mixture of blood (arterial
`and venous) and tissue, the detected light intensity is described by the Beer-Lambert law: 11
`(
`)(
`)(
`)A
`µ
`−
`µ
`−
`−
`µ
`=
`T
`V
`(1)
`
`
`I
`eI
`e
`e
`av
`aa
`at
`0
`t
`where I0 is the incident light intensity, It is the light intensity detected by the photodetector, and
`µat, µav, and µaa are the absorption coefficients of the bloodless tissue layer, the venous blood
`layer, and the arterial blood layer, respectively, in units of cm-1.
`
`The heart’s pumping action generates arterial pulsations that result in relative changes in arterial
`blood volume, represented by dA, which adds an “ac” component to the detected intensity:
`(
`)(
`)(
`−=
`µµ
`−
`µ
`−
`µ
`−
`V
`T
`A
`dI
`e
`I
`e
`e
`av
`aa
`
`
`
`
`t
`
`0
`
`aa
`
`at
`
`)dA
`
`
`
`(2)
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`2
`
`
`
`Multiple elements contribute to the attenuation of light traveling through tissue, and arterial
`pulsation has only a small relative effect on the amount of light detected (on the order of one
`percent or less; see Figure 1).
`
`
`Absorption due to
`pulsatile arterial blood
`
`1
`
`10
`
`100
`
`Absorption due to non-
`pulsatile arterial blood
`
`Absorption due to
`venous blood
`
`Absorption due to
`skin, bone and tissue
`
`
`Figure 1. Breakdown of the components in the detected photo-plethysmographic signal.12
`
`
`Dividing this change by the dc value normalizes this variation:
`I
`=
`µ−=
`
`dA
`
`aa
`
`t
`
`Id
`
`ac
`
`II
`
`
`
`t
`dc
`The ratio of the above ratio for two wavelengths (‘r’ for red, ‘IR’ for infrared) is given by
`(
`)
`/
`dI
`I
`=
`=
`(
`)
`/
`dI
`I
`,
`
`
`IRa,
`IR
`where µa,i can be expressed as a function of
`,13 arterial oxygen saturation:
`[
`]%0
`
` ( −+
`
`)σ
`µ
`=
`1
`OS
`a
`
`
` IRr,=
`
`are the wavelength-dependent optical absorption cross
`and
`, while
`Here,
`i
`sections of the red blood cells containing totally oxygenated and totally deoxygenated
`hemoglobin, respectively. One can therefore calculate arterial oxygen saturation using
`
`σσ −
`%0
`%0
`R
`(σ
`)+
`( σσ
`)%100
`=
`
`,ra
`
`,IRa
`σ
`
`
`%0
`%0
`R
`
`,IRa
`,ra
`
`
`(3)
`
`(4)
`
`(5)
`
`(6)
`
`µµ
`
`
`
`ra,
`
`OS
`a
`
`2
`
`a
`
`t
`
`t
`
`r
`
`t
`t
`2OSa
`σ
`
`%100
`a
`
`2
`
`R
`
`vH
`
`i
`
`
`
`%100
`
`,ra
`
`−
`
` −
`
`
`
`,IRa
`
`
`
`,
`ia
`
`aσ
`
`%0
`
`aσ
`
`%100
`
`OS
`2
`a
`
`
`
`Equation (6) provides the desired relationship between the experimentally-determined ratio R
`and the arterial oxygen saturation SaO2. Researchers assume this relationship applies to
`monochromatic light sources. In reality, commonly available LEDs are used as light sources and
`typically have spectral widths of 20 to 50 nm. Therefore, the standard molar absorption
`coefficient for hemoglobin cannot be used directly in (6). Furthermore, the simplified
`mathematical description above only approximates a real system that incorporates
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`3
`
`
`
`2
`
`
`
`(7)
`
`2
`
`inhomogeneities and mechanical movement. Consequently, (6) is often represented empirically
`by fitting clinical data to the following generalized function:
`=
`+
`kRk
`OS a
`
`1
`where, e.g., k1= -25.6, k2= 118.814 or k1= -25, k2= 110.15
`
`III. Methods
`A. Pulse Oximeter Development
`As shown in the functional block diagram in Figure 2, a pulse oximeter consists of three main
`units: (1) an optical probe, (2) a circuit module that hosts an analog amplifier, signal
`conditioning element, and microcontroller, and (c) a personal computer that receives data from
`the circuit module and processes, displays, and stores these data.
