FACULTY OF ENGINEERING UNIVERSITY OF MALAYA

DEVELOPMENT AND EVALUATION OF AN IMPROVED MULTIPLE FREQUENCY BIOIMPEDANCE ANALYZER

FOR DISEASE MANAGEMENT SYSTEM

SAMI FATHI ALI KHALIL

FACULTY OF ENGINEERING UNIVERSITY OF MALAYA

KUALA LUMPUR

2016

University

of Malaya

DEVELOPMENT AND EVALUATION OF AN IMPROVED MULTIPLE FREQUENCY BIOIMPEDANCE ANALYZER FOR DISEASE

MANAGEMENT SYSTEM

SAMI FATHI ALI KHALIL

THESIS SUBMITTED IN FULFILMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR

OF PHILOSOPHY

FACULTY OF ENGINEERING UNIVERSITY OF MALAYA

KUALA LUMPUR

2016

University

of Malaya

UNIVERSITY OF MALAYA

ORIGINAL LITERARY WORK DECLARATION Name of Candidate: Sami Fathi Ali Khalil

Registration/Matric No: KHA130004

Name of Degree: PhD in Biomedical Engineering

Title of Thesis: Development and Evaluation of an Improved Multiple Frequency Bio impedance Analyzer for Disease Management System

Field of Study: Biomedical Engineering I do solemnly and sincerely declare that:

(1) I am the sole author/writer of this Work;

(2) This Work is original;

(3) Any use of any work in which copyright exists was done by way of fair dealing and for permitted purposes and any excerpt or extract from, or reference to or reproduction of any copyright work has been disclosed expressly and sufficiently and the title of the Work and its authorship have been acknowledged in this Work;

(4) I do not have any actual knowledge nor do I ought reasonably to know that the making of this work constitutes an infringement of any copyright work;

(5) I hereby assign all and every rights in the copyright to this Work to the University of Malaya (“UM”), who henceforth shall be owner of the copyright in this Work and that any reproduction or use in any form or by any means whatsoever is prohibited without the written consent of UM having been first had and obtained;

(6) I am fully aware that if in the course of making this Work I have infringed any copyright whether intentionally or otherwise, I may be subject to legal action or any other action as may be determined by UM.

Candidate’s Signature Date:

Subscribed and solemnly declared before,

Witness’s Signature Date:

Name:

Designation:

University

of Malaya

ABSTRACT

This thesis presents the development of a novel multiple frequency bioimpedance analyzers (MFBIA). The analyzer is designed for diseases management, focusing on the management of dengue infection and the monitoring of high cholesterol level. The developed system consists of the hardware and the software modules. The advanced hardware (assembled in 8x4.5x2 cm box) measures the impedance parameters at 5 KHz, 50 KHz, 100 KHz and 200 KHz at the accuracy of 3.6 %, with an output current of 0.2 mA. The body composition prediction module was modeled from predefined equations using the data from 825 healthy Malaysian subjects. The diseases management system was developed based on the Bioimpedance Vector Analysis (BIVA) technique using data from 299 dengue patients and 339 high cholesterol elderly patients. The obtained results from the electrical testing showed maximum error at only 3.6% from the theoretical values. The validation using biological phantom presented significant correlations in estimating fat and fluid volumes using the developed MFBIA analyzer. The analysis of the equations showed that the Deurenberg’s model are the most appropriate method for the estimation of Fat-Free Mass, Fat Mass, Total Body Water, Extracellular Fluid and Intracellular Fluid as referenced to Dual Energy X-ray Absorptiometry. The investigation on the healthy subjects indicates significant correlation (Pearson’s correlation ≥ 0.8) between commercial analyzers and the developed MFBIA analyzer. The results from the BIVA analysis disclosed a significant vector shifting between the healthy and Dengue infected subjects (P<0.05) and the means impedance vector of normal and high cholesterol groups do differ significantly (P<0.05). In conclusion, a new MFBIA was designed and developed, and a novel diseases management system for dengue infection, and cholesterol level introduced.

