Content uploaded by Dr Ashok Kumar
Author content
All content in this area was uploaded by Dr Ashok Kumar on Jun 28, 2019
Content may be subject to copyright.
Non-Invasive Bilirubin Detection Technique for
Jaundice Prediction Using Smartphones
Nainika Saini1, Ashok kumar2, Preeti Khera3
1,2,3Electronics and Communication Engineering Department, Ambala College of Engineering and Applied Research,
Ambala, India
1nainikasaini26@gmail.com
2ashokcalicut1993@gmail.com
3kherapreeti33@gmail.com
Abstract – Objective: Jaundice is commonly occurring
ailment in newborns due to rise in the amount of bilirubin
concentration in the body. The aim of our study is early
detection of extreme jaundice to prevent permanent brain
damage or death. Methods: Current detection techniques
involve clinical testing with blood samples or
Transcutaneous Bilirubin (TcB) measurement. A non-
invasive technique of bilirubin detection for jaundice
prediction based on yellow discoloration of skin is
developed. A standalone smartphone based jaundice level
detection app is also developed that jaundice patients can
use to determine biliruibin levels. Results: The proposed
technique smartphone based jaundice detection provides
rank order correlation of 0.93 and provides an improved
method for jaundice prediction with bilirubin
concentration upto 24mg/dl. Conclusion: Since bilirubin
levels peak well after most infants are discharged from
hospital thus a screening system is essential for monitoring
newborn jaundice at home when clinical technologies are
unavailable. Significance: The ability to detect jaundice
early in convenience to patients at home can lead to early
jaundice detection which ultimately reduces healthcare
cost.
Keywords : Jaunidce; Non-invasive; Transcutaneous Bilirubin;
Transcutaneous Serum Bilirubin
I. INTRODUCTION
Jaundice is defined as the yellow discoloration [12] of skin
which can occur in babies, children and adults. It is not a
disease but a medical condition in which excess bilirubin is
produced as a chemical byproduct due to the result of the
breakdown of red blood cells. About half of the newborns
develop jaundice within first few days after birth. Newborns
tend to metabolize bilirubin slower because their liver may not
function to full capacity and also they have higher
concentration of Red Blood Cells (RBCs) than adults.
Consequently, there is an excessive amount of bilirubin in the
blood. This condition is called hyperbilirubinemia [6]. High
concentrations of bilirubin are very harmful in newborns and
can irreversibly damage the brain hence early detection of
jaundice is required.
Clinicians measure the bilirubin levels with either TSB or
TcB. Total Serum Bilirubin (TSB) is an invasive method to
measure the amount of bilirubin directly from blood samples
whereas Transcutaneous Bilirubinometer (TcB) is a non-
invasive alternative that indirectly measures bilirubin. Both of
these methods have their respective limitations. Although
Fig.1. Newborn Baby suffering from Jaundice
invasive technique is the most accurate way but is painful for
newborn babies and also cause delay in the treatment due to
long procedures in laboratory. Transcutaneos Bilirubinometers
which are non-invasive tools, cost thousands of dollars.
Hence, this screening tool is unavailable in most clinics due to
its high cost.
Nowadays, more than a billion people own smartphones
which increase the reach of smartphone medical device apps
[9]. Medical device smartphone apps are software programs
that implement algorithms, which sample the phone cameras
and process their outputs to yield medical information which
is then displayed to its user. In this paper, smartphone based
non-invasive technique for jaundice prediction has been
proposed which provides a better approach to detect jaundice
than the existing methods in terms of both precision and cost.
The rest of paper is divided in four sections: Section II
reviews related work in the area of jaundice prediction using
invasive and non-invasive techniques. Section III presents our
proposed work. Section IV shows results as compared to
existing approaches. Finally, in Section V the study is
concluded with summary and future scope.
II. RELATED WORK
Several researches have been reported over past few years
about non-invasive bilirubin detection and in the area of using
smartphones to diagnose various medical conditions.
