Proceedings of the National Conference on Trends in Engineering Applications-2019

Velalar college of Engineering and Technology, Erode,Tamilnadu-2 nd March,2019

NON INVASIVE DETECTION OF HEMOGLOBIN

RANGE USING SMARTPHONE
V.Monisha*, S.Pavithra*, S.Sabareeswari*, T.Thirumalaivasan*, Dr.N.Jeyashanthi**
*Student, **Professor, Department of Biomedical EngineeringVelalar College of Engineering and
Technology, Erode
.
Abstract—— ANEaMAPP, a smartphone (Samsung Galaxy A7). Also, the
application that noninvasively monitors photoplethysmographic technique
blood hemoglobin concentration using the successfully identifies the frame regions
smartphone’s camera and LED. Hemoglobin which provide intensity,leads to hemoglobin
measurement is a standard clinical tool estimation.
commonly used for screening anemia and Keywords: Hemoglobin; Mobile Health;
assessing a patient’s response to iron Photoplethysmography; Anemia; Camera;
supplement treatments. A light source Blood Screening
shining through a patient’s finger,
performing a chromatic analysis, analyzing I. INTRODUCTION
the color of their blood to estimate Hemoglobin in the blood that
hemoglobin level. evaluate ANEaMAPP on carries oxygen throughout the body, an iron rich
15patients ranging from 19-25 years of age, protein that carries oxygen inside the RBC and
yielding a appropriate value with the gold gives blood its red colour. Conceptually, the
standard blood test. The estimation of measure of hemoglobin is a representation of
hemoglobin range with an imaging array on the oxygen carrying capacity of the patient’s
a mobile phone (camera oximetry) has great blood. This is distinct from oxygen saturation,
potential for mobile health applications as no which measures the oxygen carrying efficiency
additional hardware other than a camera of the blood. The oxygen combining capacity of
and LED flash enabled phone are required. blood is 1.34ml oxygen per gm of Hb. This is a
However, this approach is challenging as the measure of the function of Hb. Hemoglobin
configuration of the camera can negatively level is an important clinical parameter for
influence the estimation quality. Further, the assessing anemia in both chronic and acute
number of photons recorded with the photo conditions, and is among the most commonly
performed blood tests. ANEMIA is a condition
detector is largely dependent on the optical
where the blood has insufficient red blood cells
path length, resulting in a non-homogeneous to carry oxygen from lungs to the rest of the
image. In this paper we describe a novel body or not enough hemoglobin. Anemia takes
method to calculate the intensity for the several forms and may vary in severity and
captured image to extract a hemoglobin duration. There are several types of anemia
range. We also present a study to select the such as iron deficiency anemia, Sieroblastic
optimal camera settings, notably the white anemia, Thalassemia, normocytic anemia,
balance. The experiments show that the Macrocytic anemia, Sickle cell anemia,
incandescent white balance mode is the Aplastic anemia, Myelodysplatic anemia and
preferable setting for camera oximetry other anemias. The interaction of light with
applications on the tested mobile phone blood plays an important role in optical
 Proceedings of the National Conference on Trends in Engineering Applications-2019
Velalar college of Engineering and Technology, Erode,Tamilnadu-2 nd March,2019