`
`
`Display
`Processing
`Storage
`
`
`RS-232
`
`Control signals
`
`Microcontroller
`
`Circuit
`Module
`
`Probe
`
`LEDs
`
`Finger
`
`Photodetector
`
`
`
`Current feedback
`
`Light-feedback
`amplifier
`
`Differentiator with
`holding circuit
`
`A/D
`
`Optical signal
`Analog signal
`Digital signal
`
`
`
`Figure 2. Functional block diagram of the pulse oximeter.
`
`
`The analog portion of the pulse oximeter consists of a light-feedback amplifier and an analog
`differentiator with a specialized sample and hold circuit. The current feedback design adjusts the
`light level at the excitation LEDs such that the detected light intensity is constant, keeping the
`photodiode centered in its active region. To improve the stability of this feedback loop, a
`photodiode with smaller gain, rather than a phototransistor, is used as a photodetector. Two
`LEDs with wavelengths of 660 nm and 940 nm were selected as excitation sources.
`
`As discussed earlier, the “ac” component resulting from arterial blood volume variation is very
`small. If A/D conversion is performed on the overall signal, this tiny “ac” component will be
`buried in the “huge” “dc” component after conversion. A differentiator addresses this issue. It
`removes the “dc” component by subtracting the previous signal voltage-level from the present
`signal voltage-level and amplifies this difference, yielding the “ac” component. A hold circuit is
`added to store voltage-levels from the previous sample cycle. The differentiator improves signal
`resolution by allowing one to take advantage of the full range of the A/D converter.
`
`This circuitry is coordinated by a PIC microcontroller. Three output lines control the operation of
`the circuitry, and two A/D inputs sample the desired signal. Two outputs modulate the two light
`sources and switch the charging and discharging of their corresponding hold capacitors. The
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`4
`
`
`
`other output operates the differentiator. The two A/D inputs acquire and digitize two signals: the
`“dc” signal when the differentiator is turned off (it is actually the original signal that includes
`both “dc” and “ac” components) and the amplified difference of the present and previous voltage
`level when the differentiator is turned on.
`
`The PIC microcontroller also operates an RS-232 port to a personal computer running a
`LabVIEW interface. Digitized data are sent to the PC over this RS-232 interface. Because the
`sensor module and personal computer communicate asynchronously, and 8 bytes (two bytes for
`each signal) are sent in each RS-232 packet, a handshaking protocol is used to synchronize the
`two devices. The PC generates an acknowledgement after successfully receiving each data
`packet so that the pulse oximeter module can transmit the next data packet.
`
`On the PC, LabVIEW virtual instruments (a) reconstruct the differentiated data, (b) filter the
`pulsatile signal with motion artifact reduction algorithms, (c) display the differentiated and
`reconstructed waveforms, (d) compute and display values for heart rate and blood oxygen
`saturation (see Figure 4), and (e) store the original and processed data to a text file for follow-up
`analysis. The data in the file are in columnar format:
`Column 1 – Time in milliseconds,
`Column 2 – d(Iac)ir/dt (derivative of the near-infrared signal)
`Column 3 – (Idc)ir
`Column 4 – d(Iac)red/dt (derivative of the red signal)
`Column 5 – (Idc)red
`Column 6 – (Iac)ir/dt (reconstructed near-infrared signal)
`Column 7 – (Idc)red (reconstructed red signal)
`
`
`
`
`Pulse Oximeter
`Module
`
`Reflectance
`Sensor
`
`RS-232
`to PC
`
`Figure 3. Pulse oximeter module and reflectance probe.
`
`
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`5
`
`
`
`
`
`
`
`
`Figure 4. LabVIEW virtual instrument for the pulse oximeter. In addition to heart rate
`and blood oxygen saturation (%), the interface displays the red and infrared derivative
`data (top two waveforms) and the red and infrared reconstructed data (bottom two
`waveforms).
