Computers in Biology and Medicine 35 (2005) 275 286

http://www.intl.elsevierhealth.com/journals/cobm

Design of a portable urine glucose monitoring system for health care

Ho Dong Parka , Kyoung Joung Leea; , Hyung Ro Yoona , Hak Hyun Namb
Department of Biomedical Engineering, College of Health Science, Yonsei University, 234, Maeji-Ri, Heungup-Myon, Wonju City, Kangwon-Do 220-710, Republic of Korea b Department of Chemistry, Chemical Sensor Research Group, Kwangwoon University, Republic of Korea Received 20 May 2003; accepted 13 February 2004
a

Abstract This paper describes the design of a monitoring system that can be used to measure urine glucose during daily life. It consists of a bio-chemical sensor, hardware with PIC microcontroller and control circuits, and signal analyzing part. To evaluate the performance, we compared the analyzed glucose levels of the developed system to a standard instrument, YSI glucose analyzer, based on regression analysis using standard glucose solutions mixed with urine. Also, standard deviation and coe cient of variation were computed. In conclusion, the developed system showed it could be used for the measurement of urine glucose. ? 2004 Elsevier Ltd. All rights reserved.
Keywords: Urine glucose; Regression analysis; Standard deviation (SD); Coe cient of variation (CV)

1. Introduction Diabetes is a chronic disease found in many people. This disease can cause problems for families, society, and especially the patients that su er from its various side e ects and complications, unless proper medical treatment is received. To be treated properly, glucose monitoring is very important because regularly measured blood glucose levels over a long period can serve as a basis for diagnosis and management of diabetes. Also, for diagnosis and prevention of diabetes in a healthy person, the need for regular measurements of glucose levels over time is suggested [1,2]. Recently,
This work was supported by RRC program of MOST and KOSEF. Corresponding author. Tel.: +82-33-760-2433; fax: +82-33-760-2197. E-mail address: kjlee@dragon.yonsei.ac.kr (K.J. Lee).

0010-4825/$ - see front matter ? 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.compbiomed.2004.02.003

276

H.D. Park et al. / Computers in Biology and Medicine 35 (2005) 275 286

the self-monitoring of blood glucose in patients with Diabetes mellitus is now widely established and supported by a variety of test strips and meters. Nowadays, most of the blood glucose measuring devices on the market are invasive types, like lancet, which are painful and require blood-gathering. There is a noninvasive type using a colorimetric method that decodes the color response shown on a urine glucose measuring strip. This method is routinely used in health examinations as an indirect diagnostic index for diabetes. But the colorimetric method has many defects in terms of carrying a sample, preprocessing, quantity, and response time [3]. Recently, a new kind of glucose measuring system was developed. The GlucoWatch is an example. The GlucoWatch, which automatically extracts and measures glucose every 10 min for a period of up to 13 h before requiring replacement, is a noninvasive and painless device for measuring and monitoring the glucose level, but the GlucoWatch must be calibrated with a standard blood glucose measurement and takes 2 h to warm up. Because measurement of glucose level is available after long warm up period, it is not convenient to use [4]. In this paper, we describe our design of a biosensor that measures urine glucose level noninvasively and of a new portable urine glucose monitoring system for health care, where the measured glucose level is displayed digitally. The proposed system improves on the existing blood glucose measurement systems by reducing the amount of pain patients receive and also improves on the limitations of the type using the colorimetric method. The developed urine glucose measurement system, based on amperometric electrochemistry, is a light-weight, card-sized instrument designed for self-monitoring of urine glucose concentrations. The sample is aspirated by capillary action directly into the chamber housing the strip, and then measurement commences automatically detecting the result within 23 s. The user is not required to regulate the sample volume, remove excess urine from the strip or monitor the measurement steps in any way. The measured glucose levels are stored in the memory and can be transferred to the home health care monitoring system through the Infrared Data Association (IrDA). The home health care monitoring system uses the transferred glucose data to generate statistical reports and graphs that are designed to create a comprehensive picture of a patients recent glucose level history. 2. The design of urine glucose monitoring system The designed system computes glucose level by measuring the di erential current caused by response of urine with the biosensor. This system consists of biosensor for measuring urine glucose level, system hardware and signal processing unit. 2.1. Bio-sensor for urine glucose measurements We designed a biosensor for urine glucose measurements with the electrochemical properties of screen-printed carbon paste electrodes (CPEs) assembled with immobilized glucose oxidase and hexamine ruthenium (III) chloride ([Ru(NH3 )6 ]C13 ) containing a nitrocellulose (NC) strip. Because YSI Glucose Analyzer, which is widely used as the gold standard in examination and comparison of glucose level, uses the glucose-oxidase method, the developed biosensor as shown in Fig. 1 also adopt same method as mentioned above.