diagnostics and therapeutics—for instance for Camera sensors in modern consumer electronic
the noninvasive assessment of blood devices such as mobile phones or webcams are
composition (Murdy et al., 2009). The oxygen mostly based on lowcost, power efficient and
saturation of blood is defined as the ratio of the highly integrated complementary metal-oxide-
HbO2 concentration to the total haemoglobin semiconductor (CMOS) technology. The sensor
concentration, oxygen saturation (SO2)= is an active pixel array with each pixel
[HbO2]/([HbO2]+[Hb]), and amounts to ∼97.5 consisting of a photodiode and and amplifier
% in arterial blood and ∼75 % in venous circuit. Smartphone cameras offer resolutions
blood(Boulpaep , 2009). Of all blood particles, of up to 41 Mpixel (808 PureView, Nokia,
red blood cells also predominate the scattering Finnland), but the majority of current cameras
of blood with two to three orders of magnitude, are between 2 and 8 Mpixel. Video recording is
arising from the difference in refractive index possible with resolutions of up to HD 1080p
between red blood cells and the surrounding quality. The frame rate is usually limited in
blood plasma .Without the presence of red software to 30 fps, although 60 fps and 120 fps
blood cells, plasma absorption inthe250–1,100- are possible for lower resolutions. Color images
nm region is dominated by various proteins, are obtained by interpolating red, green and
nutritive compounds and/or pharmaceuticals, blue (RGB) pixels in a single image.
while plasma scattering is dominated by II .REVIEW OF RELATED LITERATURE
proteins and platelets .(Meinke et al.,2007)
Under pathological conditions, the absorption WHO report of 1992 stated an
contribution of certain plasma proteins can overall prevalence of 42% of anemia in women
become significant even in the presence of red of developing countries; 51% in pregnant
blood cells, e.g. the absorption of bilirubin women and 41% in non-pregnant women. Out
around 460 nm for jaundiced patients. of these nearly half the total numbers of women
(Bosschaart et al., 2012) In the wavelength are from South East Asia.Later in 1993-2005,
range beyond 1,100 nm, blood absorption is WHO database on global prevalence of anemia
dominated by the absorption of water. Only reported an overall prevalence of 25% where
when water is removed from the blood, several 65.5% of pre-school children, 48.2% of
absorption features due to the presence of pregnant women and 45.7% of non-pregnant
haemoglobin, albumin and globulin can be women were estimated to have anemia. A
identified as small absorption peaks between recent systematic analysis of global anemia
1,690 and 2,400 nm. (Edelman et al., 2012) The burden has reported that the prevalence of
principle of the measurement is based on the anemia has decreased from 40.2% in 1990 to
fact that there is a substantial 32.9% in 2010. Also, females had higher
absorption/transmission difference for light in prevalence and mean severity of anemia in
near infrared region is dominated by diverse virtually all regions and throughout adulthood
hemoglobin derivatives and blood plasma that whereas male children had higher anemia
consists mainly of water. It is well known that prevalence than females. Anemia is a public
pulsatile changes of blood volume in tissue can health challenge in India and mostly affects
be observed by measuring the transmission or women of reproductive age and preschool
reflection of light through the blood volume. children. The District Level Health Survey
(DLHS) 2002-2004 reports an extremely high
Cellular phone application in medical imaging burden of anemia with 98% of adolescent girls
is an emerging area and for the task at hand the and around 96% of pregnant women suffering
technology is exploited for data acquisition as from this condition. The National Family
well as a data conduit. Health Survey (NFHS)-3rd in 2005-2006
 Proceedings of the National Conference on Trends in Engineering Applications-2019
Velalar college of Engineering and Technology, Erode,Tamilnadu-2 nd March,2019