`
`B. A Pulse Oximetry Lecture/Laboratory Pair
`At Kansas State University, the 4-credit-hour Bioinstrumentation course sequence (URL:
`http://www.eece.ksu.edu/~eece772/) consists of three courses instructed by faculty from the
`Department of Electrical & Computer Engineering (EECE) and the Department of Anatomy and
`Physiology (AP). These courses are EECE 772 (Theory and Techniques of Bioinstrumentation, 2
`hours), EECE 773 (Bioinstrumentation Design Laboratory, 1 hour), and AP 773
`(Bioinstrumentation Laboratory, 1 hour). These courses can be taken for either undergraduate or
`graduate credit. The two laboratory hours provide hands-on experience and are intended to help
`students obtain a deeper understanding of concepts learned in lectures.
`
`The pulse oximeter discussed earlier serves as a basis for a lecture/laboratory pair in the
`Bioinstrumentation course sequence. In order to improve the quality of the laboratory, the second
`author designed a laboratory session for AP 773 that uses the pulse oximeter developed by the
`first author. Four sets of devices were constructed and have been used as teaching tools in these
`laboratory sessions. The learning objectives of this laboratory (i.e., what a student should be
`able to do upon completion of the laboratory) are the following:
`• Explain the physiological origin of a photoplethysmogram
`• Describe the hardware and software components required to determine blood oxygen
`saturation using light-based sensors
`• Calculate blood oxygen saturation given a set of red/infrared plethysmograms
`• Assess the character and spectral content of the time-varying signals
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`6
`
`
`
`• Extract physiological data from a photoplethysmogram
`• Describe person-to-person variations in plethysmographic signal data
`• Calculate calibration coefficients using different approaches
`• Counteract the effects of mild motion artifact
`During the laboratory, the class is divided into groups of 2~3 students. Each group is equipped
`with a collection of components: a reflectance probe, a circuit module, a serial cable, and a
`personal computer with the LabVIEW interface installed. The students are first taught how to use
`the modules properly. They then gather PPG data from their team members at different body
`locations and save these data to files for later signal processing.
`
`
`
`Figure 5. Two students acquire photoplethysmographic data in the AP 773 pulse oximetry
`laboratory (Fall 2002).
`
`These data are processed using Microsoft Excel or MATLAB. In addition to observing and
`analyzing time domain data, the students are also required to interpret and understand the
`spectral components of the signal by performing Fast Fourier Transforms (FFTs) on the data sets.
`They implement different methods for calculating the “ac/dc” ratios required to obtain arterial
`oxygen saturation. Two calculation methods are used to compute these ratios. The methods
`correspond to Equations 3 and 4, which supply a parameter for Equation 7. The ‘peak/valley’
`method considers the peak-to-valley amplitude of the reconstructed signal as Iac when calculating
`the “ac/dc” ratio. This method is evaluated with two different filtering techniques: a sliding
`average filter and a sliding median filter. The FFT method uses the spectral peaks of the red and
`near-infrared signals to represent Iac in the calculations. The students are then asked to compare
`the calculation methods and choose the best one.
`
`Students are also encouraged to experiment with other noise reduction filters. Additionally, by
`observing and analyzing waveforms acquired from different team members, students can realize
`that factors such as skin color and perfusion affect the quality of acquired PPG data. They are
`also asked to evaluate the differences between PPG signals acquired at different body locations
`(e.g., wrist, forehead, or ear lobe) that have noticeably different vascular profiles.
`
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`7
`
`
`
`C. Pulse Oximeter Applied to Other Educational Venues
`In addition to the lecture/laboratory pair noted in the previous section, the pulse oximeter design
`and the signal data gathered from various implementations of this design have been applied in
`multiple undergraduate (EECE 499 – Honors Research; EECE 512 – Linear Systems) and
`graduate (EECE 840 – Scientific Computing) educational venues. The signals acquired from this
`platform have been used in the following ways:
`• data for time-domain smoothing algorithms (see Figure 6),
`•
`signals for time- and frequency-domain filtering projects (see Figure 7 and Figure 8),
`• waveforms for Fourier series reconstruction projects (see Figure 9), and
`•
`signals for time-frequency spectrogram projects (see Figure 10).