H.D. Park et al. / Computers in Biology and Medicine 35 (2005) 275 286

277

Working Electrode

Count Electrode(1)

Count Electrode(2)

URINE

Fig. 1. Diagram of the developed biosensor.

Fig. 2. Diagram of chemical reaction in biosensor.

The electrochemical reactions in the developed biosensor are shown in Fig. 2. The output of the biosensor consists of two sources of current level from electrodes. One source is a mixture of glucose and interference and the other is only interference, containing uric acid, vitamins, proteins, etc., that makes monitoring urine glucose level di cult. The density of pure glucose in urine can be detected by the di erential amperometric method. In order to measure the level of a standard glucose solution, analysis software, BAS100B electro chemical analyzer (Bioanalytical System Inc.) and Igor Pro 4.0(Wavemetrics Inc.), were used. The values for the bu er solution, which has negligible glucose density, are shown separately because of the very minute values. At normal temperature, the test of sensor lifetime is performed by placing 10 randomly selected sensors in the experimental oven for one day. The characteristics of the developed biosensor were rarely changed after 66 days in the experimental oven at 50 C. This result correspond to 2-year guarantee when converted to the normal temperature [5].

278

H.D. Park et al. / Computers in Biology and Medicine 35 (2005) 275 286

-30 -25

53th days

F irst day

-20 i, A
19th days 11th days

-15 -10 -5 0
66th days Y = - 0.1504X + 1.1557 NO. : 10

0 (a) -25 -20 -15 i, A

101 mg/dL 108

50

100 150 Glucose, mg/dL

200

100

100

104 101

-10 -5 0 (b)
20 mg/dL 18.7 20.2 20.7 21.2

20.1

20

40 Time, Days

60

Fig. 3. (a) Graph of life time tested by the developed sensor. (b) Data of life time tested by the developed sensor.

Fig. 3(a) shows the graph of the 66 day time span experiment in which the developed sensor was soaked in a 100 mg=dl glucose solution and shows the change in calibration curves over time. As can be seen in Fig. 3(b), the observed current levels are consistent at the point corresponding to a density of 100 mg=dl of glucose. The gure shows that there is little di erence between the calibration curves on the rst day and on the 66th day and there is no signicant change in the density of 100 mg=dl. 2.2. System hardware System hardware consists of a data acquisition circuit and a digital control circuit. The data acquisition circuit executes analog signal processing such as conversion of current signal from the

H.D. Park et al. / Computers in Biology and Medicine 35 (2005) 275 286

279

Fig. 4. System block diagram.

biosensor into a voltage signal and removal of noise by low-pass lter. The digital control circuit is designed based on PIC16c773 and a block diagram for the total system is shown in Fig. 4. The microcontroller controls A/D conversion of analog signal and ON/OFF control for various functions such as key input, alarm buzzer, external EEPROM for temporary storage of calculated glucose level, LCD display, low battery alarm and serial interface. We designed LCD panel to display the analyzed result and conditions of the glucose measurement system. The developed device can store up to 99 measurements. Also LCD can display a density of 999:9 mg=dl. 2.3. Signal analysis When the sensing part of the strip sensor comes in contact with urine, the process of urine analysis begins. In order to measure glucose level, we adopted the amperometric method that measures currents resulting from the oxidationreduction process generated by the chemical reaction between sensor and glucose. Also, there is an adequate time delay allowing for the full progression of the oxidationreduction reaction. For the work presented in this paper, 15 s was set as the appropriate period for the reaction between sensor and urine. Then, 240 data points/s were measured during the next 8 s. Fig. 5 shows examination lines of current versus time according to various glucose densities when the developed biosensor is stained with standard glucose solutions. Because the current level in the bu er solution is relatively low, the graph of the bu er solution is zoomed in the lower right separately.