In the beginning, researchers have introduced [1] an
electronic device called Transcutaneous Bilirubinometer to
determine that which neonates required TSB detection. This
primary electronic TcB device proved to be useful when
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 14, No. 8, August 2016
1060
https://sites.google.com/site/ijcsis/
ISSN 1947-5500
utilized as a screening strategy for recognizing newborns who
require serum bilirubin determination but newer TcB gadgets
can be utilized as screening devices as well as solid substitutes
for total serum bilirubin testing. A study was undertaken to
relate the correlation between the transcutaneous
bilirubinometer with total serum bilirubin without using
phototherapy [2]. The action levels were found for TcB so as
to evaluate the bilirubinometer as a screening device for
hyperbilirubinemia. Transcutaneous bilirubinometry was used
to determine the amount of yellow color in the skin and
subcutaneous tissue which was further correlated with the
serum bilirubin concentration in neonates. Another
Transcutaneous Bilirubinometer namely, JM-103 was
designed and its usefulness was evaluated [3] in Taiwanese
neonates in order to reduce the frequency of blood sampling
required in newborns to detect jaundice hence prevents severe
neonatal hyperbilirubinemia. The significance in the
prevention of severe hyperbilirubinemia may therefore lie
equally in the post-release follow-up of all neonates as per
guidelines for performing phototherapy or exchange
transfusion [4]. It was discussed [5] that Transcutaneous
Bilirubinometer could be used for early detection of severe
hyperbilirubinemia. It also discussed neonatal jaundice and the
utilization of transcutaneous bilirubin (TcB) measurements for
identification of neonates at risk of extreme
hyperbilirubinemia. Since newborns required serum bilirubin
level measurements before hospital discharge thus assertion
[7] between a Transcutaneous Bilirubin (TCB) and Total
Serum Bilirubin (TSB) was assessed. The TCB correlated but
was not much precise in predicting TSB. Later on, another
bilirubinometer, BiliCheck [8] point-of-care device, was
assessed which performs transcutaneous estimation of
bilirubin by multiwavelength spectral analysis and the results
obtained were compared with TSB. Further, a device for
detection of bilirubin which was ten times cheaper and easy to
use was proposed [10]. It was successfully tested and
confronted with accurate laboratory instruments for bilirubin
measurements. All these Transcutaneous bilirubinometers cost
thousands of dollars as a result of which they are unavailable
in most clinics due to high cost. Then the use of optical
techniques for non-invasive jaundice detection came into
account. The bilirubin detection using optical method [11] was
a better solution to eliminate the baby trauma in the process of
jaundice detection but the measurements were inaccurate in
case of jaundiced neonates receiving phototherapy. Thus
BiliCam, a low cost system that used smartphone cameras to
assess newborn jaundice was presented [12]. BiliCam
evaluated on 100 newborns, yielding a 0.85 rank order
correlation with the serum bilirubin test but it could not
replace TSB testing and can only be used as screening tool.
Also it was not able to detect high bilirubin concentrations
greater than 15mg/dl which may lead to permanent brain
damage or death.
Motivated by these studies, we have proposed a non-
invasive method for jaundice detection using smartphone
where invasive testing is done by the use of spectrophotometer
and non-invasive measurements are first done by using serum
birubin coloration on strips and then compared with those
directly from skin. In case of non-invasive testing, the images
of strips as well as skin were taken from smartphone camera
without causing any pain to newborns. The proposed method
provides a 0.93 rank order correlation and is able to detect
bilirubin concentration upto 24mg/dl to prevent
hyperbilirubinemia.
III. PROPOSED WORK
We propose a non-invasive bilirubin detection technique
for jaundice prediction based on yellow discoloration of skin in
which jaundice can be detected directly from the image of skin.
Since the current gold standard to detect bilirubin levels is total
serum bilirubin (TSB) determination by invasive blood
sampling which is stressful and painful for the neonates thus a
non-invasive technique is needed which is cheaper than any of
the available Transcutaneous Bilirubinometers. We have also
proposed a standalone smartphone based jaundice level
detection app that jaundice patients can use to determine
biliruibin levels.
A. Method of Measuring Bilirubin Concentration using
Invasive Technique
Since the TSB measurement is turned out to be most
successful method to determine hyperbilirubinemia thus we
have first estimated bilirubin levels in laboratory. To measure
bilirubin concentration invasively, serum is extracted from
blood sample. This serum is then mixed with direct bilirubin
reagents and mixture is poured in cuvette. The cuvette is then
placed into a device called spectrophotometer at a wavelength
[10] of 546nm. The absorption value of solvent in
spectrophotometer gives the amount of bilirubin. This gives us
the results for TSB estimation.