reported that 55% of women and 24% of men (M8) Android based smartphone. The light
were also found to have hemoglobin levels intensity of the mobile spectrometer was
lower than the cut off. Studies have reported an calibrated using Ocean Optics spectrometer.
increasing trend in prevalence of anemia from The light source in their experiment was
49.7% (NFHS-2, 1998-1999) to 50.5% (NFHS- halogen light (HL-2000-HP). They measured
3, 2005-2006) (Bharati et al., 2012). the absorbance at 480 nm and 524 nm where
Moreover, the severity of anemia they found the linear regression with R2 =0 .
was higher in pregnant population (Bharati et 9995 (524 nm) and 0.9971 (480 nm).
al., 2008). A survey of 16 districts from 11 II. EXISTING METHOD
states of India reported 90% prevalence of Direct cyanmethemoglobin method
anemia in adolescent girls and 84.9% in has been the gold standard for hemoglobin
pregnant women (Toteja et al., 2006). A similar estimation and is cheap but time consuming.
prevalence of 90% ingirls has also been The other methods are available such as
reported from a survey of 35 states and union hemoglobin color scale, Sahli technique,
territories of India (Bharati et al., 2009). Lovibond-Drabkin technique, Tallqvist
Despite several interventions through national technique, copper-sulfate method, HemoCue
programmes, anemia is still endemic in women. and automated haematology analyzers. Simple
Low level of hemoglobin among pregnant techniques to measure Hb exist but they are
women results in premature delivery, low birth relatively expensive and require ommercial
weight and prenatal mortality DLHS-2 reported reagents and good technical skills to interpret,
that more than 96% of children are suffering however most of these methods are no longer
from anemia whereas NFHS-3 reported that used.
70% of children in the age group of 6-59
Haldane method for hemoglobin
months are anemic. NFHS-3 also observed that
estimation is one of the oldest methods, based
even for mothers who were not anemic, 62%
had anemic children (659 months). The on carbon monoxide (CO) carrying ability of
prevalence of anemia in children rose steadily blood and comparison with a standard.
with the mother’s level of anemia, reaching However, commercially available standards are
82% for children of mothers who were severely unreliable and fade quickly. Also, the method
anemic. Phillips et al. analyzed the degree of requires a source of CO. (clegg and King, 1942)
agreement between Lab-Hb and SpHb where Dare method uses a small glass chamber which
SpHb was calculated using Masimo pronto is filled with whole blood by capillary action.
SpHb monitor. To collect the user’s data, they The chamber is illuminated by a battery lit bulb
used the ring or index finger where there are no and the color of blood is matched with a
recent injury, surgery, constrictive dressing or standard after viewing through an eyepiece.
any issues that can impede the normal blood (Elwood and Jacobs, 1966)
flow. Finally, the author concluded for ill Lovibonddrabkin technique
people. But the degree of understanding was measures cyanmethemoglobin wherein the
not found so high that we can fully depend on it blood mixes with a solution and after standing
to decide whether the patient needs a time of 3 minutes the color of blood is matched
transfusion or not. with a color standard on discs from which the
Zhang et al. developed a smartphone based hemoglobin concentration is interpolated. The
spectrometer which also works in a visible method was shown to outperform other
range. They used a G- Fresnel diffractive comparators in a study conducted way back in
optical device which worked with HTC One
 Proceedings of the National Conference on Trends in Engineering Applications-2019
Velalar college of Engineering and Technology, Erode,Tamilnadu-2 nd March,2019

1983. The discs used are costly and the estimate give quantitative results, has a subjective end
is prone to error due to subjective interpretation point and is inaccurate.(Sawant et al.,2007)
of the color. (van Lerberghe et al.,1983)
In tallqvist method, a drop of blood III. PROPOSED METHOD
is placed on a strip of blotting paper and the A smartphone application that
concentration of hemoglobin is interpolated by noninvasively estimates blood hemoglobin
comparing the color of the blood with a set of concentration using a smartphone camera.
color standards on paper but the matching of Analysis of the color of the blood in a user’s
standards and sample is extremely difficult. finger yields an estimate of the user’s
(Stone et al., 1984) hemoglobin level. We evaluated the system
In sahli’s/acid hematin method, a using the smartphone’s LED flash and
blood is mixed with N/10 HCl resulting in the incandescent light bulbs as illuminating
conversion of Hb to acid hematin which is sources.
brown in color. The solution is diluted till its Reflection photoplethysmography (PPG) is a
color matches with the brown colored glass of non-invasive technique for detecting relative
the comparator box. The concentration of Hb is blood volume changes due to the cardio-
read directly. vascular pulse travelling throughout the body
A.O.spencer instrument was from the heart. The technique, first reported in
manufactured by an American company uses a the 1930s by Hertzmann (A.B.Hertzman et al.,
drop of whole blood which is placed on one- 1937), exploits the fact that blood absorbs light
half of a glass chamber and agitated with a more than surrounding tissue and variations in
small stick coated with saponin. The chamber is blood volume affect the transmitted and
completed by a cover-glass held firmly by a clip reflected signal components of a suitable
and inserted into a slot in the instrument. optical probe. Clinical PPG applications span
Transmitted green light from a batterylit bulb is monitoring of blood pressure, cardiac output,
compared with a variable standard on a split oxygen saturation (pulse oxymetry), heart (HR)
screen and haemoglobin level is read as g/100 and respiration (RR) rates to assessment of
ml of blood. (Elwood and Jacobs, 1966). autonomous functions and detection of
Copper sulphate method , which is peripheral vascular diseases. PPG systems
commonly used to ascertain Hb for blood combine a dedicated light source operating in
donation is based on specific gravity of blood. the red and/or infra-red (IR) wavelength range
A blood droplet is allowed to fall into copper and a photo detector assembled into a compact
sulphate solution of a specific gravity skin irradiation probe for single spot
equivalent to that of blood with hemoglobin measurement. Here, the focus is PPG imaging
content of 100g/l and the movement of droplet an emerging area (Verkruysse et al.,
is observed. The test is repeated with 2008)realised by replacing the photodiode with
hemoglobin content of 80g/l. The samples are a camera, offering the advantages of real-time
then categorized as below 80g/l, between 80- large area measurement and sensitivity.
100g/l or over 100g/l. (Elwood and Jacobs,
1966) This method evaluated the impact of
CuSO4 method on donor deferral and
concluded that the CuSO4 method does not
 Proceedings of the National Conference on Trends in Engineering Applications-2019
Velalar college of Engineering and Technology, Erode,Tamilnadu-2 nd March,2019