`The modules have also been used as starting points for various undergraduate honors research
`projects, as depicted in Figure 11.
`
`Course Projects. In the smoothing exercises (see Figure 6), students are asked to perform signal
`processing exercises to ‘smooth out’ variations in signals corrupted with noise. Two of the
`common techniques are illustrated here. Polynomials, by their nature, are smooth curves whose
`numbers of peaks and valleys correspond to the order of the polynomial. In this figure, a
`polynomial of order 12 provides a reasonable representation of the original data set. Note that
`the behavior of the fitting polynomial is unpredictable outside of the original bounds. Sliding
`average and median filters are also a smoothing approach that can be implemented by a young
`student without much programming experience (the graph on the right in Figure 6 was produced
`with an Excel spreadsheet). For this photoplethysmograph (sampled at 160 Hz), a 7-wide sliding
`window appears to provide a reasonable job of smoothing out the noise while retaining the
`fundamental shape of the waveform.
`
`
`
`1.018
`
`0.8
`Time (sec)
`
`1
`
`1.2
`
`1.4
`
`1.6
`
`Sliding average filters:
`7-wide, 25-wide, and 51-wide
`
`1.016
`
`1.014
`
`1.012
`
`1.01
`
`1.008
`
`1.006
`
`1.004
`
`1.002
`
`Signal (V)
`
`1
`
`0
`
`0.2
`
`0.4
`
`0.6
`
`
`Figure 6. Data smoothing algorithms (polynomial fits and sliding average filters) applied to
`photoplethysmographic data. These exercises were assigned in EECE 772
`(Bioinstrumentation) and EECE 840 (Scientific Computing).
`In the EECE 512 project depicted in Figure 7, a student’s code (1) loads a signal from an input
`ASCII text file, (2) performs a convolution (i.e., filtering operation) between the input signal and
`a cascade of 2nd-order Butterworth lowpass and highpass filters (which can be combined to
`create lowpass, highpass, or bandpass filters), (3) saves the output signal to disk, and (4) plots the
`original and filtered signals to the screen. Input signals for these simulations include both ideal
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`8
`
`
`
`signals (e.g., pulses, square waves, and sinusoids) and real-world signals (e.g., biomedical
`signals such as electrocardiograms and light reflectance signals from the pulse oximeter modules
`presented here).
`
`
`
`Lowpass Filter
`Unit Impulse
`Response
`
`Input Signal
`Corrupted by
`Artifacts
`
`
`
`
`
`…
`
`
`
`Filtered
`Output
`Signal
`
`
`
`HPF
`
`LPF
`
`LPF
`
`HPF
`
`
`Figure 7. Multi-stage filtering of photoplethysmographic data via time-domain convolution
`in EECE 512 (Linear Systems). Stages: 2nd-order lowpass and highpass filters.
`Frequency-domain filters are also an important part of a signals and systems course. In these
`projects, a student’s program typically (1) loads an input signal from a file and calculates its
`Fourier transform, (2) calculates the frequency response of a filter chosen by the user, and (3)
`performs a frequency-domain filtering operation on the input signal: it multiplies the input
`signal spectrum by the spectrum of the filter and then takes the inverse Fourier transform of the
`result. The program then saves the input/output signals, their spectra, and the filter spectra to a
`set of ASCII text files and creates a plotting script that can be called by MATLAB or GNUPLOT.
`In the example illustrated in Figure 8, an ideal bandpass filter with a low cutoff of 0.3 Hz and a
`high cutoff or 15 Hz was used to remove the drift and 60 Hz noise present in the original
`plethysmographic signal.
`
`
`
`Output Magnitude
`Spectrum
`
`Output Phase
`Spectrum
`
`Filtered Signal
`
`Original Signal
`
`
`
`Figure 8. Frequency-domain filtering of pulsatile light reflectance data to remove signal
`drift and 60 Hz noise. Course: EECE 512 (Linear Systems).
