2. Goal: Detect Medical Problems by non-
Invasively Measuring Optical
Properties of Blood
Optical Blood Medical Example
Property Diagnostic Problem
(Symptom)
Fluorescence Zinc Protoporphyrin Iron Deficient
(ZPP) Concentration Anemia, Lead
Poisoning
Absorption Hemoglobin Anemia
Concentration
Scattering Hematocrit Low- Blood loss,
High - Dehydration
3. Beer-Lambert Law
Optical approaches leverage the Beer-Lambert
law which uses 3 variables to model light
entering (Ii) and exiting (Io) a sample.
Io = Ii 10 –2.303 ε(λ) c t /(64,500 g Hb/mole)
t – Sample thickness – cm
c – concentration of absorbent - g/liter (A typical
value of c for whole blood is 150 g Hb/liter.)
ε(λ) – extinction coefficient of absorbent, which
is a function of the light’s wavelength.
– Blood’s ε(λ) is very large for λ < 600 nm
– This is why a flashlight shone through the hand only
transmits red light.
4. Prior Work: Pulse Oximetry
A Pulse Oximeter non-
invasively determines Pulse
and Blood Oxygenation using
differential light absorption.
A Photodiode measures light
intensity of Red (660nm) &
Infrared (940 nm) LEDs shone
through a finger during
systole/diastole.
Blood Oxygenation is a
function of the “Ratio of
Ratios” of Light Intensities at
these points.
Webster, JG. Design of Pulse Oximeters
5. Prior Work: Masimo pt 1
Most non-invasive blood λ Hb02 ε(λ) Hb ε(λ)
nm cm-1/M cm-1/M
sensors extend the pulse
oximeter’s differential 610 1506 9443.6
absorption trick. Like the
620 942 6509.6
pulse oximeter, they
630 610 5148.8
target the finger using
660 319.6 3226.56
wavelengths above 600
nm, the so called ”Medical 700 290 1794.28
Spectral Window”, to 730 390 1102.2
collect enough light. 805 844 730.28
905 1209.2 769.8
Source: Masimo [US7377794]
7. Prior Work: Samsung pt 1
Samsung’s work states “Three λ Hb02 ε(λ) Hb ε(λ)
variables of R569,805, cm-1/M cm-1/M
R569,940, and R569,975
were used for calibration and 569 44496 45072
prediction models.”
producing this comment by 660 319.6 3226.56
one author:
805 844 730.28
G Yoon <gyoon@snut.ac.kr> -
“569nm is highly absorbing in 940 1214 693.44
tissue and, at the same time,
569nm intensity is small 975 1144 389.288
compared with that at longer
wavelength. That is why you
may not get good signal. We Samsung reported >8% error for
used a custom-made LED array Hemoglobin [Jeong et. Al. 2002] and
that has several chips of 569nm
to increase intensity.” Hematocrit [Yoon et. al. 2005].
8. Prior Work: Samsung pt 2
569 nm lies
outside the
Medical
Spectral
Window
Medical
Spectral
Window
9. Our Research: Target Thumb Webbing
for Absorption instead of Finger
Our hypothesis: We can
extend the medical spectral
window by offsetting the
increase in extinction
coefficient at λ< 600 nm with a
decrease in sample thickness
[Sabrina Paseman 2008] did
this by targeting the thumb
webbing for fluorescence
measurements.
We do this for absorption.
This allows us to get better
absorption measurements by
increasing the signal and so
decreasing noise based error.
It also allows us to detect
additional sources of error.
10. The Apparatus
We created an adjustable width clip
that fits either the subject’s index
finger or thumb webbing.
One end of the clip holds the same
5-LED package used by Samsung
and the other holds a fiber optic
cable which connects to an Avantes
AvaSpec-2048 spectrometer.
The LEDs and spectrometer are
controlled by an Arduino
microcontroller.
This allows us to double check the
LED’s wavelength, determine LED
intensity, see if there are any LED
artifacts (emissions at wavelengths
besides the primary wavelength),
and observe any swamping or
fluorescing effects.
11. Experiment 1: Approach
A clamp is adjusted to fit the subject’s index finger.