University of Malaya

ABSTRAK

Tesis ini membentangkan rekabentuk baru sebuah mesin analisis impedans berbilang (MFBIA) untuk pengurusan penyakit. Peranti tersebut memberi tumpuan kepada pengurusan jangkitan denggi dan pemantauan tahap kolesterol yang tinggi. Sistem yang dibangunkan terdiri daripada perkakasan dan modul perisian. Perkakasan (dipasang di dalam kotak 8x4.5x2 sentimeter) boleh mengukur parameter impedans pada 5 KHz, 50 KHz, 100 KHz dan 200 KHz pada ketepatan 3.6%, dengan arus keluaran kurang daripada 1 mA. Modul komposisi badan dimodelkan menggunakan persamaan sediaada dengan menggunakan data dari sampel 825 penduduk Malaysia yang sihat. Modul pengurusan penyakit dibangunkan dengan menggunakan kaedah analisis vector bioimpedans (BIVA).

Data yang yang digunakan dikumpul daripada 299 pesakit denggi dan 339 pesakit berumur yang mempunyai kolesterol tinggi. Keputusan ujian elektrik menunjukkan ralat maksimum hanya 3.6% daripada nilai teori. Keputusan menggunakan modul biologi menunjukkan korelasi yang signifikan diantara mesin baru dalam menganggarkan jumlah lemak dan air. Analisis persamaan menunjukkan model Deurenberg adalah kaedah yang paling sesuai untuk anggaran peratusan bebas lemak, peratusan lemak, jumlah air di dalam badan, air di luar sel dan air di dalam sel apabila dirujuk kepada Dual Energy X- ray absorptiometry. Kajian menunjukkan hubungan yang signifikan (korelasi Pearson ≥ 0.8) antara mesin komersil dan mesin MFBIA baru. Keputusan analisis BIVA menunjukkan perbezaan vektor antara subjek yang dijangkiti Denggi dan yang sihat dan (P <0.05) dan vektor kumpulan kolesterol normal dan tinggi adalah berbeza dengan ketara (P <0.05). Kesimpulannya, sebuah mesin MFBIA baru telah direka dan dibangunkan dan sistem pengurusan penyakit untuk jangkitan denggi dan tahap kolesterol telah baru juga diperkenalkan.

University

of Malaya

ACKNOWLEDGEMENTS

In the name of Allah, Most Gracious, Most Merciful

All praise and glory to Almighty Allah (Subhanahu Wa Taalaa) who gave me courage and patience to carry out this work. Peace and blessing of Allah be upon last Prophet Muhammad (Peace Be Upon Him).

The unrestrained appreciation to the supervisors Dr. Mas Sahidayana Mohktar and Professor Ir. Dr. Fatimah Binti Ibrahim, who has supported me throughout my thesis with their expertise, constant help, and guidance.

Deepest gratitude to all CIME members for sharing the literature and invaluable assistance. Not forgetting to all staff in Biomedical Engineering Department, Faculty of Engineering, University of Malaya. Without their cooperation and kindness, this work could not complete.

My parents, deserve special mention for their integrated support and prayers. Words fail to express my appreciation to my wife and my son whose dedication, love and confidence in me, has taken the load off my shoulder.

Finally, I would like to thank everybody who was critical to the successful realization of this thesis, as well as expressing my apology that I could not mention personally one by one.

University

of Malaya

TABLE OF CONTENTS

ABSTRACT ... III ABSTRAK ... IV ACKNOWLEDGEMENTS ... V TABLE OF CONTENTS ... VI LIST OF FIGURES ... IX LIST OF TABLES ... XIII LIST OF SYMBOLS AND ABBREVIATIONS ... XV LIST OF APPENDICES ... XVIII

