See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/312678134

Hypervolemia screening in predialysis healthcare for hemodialysis patients

using fuzzy color reason analysis

Article  in  International Journal of Distributed Sensor Networks · January 2017

DOI: 10.1177/1550147716685090

CITATIONS READS

5 3,783

5 authors, including:

Wei-Ling Chen Chung-Dann Kan

National Cheng Kung University National Cheng Kung University
73 PUBLICATIONS   437 CITATIONS    166 PUBLICATIONS   1,654 CITATIONS   

SEE PROFILE SEE PROFILE

Chia-Hung Lin
National Chin-Yi University of Technology
252 PUBLICATIONS   4,424 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Phonographic signal with a fractional-order chaotic system: A novel and simple algorithm for analyzing residual arteriovenous access stenosis View project

Assessment of flow instabilities in in-vitro stenotic arteriovenous grafts using an equivalent astable multivibrator View project

All content following this page was uploaded by Wei-Ling Chen on 03 August 2017.

The user has requested enhancement of the downloaded file.

Research Article

International Journal of Distributed

Sensor Networks
2017, Vol. 13(1)
Hypervolemia screening in predialysis Ó The Author(s) 2017
DOI: 10.1177/1550147716685090
healthcare for hemodialysis patients journals.sagepub.com/home/ijdsn

using fuzzy color reason analysis

Wei-Ling Chen1, Chung-Dann Kan2, Chia-Hung Lin3, Ying-Shin Chen3 and

Yi-Chen Mai4

Abstract
Maintaining adequate dry weight and fluid volume balance is an important issue for dialysis patients. Malnutrition and
sodium intake are the primary factors that cause fluid volume imbalance and changes in body weights. Inadequate dry
weight control results in higher levels of blood pressures and is related to various complications, such as volume over-
load, hypertension, congestive symptoms, and cardiovascular diseases. Moreover, inadequate fluid removal provokes
hypotension during dialysis treatment. Thus, we propose an early warning tool based on fuzzy color reason analysis in
predialysis healthcare for hypervolemia screening. The anthropometric method is a rapid, non-invasive, and simple tech-
nique for estimating the total body water. In this study, Watson standard formula is employed to estimate cross-sectional
standard of total body water with the patient characteristics, including gender, age, height, and weight. In contrast to the
experienced anthropometric formulas, Watson formula has less than 2% of margin errors and provides a criterion as a
reference manner to estimate the total body water in patient’s normal dry weight. In addition, inadequate dry weight
and total body water controls will lead to higher blood pressures. The systolic blood pressure is also an indicator to
evaluate pre-hypertension of 120–139 mmHg and hypertension of greater than or equal to 140 mmHg. Therefore, the
levels of two indicators, total body water and systolic blood pressure, are parameterized with fuzzy membership grades
to describe the normal and the specific ranges of undervolemia and hypervolemia. A color reason analysis utilizes a hue–
saturation–value color model to design a color perceptual manner for separating normal condition from hypervolemia
or undervolemia. Normalized hue angle and saturation value provide a promising visual representation with color codes
to realize the patients’ diagnosis. Dialysis patients with hypertension demonstrated that the proposed model can be used
in clinical applications. In addition, a healthcare chair is carried out to measure blood pressure and weight in predialysis.
The proposed assistant tool integrates an electronic pressure monitor and an electronic weight monitor, and fuzzy color
reason analysis is also intended to be established in an intelligent vehicle via a WiFi wireless local area network for cloud
computing.

1
Department of Engineering and Maintenance, Kaohsiung Veterans General Hospital, Kaohsiung City, Taiwan
2
Department of Surgery, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, Tainan City, Taiwan
3
Department of Electrical Engineering, Kao Yuan University, Kaohsiung City, Taiwan
4
Department of Aeronautics and Astronautics, National Cheng Kung University, Tainan City, Taiwan

Corresponding authors:
Chung-Dann Kan, Department of Surgery, National Cheng Kung University Hospital, College of Medicine, National Cheng Kung University, Tainan City
70101, Taiwan.
Email: kcd56@mail.ncku.edu.tw

Chia-Hung Lin, Department of Electrical Engineering, Kao Yuan University, Kaohsiung City 82151, Taiwan.
Email: eechl53@gmail.com