TABLE I
TEST PROCEDURE FOR LIQUIMAX BILIRUBIN METHOD
Sr. No.
Reagent
Sample Blank
Test
1.
Direct Bilirubin Reagent
1.00ml
1.00ml
2.
Sodium Nitrite
-
50µl
3.
Distilled Water
50µl
-
4.
Sample
50µl
50µl
The absorption value is noted for sample blank and
bilirubin can be calculated by using equation (1)
Direct Bilirubin mg
dl =Abs of TestAbs of Sample Blank×13.5 (1)
B. Proposed Method for Measurement of Bilirubin
Concentration using Non-Invasive Technique
The bilirubin concentration can be determined from the
yellow discoloration of skin because we have assumed that
visual characteristics of skin can estimate the amount of
bilirubin in neonates. The algorithm to determine bilirubin
concentration involves (1) color balancing, (2) image
segmentation, (3) feature extraction and (4) bilirubin
estimation.
Color Balancing involves the computation of mean values
of Red, Green and Blue pixels of an image to overcome the
effects of different conditions of light. A number of color
transformations are employed to approximate the
characteristics such as conversion to Hue, Saturation and
Value.
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 14, No. 8, August 2016
1061
https://sites.google.com/site/ijcsis/
ISSN 1947-5500
Fig.2. Algorithm for Jaundice Prediction
This conversion can be carried out with the following
equations taken from previous studies:
=
255 (2)
=
255 3
=
255 (4)
= max
,
,
(5)
=min
,
,
(6)
= (7)
HSV i.e. (Hue, Saturation and value) calculation can be
calculated as below.
Hue Calculation:
= 0°,= 0 (8)
=60°×
6, =
9
=60°×
6, =
10
=60°×
6, =
11
Saturation Calculation:
= 0, = 0 12
=
, 0 (13)
Value Calculation:
= (14)
Image Segmentation is used to segment a small portion of
image in order to decrease the search area so the algorithm
ignores the pixels outside the segmented area. This segmented
part is then compared with the standard bilirubin level color
chart. The algorithm applies threshold to find the best possible
match of segmented image from the color chart.
Feature Extraction includes colormap transformations and
feature calculation. In order to detect the discoloration in a
better way, algorithm transforms the original RGB values into
Lab color spaces. We calculate the mean value for each color
channel thus giving 6 features. These features will be used to
determine the bilirubin levels.
To avoid the usage of costly Transcutaneous
Bilirubinometers, first we have performed the non-invasive
testing by using serum bilirubin coloration on strips instead of
TcBs. A few drops of each bilirubin sample are put on the
detection strips and the images of these colored strips are taken
from smartphone camera. Figure shows the image of biliubin
colored strips.
Fig.3. Serum Bilirubin Coloration on Strips
Since each bilirubin sample gives a different color on
detection strip thus a patch of each colored strip is segmented
and pasted in a transparent block of color chart which is formed
with various skin shades as shown in figure. The threshold is
Fig.4. Color Chart for various skin shades
then applied to find the closest match in color chart. In this
way, we have formed a standard set of colors corresponding to
each bilirubin sample. This provides us alternative results for
TcB estimates.
For jaundice prediction from skin, an image of forehead or
sternum of baby suffering from jaundice is taken since these
parts show visible changes in early stages of jaundice. The
image is segmented to reduce the search area and segmented
Capture Image from Smart Phone
Color Balancing
Image Segmentation
Feature Extraction
Bilirubin Estimation
Jaundice Prediction
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 14, No. 8, August 2016
1062
https://sites.google.com/site/ijcsis/
ISSN 1947-5500
portion is pasted onto the color chart for comparison. Again,
the threshold is applied to find the best possible match from
standard set of colors in color chart. The matched color in color
chart determines the level of bilirubin. In this way, we have
formed a non-invasive detection technique for early jaundice
prediction based on yellow discoloration of skin.