CHAPTER 1: INTRODUCTION ... 1

1.1 Overview ... 1

1.2 Objectives ... 5

1.3 Scope of work ... 5

1.4 Thesis Organization ... 6

CHAPTER 2: LITERATURE REVIEW ... 8

2.1 Introduction ... 8

2.2 Fundamentals of Bioimpedance Measurement Techniques ... 8

2.2.1 Single Frequency Bioimpedance Analysis (SFBIA) ... 12

2.2.2 Multiple Frequency Bioimpedance Analysis (MFBIA) ... 12

2.2.3 Bioimpedance Spectroscopy (BIS) ... 13

2.2.4 Whole Body Bioimpedance Measurement... 16

2.2.5 Body Segment Bioimpedance Measurement ... 17

2.2.6 Alternative Bioimpedance Analysis Method ... 19

2.3 Body Composition Prediction Using Bioimpedance Analysis ... 22

2.3.1 Fat Mass (FM) and Fat-Free Mass (FFM) ... 23

2.3.2 Body Fluids ... 26

2.4 Bioimpedance Measurement Biasing Factors ... 35

University

of Malaya

2.4.1 Anthropometric Measurements ... 35

2.4.2 Gender ... 35

2.4.3 Age ... 36

2.4.4 Ethnic Groups ... 37

2.4.5 Measurements Protocols and Posture ... 38

2.4.6 Electrode Shape and Measurement Error ... 39

2.5 Bioimpedance Analyzers ... 40

2.6 Bioimpedance Analyzer Testing and Validation Approaches ... 42

2.7 Applications of Bioimpedance Analysis in Clinical Status Monitoring and Diagnosis of Diseases ... 42

2.8 Summary ... 48

CHAPTER 3: METHODOLOGY ... 49

3.1 Introduction ... 49

3.2 Design and Development of MFBIA Hardware ... 50

3.2.1 Analog Front Unit ... 52

(a) Current Modification Circuit ... 53

(b) Instrumentation Amplifier Circuit ... 56

3.2.2 Impedance Analyzer Unit ... 59

3.2.3 Control and Communication Unit ... 62

3.2.4 Power Unit ... 64

3.2.5 Graphical User Interface ... 66

3.3 Testing and validation of MFBIA Hardware ... 67

3.3.1 Testing using Electrical Module ... 67

3.3.2 Validation Using Biological Phantom ... 70

3.4 Development of Body Composition Prediction Module ... 74

3.4.1 Investigation of Body Composition Prediction Equations ... 75

3.4.2 Validation and Evaluation of Body Composition Measurement ... 78

3.5 Comparison of Body Composition Measurement ... 80

3.6 Development of Disease Management Modules utilizing MFBIA ... 81

University

of Malaya

3.6.1 Elderly Cholesterol level Assessment Module ... 82

3.6.2 Dengue Severity Diagnosing Module ... 85

3.7 Summary ... 87

CHAPTER 4: RESULTS AND DISCUSSION ... 88

4.1 Introduction ... 88

4.2 Design and Development of MFBIA Hardware ... 88

4.3 Testing and Validation of MFBIA Hardware ... 96

4.3.1 Testing using Electrical Phantom... 96

4.3.2 Validation Using Biological Phantom ... 103

4.4 Development of Body Composition Prediction Module ... 107

4.5 Comparison of Body Composition Measurement ... 130

4.6 Development of Disease Management Module utilizing MFBIA ... 156

4.6.1 Elderly Cholesterol level Assessment Module ... 156

4.6.2 Dengue Severity Diagnosing Module ... 169

4.7 Summary ... 182

CHAPTER 5: CONCLUSION AND RECOMMENDATIONS ... 184

5.1 Introduction ... 184

5.2 Conclusion ... 184

5.3 Recommendations ... 185

REFERENCES ... 187 APPINDIX ...

University

of Malaya

LIST OF FIGURES

Figure 2.1: Main body segments and compartments (Khalil et al., 2014). ... 11

Figure 2.2: Cole-Cole module plot and Cole module parameters (Khalil et al., 2014). . 14

Figure 2.3: Whole body bioimpedance measurement techniques, (a) hand to foot and (b) foot to foot electrodes positioning (Khalil et al., 2014). ... 16

Figure 2.4: Segmental bioimpedance analysis techniques. (a) right, side dual current and quad voltage electrodes. (b) right, side dual current and quad voltage electrodes. (c) dual side, dual current and quad voltage electrodes and (d) dual side, quad current and quad voltage electrodes (Khalil et al., 2014). ... 19

Figure 2.5: Bioimpedance vector analysis (BIVA) and tolerance ellipses. Reproduced with permission (Piccoli et al., 2002a)... 20

Figure 3.1: Flowchart of the implemented methodology. ... 50

Figure 3.2: Block diagram of designed general purpose MFBIA. ... 52

Figure 3.3: Block diagram and Interfacing of the analog front unit. ... 53

Figure 3.4: Modified Howland circuit (Franco, 1988). ... 54

Figure 3.5: Integration of current modification and isolation circuit, and impedance analyzer (AD5933) (ADInc., 2015). ... 55