Creative Commons CC-BY: This article is distributed under the terms of the Creative Commons Attribution 3.0 License
(http://www.creativecommons.org/licenses/by/3.0/) which permits any use, reproduction and distribution of the work without
further permission provided the original work is attributed as specified on the SAGE and Open Access pages (http://www.uk.sagepub.com/aboutus/
openaccess.htm).
2 International Journal of Distributed Sensor Networks
Keywords
Fuzzy color reason analysis, hypervolemia, total body water, systolic blood pressure, hue–saturation–value, healthcare
chair
Date received: 6 August 2016; accepted: 28 November 2016
Academic Editor: Shinsuke Hara
Introduction
The state of body water is an important factor in rou-
tine hemodialysis (HD) treatment and postdialysis
healthcare. The so-called hypervolemia/fluid overload
is a medical condition manifested as excess body water
in the blood, leading to increase in body sodium con-
tent and a consequent increase in extracellular body
water. In dialysis patients, this condition causes certain
complications, such as hypertension,1–3 increase in body
weight (BW), and peripheral edema in the legs and
arms. Sodium plays an important role in regulating the
total amount of water in the human body. Extremely
high levels of sodium cause congestive heart failure and
cardiovascular system instability during dialysis treat-
ment. Inadequate dry weight results in a higher level of
blood pressures, while prolonged symptoms lead to
restructuring of the heart function and induction of
heart failure and arrhythmias.4,5 Therefore, maintaining
dry weight, adequate water volume control, blood pres-
sure control, and reducing sodium concentration is
important in postdialysis healthcare for dialysis
patients. This study proposes a strategy to prevent
hypervolemia and maintain envolemia/dry weight.
In clinical examinations, hypertension for adult HD
patients is defined as 1-week predialysis systolic blood
pressure (SBP) of greater than 150 mmHg and diastolic
blood pressure of greater than 85 mmHg and requires
antihypertensive drugs for blood pressure control.6,7
The echocardiographic technique has been used to
observe the congestive symptoms of dilatation of vari-
ous compartments in the left atria (LA) and hypertro-
phy of left ventricle (LV).3,8,9 Color Doppler
echocardiography with B-/M-mode operations can
reveal the LA/LV systolic and diastolic functions,
including mitral valve injection velocity and pulmonary
venous flow, left/right ventricle tissue, LV mass, and
LA volume.9,10 Certain parameters are measured to
identify the heart function, such as LA diameter, LV
end-diastolic and end-systolic diameters, posterior wall
thickness, and interventricular septum. Although this
technique is a promising solution in routine examina-
tions, it is difficult to perform in the healthcare or
home environment of HD patients.
In addition, portable measurement techniques,
such as relative plasma volume monitors3,11 and
bioimpedance spectroscopy,1,9 have been used to assess
the response to hypervolemia and to further maintain
the dry weight. Relative plasma volume monitoring is a
photo-optical method to non-invasively measure abso-
lute hematocrit in the arterial end of the dialyzer, while
the percentage of blood volume change (plasma refill
rate vs removal fluid) during dialysis is being computed.
High refill rate and a flat slope indicate hypervolemia,
whereas a lower plasma refill rate and a steeper slope
indicate lower LV mass, better ventricular function,
and intradialytic hypotension. Hence, less antihyperten-
sive drug requirements can be achieved. Multifrequency
bioimpedance spectroscopy12 is also a non-invasive
technique using a pair of electrodes placed on the wrist
and ankle. The slopes of normovolemia and hypervole-
mia characterize the variation in extracellular water
(ECW) with BWs. However, ECW and BW measure-
ments need to be acquired at the beginning of each
treatment, which are used to determine the slopes of
normovolemia. The aforementioned clinical examina-
tions are simple methods but lack the sensitivity and
specificity to identify hypervolemia.
The anthropometric method is also a simple tech-
nique to determine the total body water (TBW). It can
be easily implemented using a mathematical model with
dialysis indices, such as height (H, cm), weight (W, kg),
body mass index (BMI), age (A, years), sex (S, male/
female), and diabetes (D). Experienced formulas2,12–15
can be obtained for estimating TBW using big data
(experimental data) processes, such as the Watson for-
mula, Hume formula, Sahlgrenska formula, and
Chumlea formula. Among these formulas, the Watson
formula has been widely used to estimate TBW in dia-
lysis patients. It is routinely used to estimate the TBW
for evaluating body composition in different popula-
tions, including gender, ages, heights, and weights. Its
mathematical model with a linear regression coefficient
for weights systematically estimates the body water,
while providing narrow margins of substantial error
(approximate average or close-to-average weights) and
a promising reference method. Watson formula has less
than 2% of margin errors13 to estimate the TBW in
patient’s normal dry weight, as normality TBW is a ref-
erence for sustained examinations. Therefore, the ratio
of estimated TBW and normality TBW provides a cri-
terion to identify the changes in fluid volume balance.
Chen et al.
In addition, blood pressure measurement is an impor-
tant factor for all aspects of cardiovascular diseases
and diabetes in dialysis patients. Predialysis and post-
dialysis blood pressure should be ‘‘less than 140 or
90 mmHg’’ and ‘‘less than 130 or 80 mmHg,’’ respec-
tively. Among patients undergoing HD treatment, SBP
variability is a strong predictor than diastolic blood
pressure of age, cardiovascular diseases, and diabetes.
Hypertension is also dependent on BW and fluid vol-
ume imbalance. Acute weight gain and weight loss
result in elevated blood pressure and orthostatic hypo-
tension, respectively.
In this study, two key indicators, TBW and SBP, are
parameterized using fuzzy membership functions (MFs)
to describe the specific ranges with membership grades.
Based on the anthropometric method, the Watson for-
mulas for male/female are used to estimate the TBW
with the patient characteristics, ages, heights, and
weights. The definition of TBW ratio is used to indicate
the TBW changes in the patient’s average normality
level, with ratios greater than 1 or less than 1 represent-
ing the fluid volume imbalance. Indicator SBP is
employed to screening the levels of blood pressure con-
ditions, while the range of 120–139 mmHg for pre-
hypertension and greater than or equal to 140 mmHg
or higher for hypertension in adult patients. Then, fuzzy
color reason analysis (CRA) is employed to automati-
cally separate the normal condition from prehypervole-
mia/hypervolemia and undervolemia. The proposed
screening model is designed as a classifier to deal with
fuzzy inference problems and multi-criteria decision-
making with hue–saturation–value (HSV) color
model.16–19 Fuzzification operations can map TBW and
SBP indications into the specific membership grades
using Gaussian and sigmoidal MFs. CRA is employed
to map membership grades into rule-weighted outputs
for decision-making. A fuzzy CRA utilizes hue angle
and saturation value to identify the three levels with
describing perceptual color relationships for normal,
prehypervolemia, and hypervolemia screening. In con-
trast to the other artificial intelligent models, such as
artificial neural networks, these methods can also be
designed as a decision-making manner for classification
applications.18 Optimization methods, such as least-
square algorithm, gradient descent algorithm, and
swarm-intelligence algorithm, need to adjust the model
parameters to enhance the estimation accuracy. Large
number of training data also affects the training process
and classification efficiency. However, determination of
the multi-layer network’s structure, large number of
experimental data, and the update of parameters with
iterative computations are three concerns that limit the
mechanism’s inclusion in a portable embedded system.
The proposed fuzzy CRA application software can be
easily implemented in a tablet PC or in smart mobile
devices and be integrated with the electronic pressure
3
and weight monitors in a healthcare chair. Sensing units
with a WiFi wireless local area network (WLAN) mode
(IEEE 802.11 Standards) can transmit physiological
parameters to a mobile device.20–25 Hence, smart
mobile devices can receive the reliability and persona-
lized physiological indications in indoor environment
or HD room.
Methodology
Anthropometric formulas: TBW estimation
TBW estimation is important information in an indi-
vidual dialysis patient for evaluating the degree of fat-
ness. In clinical measurements, a healthcare chair-based
body weightometer and a blood pressure monitor are
used to take physiological measurements. The weight is
relative to one’s height and is therefore used to deter-
mine the BMI, TBW in predialysis stage, and interdia-
lytic weight gain in postdialysis stage. Blood pressure is
an important factor that might affect physiological
functions, and maintaining a higher SBP of less than
150 mmHg to control hypertension and predialysis
blood pressure range of 140–90 mmHg have been sug-
gested for adults. The objective of this study was to rap-
idly estimate the accuracy of TBW in predialysis stage.
A chair or a wheelchair equipped with various commer-
cial sensors20–22 and wireless communication23–25 has
been integrated into a reliable sensing system in clinical
applications, remote physiological signal monitoring,
and home healthcare. Hence, critical signals, such as
electrocardiography signal, photoplethysmogram sig-
nal, and blood pressure, can provide health information
in health management or evaluation of treatment effi-
cacy. In this study, we intended to design an intelligent
healthcare chair to take personalized physiological
measurements in predialysis, including BW and blood
pressure monitors, as seen in Figure 1. The sensing unit
with a WiFi WLAN mode (IEEE 802.11 Standards25)
could transmit the data to a mobile device in the indoor
range of less than 30 m. The received signal strength
indicators were about greater than 270 dBm,20 which
envisaged deployment conditions in the indoor environ-
ment (HD room: 20 m 3 30 m). Hence, the electronic
sensing units in its communication applications provide
the reliable short wireless transmission distance.
Given the patients’ characteristics: S (male = 1/
female = 0), A (years), H (cm), and W (kg), the two
formats of Watson formula in adult subjects were used
to calculate TBW:2,12
TBW in males (S = 1)
TBWmale = 2:447  (0:09156 3 A)
ð1Þ
+ (0:1074 3 H) + (0:3362 3 W )
4 International Journal of Distributed Sensor Networks