`
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`
`
`
`
`9
`
`
`
`Figure 9 illustrates the use of light reflectance signals in a Fourier series project. In the left part
`of Figure 9, the top set of axes displays a PPG signal and its Fourier series reconstruction. The
`middle and bottom axes plot the magnitude and phase coefficients, respectively, that were
`calculated for the reconstruction. Note that 45 harmonics (or cosines with different magnitudes
`and phases) were required to replicate the shape of the initial signal. In the canine
`electrocardiogram depicted on the right hand side of the figure, 125 harmonics produced a good
`reconstruction. This is due to the higher frequency components present in each QRS complex.
`
`
`
`Human Pulse Plethysmogram
`
`Canine Electrocardiogram
`
`45 Harmonics
`
`125 Harmonics
`
`
`
`
`Figure 9. Reconstruction of biomedical signal data (human finger photoplethysmogram
`and canine electrocardiogram) using Fourier series. Class: EECE 512 (Linear Systems).
`
`It can be helpful to understand how a signal’s spectral character changes as a function of time.
`Figure 10 presents an example of a MATLAB interface that would be written by a student in a
`graduate scientific computing course. In this figure, the upper left set of axes plots the time-
`domain plethysmogram, while the lower left set of axes displays the spectrum of the signal
`versus time. The plots on the right depict the magnitude and phase spectrum of the input signal
`at the time denoted by the vertical line that occurs at ~55 seconds (see the upper left trace). The
`fields on the right side of the interface depict parameters that can be chosen by the user.
`
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`10
`
`
`
`
`Figure 10. Time-frequency analysis of reflectance data in EECE 840 (Scientific
`Computing).
`
`Honors Research Projects. The undergraduate Electrical & Computer Engineering curriculum
`at KSU allows high achieving students to perform research for course credit. The pulse oximeter
`modules presented in this paper have contributed to five EECE 499 (Honors Research) projects
`to date (see Figure 11). For the project shown at the top of the figure, Ben Young developed a
`system based upon the pulse oximeter module that acquired light reflectance data from the
`forehead using sensors mounted on a firefighter helmet. The goal of this project was to establish
`whether meaningful blood oxygen saturation measurements could be acquired continuously on
`an individual that needed to use their hands freely and could be exposed to dangerous levels of
`carbon monoxide. The second project from the top, managed by Shelly Allison and Craig
`Nelson, involved gathering light reflectance data from normal and hypertensive elderly subjects.
`These data will be analyzed for correlations between spectral behavior and the measured blood
`pressure of the subjects. The goal is to find a comfortable, noninvasive way to replicate the
`information normally provided by often painful blood pressure cuffs.
`
`As noted in Figure 11, Jonathan Hicks investigated a method to use a patient’s light reflectance
`data as a biometric indicator. This capability would allow a home monitoring system to
`authenticate the identity of a patient prior to uploading the patient’s physiological data to a
`remote electronic patient record. The benefits of this approach are two-fold: (1) no interaction is
`required on the part of the patient and (2) the data are independently verified prior to submission.
`The plots in Figure 11 show a representative light reflectance signal for a patient and the single-
`period template used to represent that time-varying signal. Two other representative templates
`are also depicted in the figure to show how these wave shapes vary from person to person. This
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`11
`
`
`
`method uses a statistical test to determine whether a patient’s current data are similar to the
`single-period template stored for the patient. Finally, Austin Wareing was supported by an NSF
`Research Experience for Undergraduates grant to optimize the light reflectance sensor design
`and improve the interaction between the pulse oximeter and the host LabVIEW program. His
`radial sensor design and a resulting set of waveforms are depicted at the bottom of Figure 11.
`
`
`
`
`Ben Young: Forehead
`Measurements of Blood
`Oxygen Saturation for
`Use with Fire Fighter
`Helmets
`
`Shelly Allison and
`Craig Nelson:
`Light-Based
`Indicators for
`Hypertension
`
`Jonathan Hicks: Photoplethysmographic Signals as Biometric Authenticators
`
`
`
`
`
`Austin Wareing: Optimization of Light Reflectance Sensors
`
`
`
`
`Figure 11. Honors research projects that have benefited from the pulse oximeter design.