280

H.D. Park et al. / Computers in Biology and Medicine 35 (2005) 275 286

Fig. 5. (a) Examination lines of current versus time according to various glucose densities. (b) Calibration curve of current versus glucose densities.

From Fig. 5(a), we can see the curve to be stabilized after 8 s and current varies with glucose density. The averaged value of 100 data points, consisting of 50 data points with one measurement taken before and one taken after 8 s, corresponding to the stabilized region in Fig. 5(a), was calculated. It

H.D. Park et al. / Computers in Biology and Medicine 35 (2005) 275 286

281

is shown that currents are proportional to density of glucose. Fig. 5(b) is calibration curve resulted from the averaged value as mentioned above. Eq. (1) is a formula obtained from the calibration curve, and nally the concentration of urine glucose is calculated. Urine Glucose(mg=dl) = (1=0:1818) Current( A) + 8:8586: (1)

3. The results and discussion To evaluate the performance of the designed system, three experiments were performed. First, the accuracy and reproducibility of the current-to-voltage conversion circuit was veried. Second, the calibration curve of BAS 100B electro chemical analyzer (Bioanalytical System Inc.) used to develop the biosensor was compared with that of the designed system. Third, to evaluate the reliability of the designed system, a known density of glucose was added to urine and then the analytical result of the designed system was compared with that of 2300 STAT Plus Glucose Analyzer (Yellow Spring, OH) used widely as the gold standard. The analyzed data was collected over 231 experiments with both YSI glucose analyzer and the designed system. Linear regression analysis was applied for data analysis. Fig. 6 is a picture of the developed system. 3.1. Current-to-voltage conversion circuit To evaluate accuracy and reproducibility of the current-to-voltage conversion circuit, the calibration sensor (similar to the developed biosensor) was designed with standard and variable resistance. Eq. (2) is a formula obtained from linear regression analysis of the output data of calibration sensor. Because Eq. (2) was calibrated for compensation of error which is in inherent component in the circuit, it can be applied to the second device. The standard error was 0.0259% in linear regression analysis. Calculated current( A) = 0:0676928 + 1:02431 Measured current( A): (2)

From these results, the accuracy and reproducibility using of the calibration sensor were proved to be good.

Fig. 6. Picture of the developed system.

282

H.D. Park et al. / Computers in Biology and Medicine 35 (2005) 275 286

Fig. 7. Calibration curves of BAS 100B and developed system.

3.2. Calibration curves of the designed urine glucose monitoring system Two calibration curves calculated by data measured with BAS 100B Electro Chemical Analyzer (Bioanalytical System Inc) and the designed system are represented by Eqs. (3) and (4). y = 0:1818 x 1:6105 : Designed system; y = 0:1819 x 1:6076 : BAS 100B Electrochemical Analyzer; (3) (4)

where the variable y is current and x represents density of glucose. Fig. 7 shows the two calibration curves mentioned above. The correlation coe cient value is 0.998, which indicates that the performance of the developed device and BAS 100B Electrochemical Analyzer are very similar. These results suggest that the reliability of the calibration curve to be used to compute density of glucose is acceptable. 3.3. Performance of the designed urine glucose monitoring system We performed the evaluation of condence for the detection of glucose level by comparing the standard instrument with the designed system using standard glucose mixed with urine. Linear regression was used to evaluate consistency for the detection of density. Fig. 8 shows the result of linear regression analysis based on Eq. (5). Gluco meter(mg=dl) = 0:992534 YSI + 0:279290: (5) Standard error is 2.8582%. This means that the designed system is closely behaved similarly to standard instrument. And its value implies that the density of glucose measured with the designed system has useful meaning clinically. In other words, supposing that the average glucose level of a normal person is about 80 mg=dl, standard error of 2.8582% is equivalent to the di erence of 2:29 mg=dl in glucose level. Also, the percentage ratio deviation factor distribution is less than 10% as shown in Fig. 9.