Figure 1. An intelligent healthcare chair for physiological measurements in predialysis healthcare.

TBW in females (S = 0)

TBWfemale =  2:097 + (0:1069 3 H) + (0:3362 3 W ) ð2Þ

The other anthropometric formulas for estimating

TBW are shown in Remark.12,15

Remark. Experienced anthropometric formulas

 Hume formulas

TBW in males (S = 1)
Figure 2. Total body water (L) versus body weight (Kg) for
TBWmale =  14:012934 + (0:194786 3 H) males/females using Watson formula.
ð3Þ
+ (0:2962934 3 W )
furosemide use and a negative correlation with age.
TBW in females (S = 0) BMI was also negatively correlated with age, while it
was positively correlated with furosemide use.12 TBW
TBWfemale =  35:270121 + (0:34454 3 H) had an independent positive association with height
ð4Þ and weight for female gender. In addition, patients hav-
+ (0:183809 3 W )
ing higher BW were more likely associated with ambu-
 latory blood pressure on a daily basis (24 h).
Sahlgrenska formulas
Hypertension is dependent on fluid volume imbalance.
For adult patients, SBPs gave the screening level for
TBW in males (S = 1)
evaluating pre-hypertension or hypertension, while the
TBWmale =  28:3497 + (0:243057 3 H) range was 120–139 mmHg and the highest value was
ð5Þ greater than or equal to 140 mmHg or higher, respec-
+ (0:366248 3 W ) tively. Hence, both TBW and SBPs could be indicators
TBW in females (S = 0) to evaluate the body fluid composition dependent on
hypertension or higher blood pressures.
TBWfemale =  26:6224 + (0:262513 3 H)
ð6Þ
+ (0:232948 3 W ) Fuzzy CRA
These criteria provided a reliable standard as a refer- According to Figure 3(a) and (b), two indicators, TBW
ence manner to estimate the body fluid composition in and SBP, were parameterized with Gaussian, Z sigmoi-
normal dry weight, as shown by the TBW (L) versus dal, and S sigmoidal MFs, varying between values 0
BW (kg) in Figure 2. It appeared that TBW had a posi- and 1, as shown in Figure 3. The ratio, TBW/TBWnor,
tive correlation with male gender, weight, and was used to indicate the changes in fluid volume
Chen et al. 5

(a)
1.2
μ1 μ2 μ3
Membership Grade

0.8 mean 1 mean 2 mean 3

0.6

0.4

0.2

0
0.9 0.95 1 1.05 1.1

Ratio, TBW / TBW nor

(b)
(c)
1.2 Prehypertension Hypertension 5
μ4 μ5 ξ = 5.0
Membership Grade

1
4 ξ = 4.0

Gray Grade
0.8
120 140
ξ = 3.0
3
mmHg mmHg
0.6 ξ = 2.0
2 ξ = 1.0
0.4
1
0.2

0 0
60 80 100 120 140 160 180 200 0 0.2 0.4 0.6 0.8 1

Blood Pressure (mmHg) M embership Grade

Figure 3. Membership functions: (a) MFs for parameterizing TBW ratios, (b) MFs for parameterizing blood pressures, and (c) gray
grades versus membership grades.