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`12
`
`
`
`
`IV. Discussion and Conclusion
`This paper presented initial efforts to apply an in-house pulse oximeter design to multiple
`secondary education venues. These efforts have indicated that students enjoy instructional
`experiences that utilize real-world devices, especially when they can manipulate elements of the
`design such as the signal processing algorithms that would normally be hidden from the user.
`The pulse oximeter modules have been used in four Fall offerings of the AP 773 laboratory
`(2001~2004). Because these home-grown pulse oximeters offer improved data access as
`compared to commercial products, instructors can experience far greater flexibility when
`assigning homework, which is especially appreciated when the background and educational
`experiences of the students vary significantly.
`
`Each laboratory session that utilized these modules has been supported by device developers.
`Interactions between the device developers and the students (users) lead to experiences that are
`hard to replicate with packaged, off-the-shelf units. These interactions help the students
`appreciate the concepts discussed in lecture and allow them to become more familiar with the
`device development process.
`
`As noted in the body of the paper, several other undergraduate and graduate courses have
`benefited from the data availability offered by these pulse oximeters. When asked, “What part of
`the project did you like the most” (on the survey for the Spring 2003 Linear Systems project
`depicted in Figure 7) one student responded, “Being able to see the ECG and pulse oximeter
`signals with the noise filtered out.” Many other individuals in this class of 65 students had
`similar opinions about working with data provided by a device in a nearby laboratory. Processing
`real-world signals stimulated the students’ interest the most, followed by the excitement of
`simply getting their code to work. The same Linear Systems student, when asked the question,
`“How could a project of this nature be improved?,” responded with, “More realistic signals to
`filter í WKDW LV ZKDW PDGH PH IHHO OLNH WKLV ZDV D UHDOLVWLF SURMHFW´
`
`The inexpensive hardware, plug-and-play features, and information-rich signals offered by these
`pulse oximeters have also provided starter platforms for honors students that wish to perform
`innovative research. These experiences not only help them to apply knowledge learned from
`their courses and understand recent developments; more importantly, they may also motivate
`these capable students to pursue careers in an expanding biomedical industry.
`
`Acknowledgements
`The authors wish to acknowledge Jerry T. Love, Sandia National Laboratories (retired), for the
`original design of the light feedback circuit. Portions of this material are based upon work
`supported by the National Science Foundation under grant BES–0093916. Opinions, findings,
`conclusions, or recommendations expressed in this material are those of the author(s) and do not
`necessarily reflect the views of the NSF. All studies addressed in this paper that involve human
`subjects have been approved by the Human Studies Board at Kansas State University under
`protocol #2211.
`
`
`
`
`Proceedings of the 2005 American Society for Engineering Education Annual Conference & Exposition
`Copyright © 2005, American Society for Engineering Education
`
`
`
`13
`
`
`
`[2]
`
`[3]
`
`[4]
`
`[5]
`
`[6]
`
`[7]
`
`[8]
`
`[9]
`
`[10]
`
`[11]
`
`[12]
`
`[13]
`
`
`
`References
`[1]
`Brinkman, R. and W. G. Zijlstra. "Determination and continuous registration of the percentage oxygen
`saturation in clinical conditions," Arch. Chir. Neurol., vol. 1, pp. 177-183.
`Swedlow, D. B. "Oximeter with motion detection for alarm modification," International Patent Application
`WO 94y22360 ~13, 1994, pp.
`Yao, J., and S. Warren. "A Novel Algorithm to Separate Motion Artifacts from the Photoplethysmographic
`Signal for a Pulse Oximeter," presented at 4th Joint EMBS-BMES Conf., San Francisco, CA, 2004.
`Rhee, Sokwoo, B.-H. Yang, and H. H. Asada,. "Artifact-resistant power-efficient design of finger-ring
`plethysmographic sensors," IEEE Transactions on Biomedical Engineering, vol. 48, May, pp. 795-805.
`Mannheimer, P.D., J.R. Cascini, M.E. Fein, and S.L. Nierlich. "Wavelength selection for low-saturation
`pulse oximetry," IEEE Transactions on Biomedical Engineering, vol. 44, March, pp. 148 - 158.
`Zonios, G., U. Shankar, and