The subject removes their finger so that the
spectrometer/LED distance can be measured with a
micrometer.
The thumb webbing is placed over the entire LED.
The spectrometer auto-adjusts the integration time
to a full scale reading and the data is collected.
The subject removes their thumb webbing and places
their index finger over the LED. The data is collected
with the same integration time and plotted on the
same axis for comparison.
12. Experiment 1: Data
569 nm Subject 1
Thumb webbing
Index Finger
54054.5 counts
Integration Time: 404.05 ms
Width: 0.7175 cm
4131.5 counts
13. Experiment 1: Findings
General: Transmission increased 13 fold.
Artifacts: Samsung’s LED chip has an artifact at
around 875nm when the 569nm LED is lit. If
Samsung’s device uses a photodiode to collect light,
especially at low intensities, much of the collected
light would come from the artifact rather than the
569nm peak.
Race: This method limited the quality of the readings
we could take from subjects with darker skin.
Age: Young people have smaller thumb webbing than
older people.
14. Experiment 2: Approach
Hold LED underneath the thumb webbing
and the spectrometer on the other side.
Let the spectrometer auto-integrate and
capture the graph.
For data analysis, find the ratio between the
ratios of the peak counts at 569nm and the
integration time. Counts thumb webbing
Integration timethumb webbing
Absorption Amplification =
Countsindex finger
Integration timeindex finger
21. Experiment 2: Data Summary
569 nm
Subject Scopetw Integrationtw Scopeif Integrationif Ampli-
(ADC (ms) (ADC (ms) fication
Counts) Counts)
1 32847.801 99.70 13640.500 3097.90 74.825
2 60849.000 208.91 33273.000 1736.15 15.198
3 55602.750 114.73 60534.000 3808.11 30.488
Subject1: 58 years old - light skinned
Subject2: 52 years old - dark skinned
Subject3: 16 years old - light skinned
22. Key Finding:
“Artifact Amplification”
Why does the artifact size increase?
Blood is a low pass filter, attenuating small λ
(high frequencies) more than large λ.
LED Artifacts (usually) appear “above” (at a
larger λ) the primary emission λ.
If the LED’s primary λ is below 600 nm and the
artifact is above, the artifact will appear
differentially amplified.
This amplification is a key consideration when
designing with LEDs that have this characteristic.
23. Conclusions
Transmission through thumb webbing is at least
13x better than through the finger, but the exact
ratio varies from subject to subject.
“Artifact Amplification” was confirmed by both
experiments and will be a key design
consideration going forward.
Both phenomena would partially explain
Samsung’s error numbers.
Low readings from people with darker skin in
the first experiment and smaller thumb webbing
for younger subjects indicate that probe design
will be a key issue going forward.
24. Further Research
See if there is a statistically significant
difference between the light absorption of
systolic and diastolic blood at 569, 660, 805,
940 and 975 nm.
Correlate more medical problems to
absorption differentials and fluorescent
phenomena.
– See if there is a statistically significant difference
between the fluorescence of systolic and diastolic
blood excited at 425 nm.
Collect
data for more subjects with varying
melanin contents and ages
25. References
[Sabrina Paseman 2008] Paseman, Sabrina. The Ferrometer:
A Device to Detect Iron Deficient Anemia via Non-Invasive
Optical Measurement of Zinc Protoporphyrin. Issue brief no.
SO499. Los Angeles: University of Southern California, 2008.
PDF file.
[US7377794] "Multiple Wavelength Sensor Interconnect” – p57
lists Masimo’s wavelengths
[Yoon et. Al. 2005] Yoon, Gilwon, Ph.D, et al. "Development of a
Compact Home Health Monitor for Telemedicine."
TELEMEDICINE AND e-HEALTH 11.6 (2005): 660-67. PDF File.
[Jeong et. Al. 2002] Jeong, Kye Jin, Su-Jin Kim, and Kun
Kook Park. "Noninvasive Total Hemoglobin Measurement."
Journal of Biomedical Optics 7.1 (2002): 45-50. PDF file.
“Tabulated Molar Extinction Coefficient for Hemoglobin in
Water” http://omlc.ogi.edu/spectra/hemoglobin/summary.html