balance. The parameter, TBWnor, is estimated by where standard deviations, si = Meani 3 0:05, i = 1,
Watson formula, which is a reference for sustained 2, 3; TBWnor is the subject’s ‘‘normality TBW’’ as a ref-
examinations in normal dry weight. The nephrologists erence for the patient herself/himself. Three MFs for
prescribe the normal dry weight for each patient with representing the ranges of TBW ratios are shown in
no extra fluid and a normal blood pressure. Hence, the Figure 3(a).
TBW ratio, TBWrat, was parameterized in the specific As shown in Figure 3(b), the blood pressures
ranges, including undervolemia, m1, normal, m2, and weighted between 0 and 1 as MFs for representing the
hypervolemia, m3, as follows normal and hypertension ranges as follows
8  !
TBW < \1, undervolemia 1 SBP  120 2
TBWrat = , TBWrat = ’ 1, normal ð7Þ m4 = exp  3 , 0\SBP\120
TBWnor : 2 20
.1, hypervolemia
8   2  ð11Þ
< ( 
m1 =
exp  12 3 TBWratsMean 1
, TBWrat .Mean1  2 
:
1
exp  12 3 SBP140 , SBP\140
1, 0:9\TBWrat  Mean1 m5 = 20 ð12Þ
1, 140  SBP\200
ð8Þ
  ! The membership grades of TBW are weighted by m4
1 TBWrat  Mean2 2 and m5 as follows
m2 = exp  3 ,
2 s2 ð9Þ
½m01 , m02 , m03  = ½m1 3 m4 , m2 3 m4 , m3 3 m5  ð13Þ
0:9\TBWrat \1:1
8   2  Then, the weighted values are converted to gray
< grade, ri, i = 1, 2, 3, by nonlinear transformation, as
exp  12 3 TBWratsMean3
, TBWrat \Mean3
m3 = 3
described earlier18,24
:
1, Mean3  TBWrat \1:1
ð10Þ ri = jexp½  j 3 (1  u0i )2  ð14Þ
6 International Journal of Distributed Sensor Networks

Figure 4. HSV color space and RGB color transformation.

 
where parameter, j, is the recognition coefficient to rg
H2 = 240 + 60 3 , rmax = b ð18Þ
adjust gray-grade intensity with interval (0, N), which Dr
is used to enhance contrast (j = 5.0 in this study). The 8  
>
> rb
gray-grade transformation is a Gaussian function, as < 60 3 , rmax = r, g  b
shown in Figure 3(c). Intensity adjustment is used to Dr  
H3 = ð19Þ
>
> rb
enhance contrast and provides better contrast to sepa- : 300 + 60 3 , rmax = r, g\b
rate testing data in two classes. The minimum and max- Dr
imum gray grades can be determined as follows Index, Hj (Hj = 1, 2, and 3), that refers to the three
primary colors is used to identify the TBW condition,
rmin = min½r1 , r2 , r3 ,
while the green series color (60°–180°) is used for under-
rmax = max½r1 , r2 , r3 , ð15Þ volemia, blue series color (180°–270°) for normal condi-
Dr = rmax  rmin tion, red series color (270°–300°) for prehypervolemia,
and (0°–60° or 300°–360°) is used for hypervolemia, as
where rmin 6¼ rmax and rmax 6¼ 0. The primary color shown in Figure 4. In addition, the saturation, S, and
grades and parameter, Dr, have capability of self- value, V, are described as pure color saturation and
regulation by the membership grades. According to the lightness, respectively, which are defined as follows19
HSV color model,19 the CRA-based classifier18,19 is
(
defined mathematically by transformations between the 0, rmax = 0
RGB color space and the HSV color space. Gray grades V = rmax , S= rmin ð20Þ
1 , rmax 6¼ 0
are converted to three primary color grades, r (red), g V
(green), and b (blue) as follows
The value of index, H, is generally normalized to lie
between 0° and 360°, while ‘‘rmin = rmax’’ and
g = r1 , b = r2 , r = r3 ð16Þ
‘‘S = 0’’ have no geometric meaning. These can help
where r, g, and b2 [0, j] are the r, g, and b coordinates the nurses or HD patients to easily visualize the screen-
in the RGB color space, respectively. Primary color ing results. For hypervolemia screening, index H is con-
grades are employed to separate the normal condition ducted to identify the three levels as angle points in
from hypervolemia or undervolemia, as seen in Figure 4. Subsequently, index, S, of 0.5–1.0 provides
Figure 4. For an HSV color space, the hue angle, Hue high confidence for confirming the possible level.
2[0°, 360°], is described by numerous specific coordi-
nates of the corresponding color around the wheel and Implementation of proposed screening model
is determined as follows
The proposed fuzzy CRA-based screening model is