other way around. The degree of infiltration is

likewise subject to the Hb fixation in blood. In
order to find out the concentration of
hemoglobin, transmittance of light through the
Figure 2.0 Imaging setup finger/earlobe was calculated using Beer
Cellular phone application in medical imaging Lambert’s law.
is an emerging area and for the task at hand the Beer’s Law: The intensity of transmitted light
technology is exploited for data acquisition as decreases exponentially as concentration of the
well as a data conduit. substance increases.
Camera sensors in modern consumer Lambert’s law: The intensity of transmitted
electronic devices such as mobile phones or light decreases exponentially as the distance
webcams are mostly based on lowcost, power traveled through the substance increases.
efficient and highly integrated complementary The Beer-Lambert law is the linear relationship
metal-oxide-semiconductor(CMOS) between the concentration of the absorber and
technology. The sensor is an active pixel array the absorbance of Electromagnetic radiation.
with each pixel consisting of a photodiode and The general Beer Lambert's law is given by:
and amplifier circuit. Smartphone cameras offer A = εCL
resolutions of up to 41 Mpixel (808 PureView, Where, A is the absorbance brought
Nokia, Finnland), but the majority of current by the solute, ε is the wavelength dependent-
cameras are between 2 and 8 Mpixel. Video molar absorptivity coefficient of the solute, L is
recording is possible with resolutions of up to the path length and C is the concentration of the
HD 1080p quality. The frame rate is usually solute in the arrangement. Trial estimations are
limited in software to 30 fps, although 60 fps made regarding transmittance (T), which is
and 120 fps are possible for lower resolutions. characterized as:
Color images are obtained by interpolating red, T =I0/I1
green and blue (RGB) pixels in a single image. Where, I0 is the light transmitted
through the sample and I1 is the light without
IV. METHODOLOGY the sample. The relation between absorbance at
a particular wavelength and transmittance is
The light vitality infiltrates through given as:
the skin of a finger or an earlobe to a profundity A= - log (T) = - log [I0/I1]
relying upon the wavelength of the source and For given launching conditions, the extent of
the concentration of Hb and different elements absorbance of light is directly proportional to
of blood. The accuracy of noninvasive optical hemoglobin concentration in blood and
estimation of the focus changes in oxy, de-oxy, transmittance is inversely proportional to
and add up to hemoglobin relies upon wave hemoglobin concentration.
length. Diverse wavelengths of the light vitality
are expected to be transmitted, assimilated and
reflected at different degrees. The measure of
transmitted light is subject to the Hb
centralization of blood. As the measures of Hb
in the blood test builds, more assimilation
happens and less light is transmitted and the
 Proceedings of the National Conference on Trends in Engineering Applications-2019
Velalar college of Engineering and Technology, Erode,Tamilnadu-2 nd March,2019

regions where the signal saturated and beats

could not be extracted. These areas were
concentrated around the edges closest to the
LED (shortest optical path). For these situations
our method was useful to estimate
thehemoglobinrange.

