Prediction of Glucose Concentration by Impedance ...

2 downloads 0 Views 406KB Size Report
Universidad Autónoma de San Luis Potosí,. Álvaro Obregón 64, San Luis Potosí, SLP, México 78000 e-mail address: [email protected].uaslp.mx. Abstract.
Prediction of Glucose Concentration by Impedance Phase Measurements Edgar Guevara, Francisco Javier González Instituto de Investigación en Comunicación Óptica Universidad Autónoma de San Luis Potosí, Álvaro Obregón 64, San Luis Potosí, SLP, México 78000 e-mail address: [email protected] Abstract. In recent years researchers have explored non-invasive techniques for glucose testing such as near infrared spectroscopy, light scattering, photoacoustic spectroscopy, and electrical impedance measurements, without achieving the accuracy of the traditional invasive method. In this paper, measurements in the 1-13 MHz band show that glucose directly affects the impedance parameters of solutions and the effect is more evident in the phase angle; therefore, these impedance phase measurements can be employed to predict the concentration of glucose in vitro, obtaining a smaller error of prediction than previous techniques. Keywords: Glucose, non-invasive monitoring, electrical impedance spectroscopy. PACS: 84.37.+q 84.37.+q,, 87.85.Ox

INTRODUCTION Diabetes mellitus is a chronic disease in which blood glucose levels are too high. Abnormally high levels of glucose can damage the small and large blood vessels, leading to diabetic blindness, kidney disease, amputations of limbs, stroke, and heart disease[1,2]. According to the Mexican National Health Information System there are more than 10 million people who are currently diagnosed with diabetes and it is the main cause of mortality in Mexico, accounting for 13.6 percent of deaths by disease[3]. Frequent determination of glucose concentrations in diabetic patients is crucial for effective treatment and reduction of the mortality of diabetes. However, the traditional testing method involves pricking your finger with a lancet (a small, sharp needle), putting a drop of blood on a test strip and then placing the strip into a meter that displays your blood glucose level; that is uncomfortable or even painful for the patient, and may be difficult to perform in long term diabetic patients due to calluses on the fingers and poor circulation, which arises the need to develop a method to monitor glucose concentration in a painless way[4]. Among the non optical approaches for measuring blood glucose concentration, electrical impedance spectroscopy, also known as dielectric spectroscopy is widely used. Several reports describe the use of impedance-based detection systems for noninvasive glucose monitoring [5,6]. It is well known that changes within physiological CP1032, Medical Physics - Tenth Symposium on Medical Physics, edited by G. Herrera Corral and L. M. Montaño Zetina © 2008 American Institute of Physics 978-0-7354-0556-1/08/$23.00

259 257 Downloaded 08 Sep 2008 to 148.224.6.12. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/proceedings/cpcr.jsp

glucose levels do not affect directly the dielectric spectrum of skin and underlying tissue in a broadband frequency spectrum. Therefore, concentration cannot be measured directly by impedance spectroscopy. However, changes in glucose concentration lead to changes in the electrolyte balance in blood, cells and interstitial fluid both in healthy subjects and in patients with diabetes. The resulting changes in the AC and DC conductivity can be measured using impedance spectroscopy. Nevertheless, there are some partially contradictory results, which state that glucose directly affects the impedance parameters of solutions, specially at frequencies below the MHz band [8]. In this work the glucose concentration is determined by electrical impedance phase measurements.

EXPERIMENTAL The experimental setup is based on a HP4192A Impedance Analyzer, which can be set to a working frequency within the range from 5 Hz to 13 MHz with 1 mHz maximum resolution. Measurement range of absolute value of impedance |Z| is 0.1m 0.1mΩ Ω to 1.2999Ω 1.2999Ω; phase angle θ is -180º to +180º, with an accuracy of 0.1%. The sample is attached to the HP16047A Test fixture, so that a four terminal device can be used. Measurement data is logged to the computer, via the IEEE-488 interface. The frequency range of 1 MHz to 13 MHz is chosen so as not to have problems with electrode polarization and the huge contribution of moisture in human tissue [7]. A total of 241 measurements per sample were made in this interval, at steps of 50 kHz. The impedance phase spectra is displayed on Fig. 1 (a), it can be shown that the phase angle behaves accordingly to the concentration of the solution.

FIGURE 1. (a) Impedance phase spectra for different concentrations of glucose. (b) Correlation between the predicted and reference glucose concentration.

RESULTS The prediction model was determined using Partial Least Squares Regression (PLS), which was applied to measurements obtained experimentally, as described above. The correlation between the reference and predicted values shows a good 260 258 Downloaded 08 Sep 2008 to 148.224.6.12. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/proceedings/cpcr.jsp

prediction performance with a correlation coefficient squared (r2) of 0.958425, and a root mean square error of prediction (RMSEP) of 717.2932 mg/dl. Since we did not have too many samples, the same solutions were used both for model estimation and testing, employing full cross–validation. One sample is left out from the data set at a time, and the model is calibrated on the remaining data points. Then the value for this left-out sample is predicted. The process is repeated with another sample, and so on until every sample has been left out once; then the RMSEP and r2 are computed. Figure 1 (b) shows the result of PLS calibration by full crossvalidation.

CONCLUSION The results suggest that the measurement of electrical impedance can be used as a non-invasive approach to determine the glucose concentration in human blood. Further studies will be carried out to evaluate the usefulness and limitations of combining information from this technique with near infrared (NIR) spectroscopy to develop a joint technique for in vivo monitoring.

REFERENCES 1. http://www.fda.gov/diabetes/ 2. http://www.diabetes.org/home.jsp 3. Sistema Nacional de Información en Salud. “Principales causas de mortalidad general” http://sinais.salud.gob.mx/mortalidad/tabs/m_005.xls 4. R. W. Waynant, and V. M. Chenault, “Overview of Non-Invasive Fluid Glucose Measurement Using Optical Techniques to Maintain Glucose Control in Diabetes Mellitus”, LEOS Newsletter 12 (2), 3-6 (1998). http://www.ieee.org/organizations/pubs/newsletters/leos/apr98/overview.htm 5. A Caduff , E Hirt , Y. Feldman , Z Ali , L Heinemann. “First human experiments with a novel noninvasive, non-optical continuous glucose monitoring system,” Biosensors and Bioelectronics 19 19,, 209-217 (2003) 6. A. Caduff, F. Dewarrat, M. Talary, G. Stalder, L. Heinemann, and Y. Feldman, “Non-invasive glucose monitoring in patients with diabetes: A novel system based on impedance spectroscopy,” Biosensors and Bioelectronics 22 22,, 598–604 (2006). 7. Caduff, Dr. A., On the occasion of the Annual Meeting of the EASD, Munich, 05-Sep-04 “Impedance spectroscopy based glucose monitoring” 8. A Tura, S. Sbrignadello, S. Barison, S. Conti, G. Pacini. “Impedance spectroscopy of solutions at physiological glucose concentrations,” Biophysical Chemistry 129 129,, 235-241 (2007) 9. C. Araujo-Andrade, I. Campos-Cantón, J.R. Martínez, G. Ortega-Zarzosa, F. Ruiz. “Modelo de predicción basado en análisis multivariante para la determinación de concentración de azúcar en solución” Revista Mexicana De Física E 51 (2) 67–73 (2005)

261 259 Downloaded 08 Sep 2008 to 148.224.6.12. Redistribution subject to AIP license or copyright; see http://proceedings.aip.org/proceedings/cpcr.jsp