gb shown in Figure 5(a), consisting of fuzzification via
H1 = 120  60 3 , rmax = g ð17Þ gray-grade transformation, maximum–minmum opera-
Dr
tion, and transformation from RGB color space to
Chen et al.
Figure 5. The proposed fuzzy CRA screening model: (a) structure of fuzzy CRA based and (b) implementation of screening model
in an embedded system.
HSV space. The value of hue angle, Hj, was normalized
and was utilized to identify the four levels as index HC
 decision confidences as inequality, S  0.50. In addi-
Hj
HC = , HC 2 ½0, 1 ð21Þ
3608
where the critical thresholds, HC = [60°, 180°, 270°,
300°, 360°]/360° = [1/6, 1/2, 3/4, 5/6, 1], were used to
separate the ‘‘normal condition’’ from ‘‘hypervolemia’’
and ‘‘undervolemia.’’ The concept of the fuzzy CRA
reason was derived from the fuzzification operations
and HSV color model to describe perceptual color rela-
tionships for hypervolemia screening. Fuzzification
operations can map mathematical input variables into
specific gray grades using Gaussian, Z sigmoidal, and
S sigmoidal MFs. The MF parameters were static or
could be changed dynamically based on different HD
patients. Subsequently, HSV color model is employed
to map gray grades into perceptual colors for
7
decision-making. The saturation value, S, varied from
0.0 (unsaturated) to 1.0 (fully saturated) to enhance the
tion, index, HC, was used to identify the possible level
and the choice of recognition coefficient, j .. 1;26
gray grades could be weighted to distinctly separate
into different levels, and the decision-making might
increasingly become distinguishing for classification of
space to enhance the screening accuracy.
The proposed screening model could be implemen-
ted in an embedded system, as shown in Figure 5(b).
Its model uses straightforward mathematical computa-
tions to achieve the inference procedures for real-time
applications. As seen in Figure 5(b), an embedded sys-
tem (National InstrumentsTM myRIO-1900, Austin,
TX, USA) can be applied to establish a prototype
screening algorithm within a short design cycle. We
also integrated electronic body weightometer and blood
pressure monitor for weight and blood pressure
8 International Journal of Distributed Sensor Networks
Figure 6. The graphical programming user interface.
measurements using a healthcare vehicle/chair. The
vehicle was equipped with commercial sensors and was
designed to carry out wireless communication for
unconstrained physiological monitoring. A WiFi
(IEEE 802.11 Standard25) WLAN was used for linking
mobile devices (smart phones, personal digital assis-
tant, or iPad) and portable computers provide support
(laptop), while being operated on a 2.4-GHz industrial,
science, and medical frequency bands. It was designed
for portable devices with low power consumption and
short-range communications in common household
and mobile appliances.
The screening model in an embedded system has pre-
dialysis healthcare functions, including (1) connectivity
measurements from the weight and blood pressure sen-
sing units, (2) built-in WiFi communication to acquire
physiological data for remote monitors, (3) easy imple-
mentation of fuzzy CRA-based decision-making algo-
rithm in a microprocessor (dual-core ARM Cortex),
and (4) remote data acquisition via wired or wireless
communication and transmitting data to a tablet PC or
a mobile vehicle. The proposed framework can be eas-
ily implemented in a portable medical device using
LabVIEW graphical programming software (NITM,
Austin, TX, USA) and the embedded design device in
real-time applications. The proposed prototype model
was designed in a graphical programming user interface
and conducted in a development platform, as shown in
Figure 6. In addition, the remote monitor’s function
can be used in telecare applications in public HD room
and home-based HD for hypervolemia screening.
Experimental results
The proposed hypervolemia screening model has been
developed in an embedded system and a WiFi wireless
connection. The vehicle has integrated commercial sen-
sors to measure SBP and BW. The fuzzy CRA-based
algorithm was conducted in LabVIEW at the PC level.
Data transmission units are both wired and wireless
communications. The measurement data and informa-
tion can be obtained and transmitted using remote data
acquisition unit with WiFi WLAN that communicates
with a tablet PC. For an indoor wireless communica-
tion, its framework also created a data dashboard to
remotely monitor and to represent HD patients’ per-
sonal parameters, including age, male/female, BW,
height, and SBP, and screening results. The perfor-
mance of the proposed methods was tested for screen-
ing accuracies on the data set, including normal
condition, undervolemia, and hypervolemia groups,
with age ranging from 37 to 67 years for the study as
detailed below.
Preliminary screening assessment
For a male, A = 50 years, W = 52 kg, H = 160 cm,
and BMI = 20.31 kg/m2, current W and SBP were
required for measurement in predialysis stage. Hence,
the vehicle provided a personalized access to measure
health information and compare with subject’s control
data. The proposed model used straightforward mathe-
matical operations to assess the numerical computa-
tions, while it required less parameter assignment, such
as mean values and standard deviations for MFs and
recognition coefficient, j = 5. The mean values were
averaged from personal baseline data, such as changes
in weight, 6 2.4 kg, and changes in predialysis-seated
SBP, and 6 16 mmHg in hypertensive HD patients.11
The proposed model has capability of self-regulation of
primary color grades, which showed good adaptability
for separating normal condition from abnormal
Chen et al.
Figure 7. Preliminary testing results: (a) testing results for
hypervolemia screening and (b) testing results for undervolemia
screening.
condition. The Watson formula was employed to eval-
uate the TBW.
By increasing the BWs and blood pressures, the hue
angle, H, gradually increased from 180°, 270°, 300°, to
360°, while normalized hue angle, HC, increased from
1/2, 3/4, 5/6, to 1, as seen in Figure 7(a). It can be seen
that both higher BWs and blood pressures were sensi-
tive to the screening results. The CRA has a flexibility
visual manner with color codes, Hj 6 60, j = 1, 2, 3,
to realize the patients’ diagnosis. The saturation value,
S  0.5, provided the confidence for confirming the
possible level. While index, S, decreased to less than
0.50, the noted critical point was used to decide the level
of translation from normal condition to hypervolemia.
Therefore, the changes in hue angle remained signifi-
cant and dependent on gross changes in weights and
blood pressures. These parameters could be indicators
to evaluate the body fluid composition in prehyperten-
sive and hypertensive HD patients. In addition, by
decreasing the BWs and increasing the blood pressures,
the hue angles approximated to 120° (normalized hue
angle, HC = 1/3) as index, S  0.5. The undervolemia
can also be identified, as seen in Figure 7(b). The pro-
posed screening model has been validated and proved
to be a promising method to evaluate the possible levels
in clinical settings.
Case study in hypertensive HD patients
The experimental data from hypertensive HD patients
were used to verify the proposed screening model. Data
9
of 12 participating subjects, aged 37–67 years, were
shown in Tables 1 and 2. For these case studies,
Watson, Hume, and Sahlgrenska formulas were used
to calculate the normality TBWs and estimated TBWs.
The normal dry weight was prescribed for each patient
by the nephrologists. Then, the TBWrat was parameter-
ized to rapidly indicate the changes in fluid volume bal-
ance, as TBWrat . 1, TBWrat ’ 1, or TBWrat \ 1. In
fuzzification operations, the membership grades of
TBWrat and SBP were converted to gray grades and
three primary color grades using the gray-grade inten-
sity adjustment. The CRA utilized HSV color model to
map gray grades into color codes, green series color,
blue series color, and red series color, for the estimation
of patients’ diagnosis. For instance, Subject 7 (male,
aged 54.9 years) had hypertension and diabetes, a case
in which changes in dry weight of + 2.0 kg could
result in higher SBPs and fluid volume imbalance.
Inadequate control of fluid volume or inability to
maintain an appropriate dry weight for chronic HD
patients was identified as a key factor to cause excess
mortality. These problems could lead to chronic vol-
ume overload with hypertension and left ventricular
hypertrophy, while subsequently causing cardiovascu-
lar symptoms and also increasing the patency rate of
complications. In addition, some patients removed
extra weights to achieve appropriate dry weight, result-
ing in uncomfortable symptoms. Therefore, it is impor-
tant to screen early in predialysis stage (three times a
week) and then perform dialysis for volume control or
administer antihypertensive drugs. The proposed
screening model is detailed according to the following
procedure:
Step 1. Given the baseline W, measured W, height,
and SBP, TBW and TBWnor were calculated using
equation (1), and then TBW = 43.6233 and
TBWnor = 42.9509 can be obtained, as seen in
Table 1;
Step 2. Calculate the TBW ratio, TBWrat = 1.0157,
using equation (7), as seen in Table 2;
Step 3. Compute the membership grades, [m1, m2,
m3, m4, m5] = [0.7681, 0.9525, 0.8067, 0.1274,
1.0000] using equations (8)–(12);
Step 4. Convert the membership grades to gray
grades using equations (13) and (14), then r = [r1,
r2, r3] = [0.0854, 0.1053, 4.1482], r2 [0, 5];
Step 5. Find the minimum and maximum grades,
rmin = 0.0854 and rmax = 4.1482, rmin 6¼ rmax;
Step 6. Convert the gray grades to primary color
grades, g = 0.0839, b = 0.1043, and r = 4.2568;
Step 7. If rmax = r3, then find the average hue
angle, H3 = 359.7064°, and saturation, S = 0.9794,
using equations (19) and (20);
Step 8. Normalize the hue angle using equation (21),
then index HC = 0.9992 is employed to identify the
10 International Journal of Distributed Sensor Networks