C. Smartphone Based Standalone Jaundice Level Detection
App
We have also proposed a standalone smartphone based
jaundice level detection app that jaundice patients can use to
determine biliruibin levels. Since bilirubin levels peak well
after most infants are discharged from hospital thus a low cost
screening system is essential for monitoring newborn jaundice
at home when clinical technologies are unavailable. Modern
smartphones are increasingly being programmed to diagnose
certain medical conditions. In addition to the features like
calling and texting, mobile phones also have high resolution
built-in cameras.
Fig.5. Flow Chart for Jaundice Detection Android App
Today’s smartphones are equipped with software
development kit that allows the device to be programmed with
algorithms in high level languages such as Java and C to
diagnose medical problems. Mobile diagnosis apps follow a
programming paradigm involving three steps procedure:
Gather raw images data
Process the data into medical information and
Display results
The steps involved in processing of image in an android
app are illustrated in flow chart shown in figure. Initially
pictures of patient’s skin are captured from the smartphone
camera and analysed by applying a series of image processing
steps. The camera image is first decompressed and then
segmented to detect the yellow discoloration of skin. Then
color segmentation is performed to demarcate yellow region of
skin. Depemding upon the amount of yellow discoloration, the
amont of bilirubin and jaundice level is displayed. Hence, it
provides an easiest way to predict jaundice with just one click
on your smartphone. Since the proposed app is standalone, it
does not require even internet connection thus making it much
simple to use for persons with low or incomplete resources.
With this method, early jaundice prediction has become
possible with least possible hardware and cost.
IV. RESULTS AND ANALYSIS
The developed technique, non-invasive jaundice detection
based on yellow discoloration of skin is evaluated using
MATLAB software. From the image of strips, RGB values are
obtained for different concentrations of bilirubin and a standard
is obtained to determine the amount of yellow discoloration of
skin. Table2. shows a standard chart for predicting bilirubin
levels from the image of skin.
TABLE II
PREDICTING BILIRUBIN LEVELS FROM SKIN USING TSB RESULTS
FOR DIFFERENT CONCENTRATIONS.
Bilirubin
Concentration
R Pixel
Values
G Pixel
Values
B Pixel
Values
TSB
Measurements
3 mg/dl
200
140
110
0.001
6mg/dl
200
150
100
0.006
9mg/dl
165
125
90
0.016
12mg/dl
145
110
80
0.020
15mg/dl
140
100
75
0.021
21mg/dl
125
85
60
0.023
24mg/dl
120
70
55
0.025
The developed technique gives a 0.93 rank order
correlation and a graph is plotted between images of different
biilirubin concentrations and pixel values of B which is the
Fig.6. Pixel Values from Skin and Strips to determine Bilirubin
20
40
60
80
100
6
9
12
15
21
24
Pixel values of B from image
Samples of various Bilirubin Concentrations
(mg/dl)
Strips
Skin
Take an Image
Smart Phone
Camera
Smart Phone
Gallery
View Image
Image Segmentation
Color Segmentation
Display Bilirubin Amount
Display Jaundice Level
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 14, No. 8, August 2016
1063
https://sites.google.com/site/ijcsis/
ISSN 1947-5500
main variant among different shades of yellow color to predict
the bilirubin levels for both skin and strips.
These results are then utilized to develop an android app for
jaundice detection. Fig.7. shows the screenshots of our android
application developed for jaundice level detection. For each
sample, image is captured from camera and is displayed to
user. The user then selects the desired area of forehead or
sternum to check the amount of bilirubin level. The selected
part of image is then compared with the available database and
the amount of bilirubin and hence the stage of jaundice is
displayed to user.
V. CONCLUSION AND FUTURE SCOPE
In this paper, a non-invasive bilirubin detection technique for
jaundice prediction based on yellow discoloration of skin is
developed. The technique works by using a smartphone to
click the image of skin and then predicting the bilirubin level
in newborn babies. It is painless, bloodless and the cheapest
non-invasive method of screening the neonates at home after
the discharge from hospitals. The developed method provides
a 0.93 rank order correlation and is able to detect bilirubin
concentration upto 24mg/dl to prevent hyperbilirubinemia.