V. RESULTS AND DISCUSSIONS

ANEaMAPP’s hemoglobin
estimations correlate with the CBC’s
predictions with a rank order correlation of
0.69, 0.74, and 0.82 with a mean error of
1.56 g/dL, 1.44 g/dL, and 1.26 g/dL The current results are based on
respectively for each embodiment. The data collected on a Samsung galaxy A7
results of the Pronto are also compared to the device and led light . As such, for the system
CBC, which yields a rank order correlation of to be more accessible, it needs to function on
0.81 with a mean error of 1.28 g/dL. various types of devices and supplemental
For photoplethysmographic lighting. Different brands and models of
imaging, where the primary focus is on pulse phone use different cameras, lenses, and
extraction, the red channel was shown to be filters. The strength of the IR filter will
adequate and with the highest amplitude over strongly influence how strong the IR LEDs
the largest sensor area. This is contrary to need to be for penetration. The necessary
another study that used the green channel for adjustments based on these factors need to be
HR estimation. The discrepancy might originate investigated before the results of
from the fact that the authors in are referring to ANEaMAPP can be generalized to different
their argumentation to non-contact hardwares.
photoplethysmographic imaging experiments
where ambient light is the light source and the REFERENCE
majority of the optical path is in air which 1. Adam I, Ahmed S, Mahmoud MH,
contains less water than tissue and does Yassin MI. Comparison of HemoCue®
hemoglobin-meter and automated
attenuate the wavelengths in question less,
hematology analyzer in measurement of
increasing the proportion of blue and red over
hemoglobin levels in pregnant women
red. Although in our case the red PPG was
at Khartoum hospital, Sudan. Diagn
robust over a large pixel area, it still showed Pathol. 2012; 7: 30.
 Proceedings of the National Conference on Trends in Engineering Applications-2019
Velalar college of Engineering and Technology, Erode,Tamilnadu-2 nd March,2019

2. Anand H, Mir R, Saxena R. (NBM-200) and portable

Hemoglobin color scale a diagnostic hemoglobinometer (HemoCue) with an
dilemma. Indian J Pathol Microbiol. automated hematology analyzer
2009; 52: 360-362. (LH500) in blood donor screening. Ann
3. Bosschaart N, Kok JH, Newsum AM, Lab Med. 2013; 33: 261-267.
Ouweneel DM, Mentink R, van 12. L. Gailite, J. Spigulis, and A. Lihachev,
Leeuwen TG, Aalders MCG (2012) Laser Med. Sci. 23, 189–193 (2008).
Limitations and opportunities of 13. McMurdy J, Jay G, Suner S, Crawford
transcutaneous bilirubin measurements. G (2009) Photonics-based in vivo total
Pediatrics 129:689– 694 hemoglobin monitoring and clinical
4. Boulpaep EL (2009) Blood (chapter 18). relevance. J Biophotonics 5:277–287
In: Boron F, Boulpaep EL (eds) Medical 14. Meinke M, Muller G, Helfmann J,
physiology, 2nd edn. Saunders, Friebel M (2007) Empirical model
Philadelphia, pp. 448– 481 functions to calculate hematocrit-
5. Clegg JW, King EJ. Estimation of dependent optical properties of human
Haemoglobin by the Alkaline Haematin blood. Appl Opt 46:1742–1753
Method. Br Med J. 1942; 2: 329-333. 15. Meinke M, Muller G, Helfmann J,
6. Edelman G, Manti V, Ruth SM, van Friebel M (2007) Optical properties of
Leeuwen T, Aalders M (2012) platelets and blood plasma and their
Identification and age estimation of influence on the optical behavior of
blood stains on colored backgrounds by whole blood in the visible to near
near infrared spectroscopy. Forensic Sci infrared wavelength range. J Biomed
Int 220:239– 244 Opt 12:014024
7. Elwood PC, Jacobs A. Haemoglobin 16. Nkrumah B, Nguah SB, Sarpong N,
estimation: a comparison of different Dekker D, Idriss A, May J, AduSarkodie
techniques. Br Med J. 1966; 1: 20-24. Y. Hemoglobin estimation by the
8. Faber DJ, van Leeuwen TG (2009) Are HemoCue® portable hemoglobin
quantitative attenuation measurements photometer in a resource poor setting.
of blood by optical coherence BMC Clin Pathol 2011; 11: 5.
tomography feasible? Opt Lett 34:1435– 17. Sawant RB, Bharucha ZS,
1437 Rajadhyaksha SB. Evaluation of
9. Haemoglobin color scale practical hemoglobin of blood donors deferred by
answer to a vital need (2001) the copper sulphate method for
Department of blood safety and Clinical hemoglobin estimation. Transfus Apher
Technology. World Health Organization, Sci. 2007; 36: 143-148.
Geneva: WHO 18. Shah VB, Shah BS, Puranik GV.
10. Horecker BL (1943) The absorption Evaluation of non cyanide methods for
spectra of hemoglobin and its hemoglobin estimation. Indian J Pathol
derivatives in the visible and near infra- Microbiol. 2011; 54: 764-768.
red regions. J Biol Chem 148: 173–183 19. Steinke JM, Shepherd AP (1988)
11. Kim MJ, Park Q, Kim MH, Shin JW, Diffuse model of the optical absorbance
Kim HO. Comparison of the accuracy of whole blood. JOSA A 5:813–822
of noninvasive hemoglobin sensor
 Proceedings of the National Conference on Trends in Engineering Applications-2019
Velalar college of Engineering and Technology, Erode,Tamilnadu-2 nd March,2019