(6) 2 (3)/(3)
level of hypervolemia and index S  0.5 denoting

+ 1.31
+ 1.00
+ 1.73
+ 1.52
+ 1.49
22.42
+ 1.70
21.32
20.38
20.67
21.83
21.61
high confidence, as seen in Table 2.
Error (%), (TBW 2 TBWnor)/TBWnor

Subject 7 satisfied with an index, HC = 0.9992

(5) 2 (2)/(2)

(359.7064°) for red series color, and S = 0.9794; thus,

this case study can be agreed as a patient with increas-
+ 1.07
+ 0.80
+ 1.34
+ 1.21
+ 1.21
21.87
+ 1.36
21.05
20.29
20.53
21.39
21.22
ing weight that led to ‘‘hypervolemia.’’ In addition, for
Subject 9 (male, aged 66.1 years), the inference results
indicated ‘‘normal condition,’’ satisfying with an index,
(4) 2 (1)/(1)

HC = 0.5369 (193.2840°) for blue series color.

However, the saturation index, S = 0.4485, was less
+ 1.17
+ 0.90
+ 1.54
+ 1.38
+ 1.39
22.19
+ 1.57
21.20
20.36
20.64
21.68
21.48
than 0.5, due to the subject having slight decrease in
weight and pre-hypertension. This observation led to a
suggestion of maintaining the appropriate BMI and dry
48.2635
47.8804
36.5811
41.5993
47.3141
35.4936
43.8392
41.2056
38.9077
37.7816
33.4841
33.4953

weight. In contrast to Subject 11 (male, aged

(6)

66.9 years), he had changes in weight of 21.7 kg, and

the inference results indicated ‘‘undervolemia,’’ satisfy-
47.6473
47.3435
38.1977
42.2601
46.8723
37.3184
44.0702
41.9432
40.0747
39.1617
35.6990
35.7102

ing with an index, HC = 0.2950 (106.2000°) for green

Estimated TBW

series color, and saturation index, S = 0.5303. For

(5)

Symbols (1) and (4) mean Watson formulas; symbols (2) and (5) mean Hume formulas; symbols (3) and (6) mean Sahlgrenska formulas.

color code in imaging version, the screening results pro-

vided a promising suggestion to gradually control fluid
49.2519
48.6543
37.7488
41.9679
46.3635
36.0093
43.6233
41.2786
37.6583
36.4410
33.3548
33.4897

volume and appropriate BMI for HD healthcares. This

(4)

finding confirmed that the proposed screening model

could detect fluid volume imbalance in its early stages
47.6409
47.4042
35.9585
40.9766
46.6183
36.3726
43.1067
41.7550
39.0542
38.0379
34.1067
34.0447

in HD patients, six with hypervolemia, two with normal

condition, and four with undervolemia. Experimental
(3)
Normality TBW, TBWnor

results for 12 HD subjects are shown in Tables 1 and 2.

47.1436
46.9584
37.6939
41.7564
46.3093
38.0295
43.4776
42.3877
40.1932
39.3691
36.2027
36.1547
(2)

Discussion
48.6804
48.2172
37.1773
41.3963
45.7248
36.8161
42.9509
41.7829
37.7928
36.6764
33.9263
33.9940

Cardiovascular disease is the most common cause of

mortality in HD patients, and hypertension is also a
(1)

significant factor for cardiovascular disease pathogen-

esis. Hypervolemia, hypematremia, and hyperkalemia
H (m)

1.73
1.75
1.72
1.71
1.77
1.72
1.72
1.69
1.70
1.76
1.69
1.68

are important risk factors for their morbidities and

mortalities. In previous studies,3,4,5,27,28 the relationship
between hypervolemia and malnutrition was regarded
Change in W, (kg)

to be the key indicator of predisposition to hyperten-

sion in HD patients, further preventing the progression
Table 1. Experimental results for TBW estimation.

of cardiovascular event rate. During dialysis treatment,

+ 1.7
+ 1.3
+ 1.7
+ 1.7
+ 1.9
22.4
+ 2.0
21.5
20.4
20.7
21.7
21.5

inadequate fluid removal and blood volume ultrafiltra-

tion control or fluid volume imbalance provoked hypo-
tension, associated with clinical symptoms for nursing
Baseline W (kg)

decreases SBP by greater than 20 mmHg or decreases

mean arterial pressure by 10 mmHg.29
Hence, anthropometric formulas, such as Watson,
92.6
94.2
61.2
75.9
87.3
62.5
81.0
78.6
67.8
64.3
58.6
59.2

Hume, Sahlgrenska, and Lee formulas,12–15 provide

well-known standard methods to estimate TBW. This
method might cause large systematic errors, while
A (year)

TBW: total body water.