H.D. Park et al. / Computers in Biology and Medicine 35 (2005) 275 286
S = 2.85282 R-Sq = 99.5% R-Sq(adj) = 99.5% 150 Glucometer (mg/dL)
Regression 95% PI

283

100

50

N: 231

0 0 50 100 150 YSI Glucose Analyzer (mg/dL)

Fig. 8. Regression results of YSI versus the developed system.

10

% Deviation from YSI (%)

-10 0 50 100 150

YSI Glucose Analyzer (mg/dL)

Fig. 9. Residual distribution of YSI versus the developed system.

The percentage ratio deviation factor distribution is dened as percentage ratio of the di erence between the value of YSI and the designed system to the value of YSI as shown in Eq. (6). The percentage ratio deviation factor distribution (%) = |The value of YSI The value of designed system| The value of YSI 100: (6)

Because the standard regulations for a portable urine glucose monitoring system are not yet prepared, it is di cult to evaluate its utility. But American Diabetes Association (ADA) suggested that the percentage ratio deviation factor distribution for a blood glucose monitoring system should be less than 15% [6]. Considering these facts, our designed system could be used to monitor urine glucose. In order to evaluate the consistency of the linear regression analysis, we performed Anderson Daring Normality Test for residual and found P-value of 0.058. As this value is over the threshold value of 0.05, it implies that the result of linear regression is reliable.

284

H.D. Park et al. / Computers in Biology and Medicine 35 (2005) 275 286

Fig. 10. (a) SD of developed system. (b) CV of developed system.

Considering that commercial glucose monitoring systems using a biosensor calculate standard deviation (SD) and coe cient of variation (CV) as parameters evaluating the characteristics of the systems, we calculated CV and SD. SD is dened as a measurement of deviation from the mean glucose concentration [79]. CV is dened as the percent of the SD to the mean value of glucose density as shown in Eq. (7). The higher CV is, accuracy and reproducibility of the developed system decrease. CV(%) = SD mean 100: (7)

Fig. 10(a) and (b) is the bar graph representing SD and CV of our designed system. It shows that SD increases as the density of glucose increases, and CV maintains a value of less than 5%. This research was performed as one part of the development of a home health care system, which can monitor health during daily activities.

H.D. Park et al. / Computers in Biology and Medicine 35 (2005) 275 286

285

Because the proposed system was tested under laboratory condition, it is necessary to perform clinical evaluation using actual urine samples from patients in the future. After that, this system can be applied not only in diabetes patients but also healthy people for detecting initial onset of diabetes. 4. Conclusions In this paper, we designed a non-invasive, quantitative and portable urine glucose monitoring system consisting of bio-chemical sensor, system hardware and signal processing part. Through evaluation of the designed system, we obtained conclusions as follows: 1. We expect the developed device to be user-friendly as a result of its small size, small sample volume, and automatic measurement. 2. After clinical evaluation, the developed device can be served as a basis for diagnosis and management of diabetes through monitoring urine glucose level during daily activities. Also, our designed system for self-monitoring of urine glucose was found to have excellent reliability, accuracy and reproducibility under laboratory conditions using standard urine glucose solutions. In addition to providing accurate glucose results, we expect the developed device to be user-friendly as a result of its small size, small sample volume, and automatic measurement. Summary Urine glucose monitoring system is a self-monitoring system that display the glucose level by non-invasive measurement method. This paper describes the design of a monitoring system that can be used to measure urine glucose during daily life. We designed a noninvasive urine glucose monitoring system that improves the defects of urine glucose measurement using a colorimeter method and invasive blood glucose measurement method. The system consists of a bio-chemical sensor for urine glucose measurements, hardware with PIC microcontroller and control circuits, and signal analyzing part. The developed bio-chemical sensor has good reproducibility, is convenient to handle and can be mass-produced inexpensively. To evaluate the performance, we compared the analyzed glucose levels of the developed system to a standard instrument, YSI glucose analyzer, using standard glucose solutions mixed with urine. The standard error was 2.852 from the evaluation of the condence interval based on regression analysis. Also, standard deviation (SD) and coe cient of variation (CV) which are important parameters to evaluate a glucose detection system using a bio-chemical sensor, were computed. SD was 10%, which is less than the clinically valid value of 15% and CV was under 5%. In conclusion, the developed system showed it could be used for the measurement of urine glucose. Acknowledgements This work was supported by RRC program of MOST and KOSEF.