Although the developed method is not able to replace TSB
detection but it can be used initially to detect jaundice upto
24mg/dl. To make this technique easily accessible for a billion
of people worldwide, especially economically disadvantaged
people, smartphone based android app has been developed. In
future, the focus should be on further data collection to
increase the diversity and reduce the limitations. Also, the
image quality can be further improved for better results.
(a). Jaundice detected with bilirubin concentration 15mg/dl.
(b). Jaundice detected with bilirubin concentration 24mg/dl
Fig.7. Jaundice Detection through Android Application
REFERENCES
[1] Giovanna Bertini, “Non-Invasive Bilirubinometery in Neonatal
Jaundice”, Semin Neonatol, volume 7, 2002, pp. 129-133.
[2] Gagan Mahajan “Transcutaneous Bilirubinometer in Assessment
of Neonatal Jaundice in Northern India”. INDIAN
PEDIATRICS. Vol. 42, 2005, pp. 41- 45.
[3] Yu-Hsun Chang, Wu-Shiun Hsieh, Hung-Chieh Chou, Chien-Yi
Chen, Jing-Yi Wu, Po-Nien Tsao “The Effectiveness of a
Noninvasive Transcutaneous Bilirubin Meter in Reducing the
Need for Blood Sampling in Taiwanese Neonates,”clinical
Neonatalogy.,vol.13, no. 2, 2006, pp. 60-63.
[4] M Kaplan, P Merlob, R. Regev, “Israel Guidelines for the
Management of Neonatal Hyperbilirubinemia and Prevention of
Kernicterus”, Journal of Perinatology, 2008, pp. 389-397.
[5] Samar N.El-Beshbishi, Karen E. Sharttuck, Amin A. Mohammad
and John R.Petersen,“Hyperbilirubinemia and Transcutaneous
Bilirubinometry,” clinical chemistry, vol.55, no.7, 2009, pp.
1280-1287.
[6] Suresh K. Alla, Adam Huddle, Joshua D. Butler, Peggy S.
Bowman, Joseph F. Clark, and Fred R. Beyette Jr., “Point-of-
Care Device for Quantification of Bilirubin in Skin Tissue”,
IEEE Transactions on Biomedical Engineering, Vol. 58, No. 3,
2011
[7] DM Campbell, KC Danayan, V McGovern, S Cheema, B Stade,
M Sgro, “Transcutaneous Bilirubin Measurement at the Time of
Hospital Discharge in a Multiethnic Newborn Population”,
Paediatr Child Health, Volume 16, Number 3, 2011, pp. 141-
145.
[8] Munevver Kaynak-Turkmen, S. Ayvaj Aydogdu, Cengiz
Gokbulut, Cigdem Yenisey, Omer Soz, Bilin Cetinkaya-
Cakmak, “Transcutaneous Measurement of Bilirubin in Turkish
Newborns: Comparison with Total Serum Bilirubin”, The
Turkish Journal of Pediatrics, Volume 53, 2011, pp. 67-74.
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 14, No. 8, August 2016
1064
https://sites.google.com/site/ijcsis/
ISSN 1947-5500
[9] Emmanuel Agu, Peder Pedersen, Diane Strong, Bengisu Tulu,
Qian He, Lei Wang, Yejin Li, “The Smartphone as a Medical
Device” IEEE International Workshop of Internet-of-Things
Networking and Control, 2013, pp. 49-52.
[10] M. Penhaker, V.Kasik, B. Hrvolova, “Advanced Bilirubin
Measurement by a Photometric Method” Elektronika Ir
Elektrotechnika, Vol.19, 2014, pp. 47-50.
[11] Nurashlida Ali “A Review of Non –Invasive Jaundice detection
usingnOptical Technique in Neonates”.Conference in Advances
In Computing, Electronics and Electrical Technology - CEET
2014. , 2014,pp.1-3.
[12] Lilian de Greef, Mayank Goel, Min Joon Seo, Eric C. Larson,
James W. Stout, James A. Taylor, Shwetak N. Patel, “BiliCam :
Using Mobile Phones to Monitor Newborn Jaundice”, ACM,
2014, pp.1-12.
International Journal of Computer Science and Information Security (IJCSIS),
Vol. 14, No. 8, August 2016
1065
https://sites.google.com/site/ijcsis/
ISSN 1947-5500