TBW varied from the average, obese, and overhydrated

37.0
43.7
46.1
52.0
53.6
54.2
54.9
56.5
66.1
66.7
66.9
67.0

patients. In particular, in obese patients, the estimated

results led to large errors, in average or close-to-average
Patient no.

weights, and the Watson, Hume, and Sahlgrenska for-

mulas provided a promising reference with narrow mar-
gins of error.2 For the prescribed normal dry weight,
10
11
12
1
2
3
4
5
6
7
8
9
Table 2. Experimental results for hypertensive HD patients.

Patient no. A (year) Baseline W (kg) Change in W (kg) H (m) SBP (mmHg) TBWrat = TBW/TBWnor Hypertension Fuzzy CRA (HC/S)
Chen et al.

diabetes
(1) (2) (3) (1) (2) (3)

1 37.0 92.6 + 1.7 1.73 168.8 1.0117 1.0107 1.0131 O 0.9998 0.9998 0.9998
0.9855 0.9887 0.9873
2 43.7 94.2 + 1.3 1.75 177.2 1.0090 1.0082 1.0100 O 0.9999 1.0000 1.0000
0.9890 0.9855 0.9893
3 46.1 61.2 + 1.7 1.72 145.6 1.0153 1.0133 1.0173 O 0.9828 0.9817 0.9834
0.8629 0.8245 0.8766
4 52.0 75.9 + 1.7 1.71 154.6 1.0138 1.0120 1.0151 O 0.9973 0.9973 0.9973
0.9578 0.9546 0.9630
5 53.6 87.3 + 1.9 1.77 163.6 1.0139 1.0121 1.0149 O 0.9995 0.9995 0.9995
0.9830 0.9820 0.9835
6 54.2 62.5 22.4 1.72 143.3 0.9780 0.9813 0.9758 O 0.2581 0.2586 0.2555
0.4691 0.3562 0.5417
7 54.9 81.0 + 2.0 1.72 160.6 1.0157 1.0136 1.0169 O 0.9992 0.9992 0.9992
0.9794 0.9780 0.9803
8 56.5 78.6 21.5 1.69 158.8 0.9879 0.9895 0.9868 O 0.1667 0.1667 0.1667
0.9062 0.9130 0.9004
9 66.1 67.8 20.4 1.70 130.0 0.9964 0.9971 0.9962 3 0.5369 0.5179 0.5153
0.4485 0.6444 0.6583
10 66.7 64.3 20.7 1.76 140.0 0.9936 0.9947 0.9933 3 0.5769 0.6368 0.5639
0.1536 0.1112 0.1672
11 66.9 58.6 21.7 1.69 140.0 0.9831 0.9860 0.9817 O 0.2950 0.3085 0.2905
0.5303 0.4191 0.5722
12 67.0 59.2 21.5 1.68 140.4 0.9852 0.9877 0.9838 O 0.3002 0.3131 0.2931
0.4401 0.3445 0.5484

HD: hemodialysis; TBW: total body water; CRA: color reason analysis.
Symbol (1) means Watson formula; symbol (2) means Hume formula; symbol (3) means Sahlgrenska formula.
11
12 International Journal of Distributed Sensor Networks

the estimated results with a total average error of less

than 2% for 12 subjects using three formulas are
shown in Table 1. The robustness range of normal
condition was 6 2% of changes in TBW and changes
in SBP from 110 to 135 mmHg. That is, the Watson
formula has been verified and can be applied to esti-
mate TBW. This indicates that a rapidly safe and sim-
ple method can be used from a cross-sectional
standard for bedside applications in predialysis stage.
In this study, the objective was to establish an intelli-
gent vehicle with a warning tool for predialysis
healthcare. For physiologica measurements to screen
hypervolemia, these findings can provide a promising
suggestion to make changes in drink/food and to con-
trol extra BWs.