286

H.D. Park et al. / Computers in Biology and Medicine 35 (2005) 275 286

References
[1] S. Amir, C. Rabin, A. Galatzer, Cognitive and behavioral determinants of compliance in diabetics, Health Soc. Work (1990) 144151. [2] The Diabetes Control and Complications Trial Research Group, The e ect of intensive treatment of diabetes on the development and progression of long-term complications in insulin-dependent diabetes mellitus, N. Engl. J. Med. 329 (1993) 977986. [3] ADA Consensus Development Panel, Consensus statement on self-monitoring of blood glucose, Diabetes Care 10 (1987) 9599. [4] www.glucowatch.com [5] S.S. Deshpande, Enzyme Immunoassays From Concept To Product Development, Chapman & Hall, New York, 1996. [6] ADA Consensus Statement, Self-monitoring of blood glucose, Diabetes Care 55 (1994) 11271135. [7] J.M. Bland, D.G. Altman, Statistical methods for assessing agreement between two methods of clinical measurement, Lancet 1 (1986) 307310. [8] B.D. Lewis, Laboratory evaluation of the glucocard blood glucose test meter, Clin. Chem. 38 (1992) 20932095. [9] B. Harrison, A comparison of statical techniques to evaluate the performance of the glucometer elite blood glucose meter, Clin. Biochem. 29 (1996) 521527. Park, Ho-Dong received his BS and MS in Biomedical Engineering from Yonsei University, Wonju, Korea, in 2000 and 2002, respectively. Since fall 2002, he has been pursuing the Ph.D. degree in biomedical engineering at Yonsei University. He is also member of the KOSOMBE and KIEE. His main research interests are in the design of embedded medical systems, design of ambulatory module and medical instrumentation based on microprocessors. Currently, he is also working in the eld of human acupuncture and emergency medical device. Lee, Kyoung-Joung received his BS, MS and PhD in Electrical Engineering from Yonsei University, Seoul, Korea, in 1981, 1983 and 1988, respectively. Since 1989, he has worked in biomedical engineering at Yonsei University. He was a visiting professor at Case Western Reserve University, Ohio, USA, in 1993. He is currently a Professor of Biomedical Engineering at Yonsei University. He is also member of the KOSOMBE, IEEE, KIEE, and KITE. His main research interests are in the eld of ECG signal processing, design of ambulatory module and medical instrumentation based on microprocessors. Currently, he is also working in the eld of gait analysis and sleep. Yoon, Hyung-Ro received his BS, ME and PhD in Electronic Engineering from Yonsei University, Seoul, Korea, in 1972, 1974 and 1986, respectively. From 1974 to 1979, he was Research Engineer at the Agency for Defense Development, Seoul, Korea where he worked on the research and development of data acquisition and control systems. Since 1982, he has been on the Faculty of Medical Engineering. Yonsei University, where his teaching speciality is medical instrumentation and computer applications. His research interests include real-time ECG analysis, microcomputer-based medical instrumentation and telemedicine. He is also a member of the KOSOMBE, KIEE and IEEE Engineering in Biomedical Engineering Society. Nam, Hak-Hyun received his B.S. degree from Seoul National University in 1982 and Ph.D in Physical Chemistry from Michigan State University in 1989. He continued his postdoctoral work in laser spectroscopy at Professor Bradley Moors laboratory at the University of California at Berkeley until 1992. He is now a Full Professor of Physical Chemistry at Kwangwoon University. His current research interests involve the development of various types of chemical and biosensors for clinical and environmental interests.