20. Stone JE, Simmons WK, Jutsum PJ, and S. E. McLean. A noninvasive
Gurney JM. An evaluation of methods hemoglobin monitor in the pediatric
of screening for anaemia. Bull World intensive care unit. Journal of surgical
Health Organ. 1984; 62: 115-120. research, 195(1):257–262, 2015.
21. Van den Bos R, Arends L, Kockaert M, 30. Toteja GS, Singh P, Dhillon BS,
Neumann M, Nijsten T (2009) Saxena BN, Ahmed FU, Singh RP, et
Endovenous therapies of lower al. Prevalence of anemia among
extremity varicosities: a meta-analysis. J pregnant women and adolescent girls
Vasc Surg 49:230–239 in 16 districts of India. Food Nutr
22. van Lerberghe W, Keegels G, Cornelis
Bull. 2006; 27: 311-315.
G, Ancona C, Mangelschots E, van
Balen H. Haemoglobin measurement:
the reliability of some simple techniques
for use in a primary care setting. Bull
World Health Organ. 1983; 61: 957-965
23. W. Verkruysse, L. O. Svaasand, and J. S.
Nelson, Opt. Express 16, 21434–21445
(2008).
24. Zijlstra WG, Buursma A, van Assendelft
OW (2000) Visible and near infrared
absorption spectra of human and animal
haemoglobin. VSP, Utrecht.
25. Bharati P, Shome S, Chakrabarty S,
Bharati S, Pal M. Burden of anemia and
its socioeconomic determinants among
adolescent girls in India. Food Nutr
Bull. 2009; 30: 217-226.
26. Bharati P, Som S, Chakrabarty S,
Bharati S, Pal M. Prevalence of anemia
and its determinants among nonpregnant
and pregnant women in India. Asia Pac
J Public Health. 2008; 20: 347-359.
27. Bharati S, Pal M, Som S, Bharati P.
Temporal Trend of Anemia Among
Reproductive-Aged Women in India.
Asia Pac J Public Health. 2012.
28. Zhang, G. Cheng, P. Edwards, M.-D.
Zhou, S. Zheng, and Z. Liu. G-fresnel
smartphone spectrometer. Lab on a
Chip, 16(2):246–250, 2016.
29. M. R. Phillips, A. L. Khoury, A. V.
Bortsov, A. Marzinsky, K. A. Short, B.
A. Cairns, A. G. Charles, B. L. Joyner,