Conclusion
A strategy to monitor and control the fluid volume sta-
tus and hypertension is an important clinical issue in
HD patients. Dietary sodium restriction and fluid vol-
ume control have been performed to improve malnutri-
tion. A promising method that can provide an accurate
assessment to achieve and maintain a stable BMI and
dry weight is needed. For the prescribed normal dry
weight, Watson formulas for male and female subjects
have been validated to estimate cross-sectional TBW
with an average error of less than 1.2% to indicate the
fluid volume imbalance. The proposed fuzzy CRA with
the TBWrat and SBP is used to separate the normal
condition from hypervolemia or undervolemia. The
fuzzy CRA has a flexibility inference mechanism and
no iterative computations to update model parameters.
The recognition coefficient, j, monotonously increases
to enhance better contrast in classification applications,
while has capability of self-regulation in the primary
color grades. Hence, this simple technique can be easily
implemented in a tablet PC and an intelligent vehicle
via wireless connection, which only requires few patient
characteristics, such as S, A, W, and H parameters, as
shown by the data in the dashboard in Figure 8. This
individualized tool can enhance the priority in data
read, cloud computing, and cloud storage for patient
demands. In addition, antihypertensive medication is a
directed manner to control blood pressure, further pre-
venting the progression of congestive heart failure and
improving cardiovascular outcomes. In routine exami-
nations, TBW and blood pressure screenings can be
used to evaluate individualized characteristics for drink,
food, and pharmacologic controls. We may have a
cross-sectional reference to objectively direct dry weight
management. In addition, this promising model is an
individualized tool for dry weight maintenance in pre-
dialysis healthcare.
Figure 8. Fuzzy CRA-based screening tool was implemented
both in a tablet PC and an iPad via wired and wireless (WiFi,
IEEE 802.11) communication.
Declaration of conflicting interests
The author(s) declared no potential conflicts of interest with
respect to the research, authorship, and/or publication of this
article.
Funding
The author(s) disclosed receipt of the following financial sup-
port for the research, authorship, and/or publication of this
article: This work was supported in part by the Ministry of
Science and Technology, Taiwan, under contract nos MOST
105-2221-E-006-087-MY2 and MOST 105-2218-E-075B-001
during 1 March 2016–31 July 2017 and is also supported in
part by the research grant of Kaohsiung Veterans General
Hospital, under contract no. VGHKS 105-070 during 1
January 2016–31 December 2016.
References
1. Chamney PW, Kramer M, Rode C, et al. A new tech-
nique for establishing dry weight in hemodialysis patients
via whole body bioimpedance. Kidney Int 2002; 61:
2250–2258.
2. Tzamaloukas AH, Murata GH, Vanderjagt DJ, et al.
Body composition evaluation in peritoneal dialysis
patients using anthropometric formulas estimating body
water. Adv Perit D 2003; 19: 212–216.
3. Agarwal R. Hypervolemia is associated with increased
mortality among hemodialysis patients. Hypertension
2010; 56: 512–517.
4. Weir MR. Hypervolemia and blood pressure: powerful
indicators of increased mortality among hemodialysis
patients. Hypertension 2010; 56: 341–343.
 Chen et al.
5. Agarwal R, Flynn J, Pogue V, et al. Assessment and
management of hypertension in patients on dialysis. J
Am Soc Nephrol 2014; 25: 1630–1646.
6. Rocco MV, Yan G, Heyka RJ, et al.; HEMO Study
Group. Risk factors for hypertension in chronic hemo-
dialysis patients: baseline data from the HEMO study.
Am J Nephrol 2001; 21: 280–288.
7. Rahman M, Fu P, Sehgal AR, et al. Interdialytic weight
gain, compliance with dialysis regimen, and age are inde-
pendent predictors of blood pressure in hemodialysis
patients. Am J Kidney Dis 2000; 35: 257–265.
8. Ozerkan F, Toz H, Ozkahya M, et al. Hypervolemia in
dialysis patients–Doppler echocardiography studies.
Nephrol Dial Transpl 1998; 13: 2151–2153.
9. Hur E, Gungor O, Musayev O, et al. Bioimpedance spec-
troscopy for the detection of hypervolemia in peritoneal
dialysis patients. Adv Perit D 2011; 27: 65–70.
10. Sahn DJ, DeMaria A, Kisslo J, et al. Recommendations
regarding quantitation in M-mode echocardiography:
results of a survey of echocardiographic measurements.
Circulation 1978; 58: 1072–1083.
11. Sinha AD, Light RP and Agarwal R. Relative plasma
volume monitoring during hemodialysis aids the assess-
ment of dry weight. Hypertension 2010; 55: 305–311.
12. Ohashi Y, Otani T, Tai R, et al. Associations of protei-
nuria, fluid volume imbalance, and body mass index with
circadian ambulatory blood pressure in chronic kidney
disease patients. Kidney Blood Press R 2012; 36: 231–241.
13. Watson PE, Watson ID and Batt RD. Total body water
volumes for adult males and females estimated from sim-
ple anthropometric measurements. Am J Clin Nutr 1980;
33(1): 27–39.
14. Hwang HS, Hong YA, Yoon HE, et al. Comparison of
clinical outcome between twice-weekly and thrice-weekly
hemodialysis in patients with residual kidney function.
Medicine 2016; 95(7): e2767.
15. Lee SW, Song JH, Kim GA, et al. Assessment of total
body water from anthropometry-based equations using
bioelectrical impedance as reference in Korean adult con-
trol and haemodialysis subjects. Nephrol Dial Transpl
2001; 16: 91–97.
16. Chen SM. Fuzzy backward reasoning using fuzzy Petri
nets. IEEE T Syst Man Cy B 2000; 30(6): 846–856.
17. Cao Y and Chen G. A fuzzy Petri-nets model for comput-
ing with words. IEEE T Fuzzy Syst 2010; 18(3): 486–499.
View publication stats
13
18. Lin C-H. Assessment of bilateral photoplethysmography
for lower limb peripheral vascular occlusive disease using
color relation analysis classifier. Comput Meth Prog Bio
2011; 103(3): 121–131.
19. Gonzalez R and Woods RE. Digital image processing.
3rd ed. Upper Saddle River, NJ: Prentice Hall, 2008.
20. Vallozzi L, Van Torre P, Hertleer C, et al. Wireless com-
munication for firefighters using dual-polarized textile
antennas Integrated in their garment. IEEE T Antenn
Propag 2010; 58(4): 1357–1368.
21. Postolache OA, Silva Girao PMB, Mendes J, et al. Phy-
siological parameters measurement based on wheelchair
embedded sensors and advanced signal processing. IEEE
T Instrum Meas 2010; 59(10): 2564–2574.
22. Baek HJ, Chung GS, Kim KK, et al. A smart health
monitoring chair for nonintrusive measurement of biolo-
gical signals. IEEE T Inf Technol B 2012; 16(1): 150–158.
23. Chen C and Pomalaza-Raez C. Implementing and evalu-
ating a wireless body sensor system for automated phy-
siological data acquisition at home. Int J Comput Sci
Inform Tech 2010; 2(3): 24–38.
24. Jia G, Zhou J, Yang P, et al. A sensing chair design for
home based physiological signs monitoring. In: Proceed-
ings of the 2013 IEEE international symposium on medical
measurement and applications, Gatineau, QC, Canada,
4–5 May 2013, pp.261–264. New York: IEEE.
25. Institute of Electrical and Electronics Engineers (IEEE)
Std 802.11ä:2007. Wireless LAN medium access control
(MAC) and physical layer (PHY) specifications (12 June
2007).
26. Lin C-H, Chen W-L, Kan C-D, et al. Detection of venous
needle dislodgement during haemodialysis using frac-
tional order shape index ratio and fuzzy colour relation
analysis. IET Healthc Technol Lett 2015; 2(6): 149–155.
27. Pogue V, Rahman M, Lipkowitz M, et al. Disparate esti-
mates of hypertension control from ambulatory and clinic
blood pressure measurements in hypertensive kidney dis-
ease. Hypertension 2009; 53: 20–27.
28. Inci A, Kursat S, Kutsal DA, et al. Hypervolemia–mal-
nutrition in renal failure: is there a relationship? Clin
Nephrol Urol Sci 2015; 2: 1–8.
29. Shoji T, Tsubakihara Y, Fujii M, et al. Hemodialysis-
associated hypotension as an independent risk factor for
two-year mortality in hemodialysis patients. Kidney Int
2004; 66: 1212–1220.