Figure 3.6: The standard design of instrumentation amplifier (Moore et al., 2009). ... 57

Figure 3.7: Integration of instrumentation amplifier circuit, and impedance analyzer (AD5933) (ADInc., 2015). ... 58

Figure 3.8: The functional block diagram of the impedance analyzer chip (AD5933) (ADInc., 2015). ... 60

Figure 3.9: The functional block diagram of the Bluno-Nano® board. ... 63

Figure 3.10: The programming flowchart of control and communication unit. ... 65

Figure 3.11: The implemented method for testing the developed MFBIA using the electrical module. ... 68

Figure 3.12: Equivalent electrical module for bioimpedance measurements validation (Sami F. Khalil, 2014). ... 69

University

of Malaya

Figure 3.14: The biological phantom filled with oil and saline mixture and electrodes placing (Sami F. Khalil, 2014). ... 72 Figure 3.15: The experimental setup of biological phantom during bioimpedance measurements (Sami F. Khalil, 2014). ... 73 Figure 3.16: The implemented method for body composition validation. ... 79 Figure 3.17: The implemented method for body composition evaluation using the commercial analyzers and the developed MFBIA. ... 81 Figure 3.18: The implemented method for assessment of cholesterol level in Elderly using BIVA. ... 83 Figure 3.19: The implemented method for diagnosis of Dengue severity using BIVA. 86 Figure 4.1: Circuit diagram of developed MFBIA... 89 Figure 4.2: PCB design of MFBIA. ... 90 Figure 4.3: The implemented MFBIA. ... 91 Figure 4.4: The developed GUI for, (a) Impedance Measurement, (b) Body Composition, (c) Dengue Analysis, (d) Cholesterol Analysis. ... 92 Figure 4.5: Sum of error percentages (ER %) between theoretical and measured impedance valuess, using the three designs of developed MFBIA. ... 97 Figure 4.6: Error percentages (ER %) between theoretical and measured impedance values, using the three designs of developed MFBIA. ... 98 Figure 4.7: Sum of error percentages between theoretical and measured impedance parameters values, using the LCR analyzer and developed MFBIA. ... 99 Figure 4.8: Sum of the standard error of mean (SEM) for each impedance parameters, measured using the developed MFBIA and the LCR analyzer. ... 100 Figure 4.9: Regression between surface impedance and, oil percentage (Red) and Saline (Blue) at (a) 5 KHz, (b) 50 KHz, (c) 100 KHz and (d) 200 KHz. ... 104 Figure 4.10: The linear correlation analysis of FFM measurements between (a) QuadScan® 4000 and DEXA, (b) developed MFBIA and DEXA, for males (red) and females (blue). ... 115 Figure 4.11: The Bland-Altman plot of mean differences of FFM measurements between (a) QuadScan® 4000 and DEXA, (b) developed MFBIA and DEXA, for males (red) and females (blue). ... 116

University

of Malaya

Figure 4.12: The linear correlation analysis of FM measurements between (a) QuadScan®

4000 and DEXA, (b) developed MFBIA and DEXA, for males (red) and females (blue).

... 117 Figure 4.13: The Bland-Altman plot of mean differences of FM measurements between (a) QuadScan® 4000 and DEXA, (b) developed MFBIA and DEXA, for males (red) and females (blue). ... 118 Figure 4.14: The linear correlation analysis of TBW measurements between (a) QuadScan® 4000 and DEXA, (b) developed MFBIA and DEXA, for males (red) and females (blue). ... 120 Figure 4.15: The Bland-Altman plot of mean differences of TBW measurements between (a) QuadScan® 4000 and DEXA, (b) developed MFBIA and DEXA, for males (red) and females (blue). ... 121 Figure 4.16: The linear correlation analysis between QuadScan® 4000 and developed MFBIA for (a) ECF and (b) ICF measurements, for males (red) and females (blue). . 122 Figure 4.17: The Bland-Altman plot of mean differences between QuadScan® 4000 and developed MFBIA for (a) ECF and (b) ICF measurements, for males (red) and females (blue). ... 124 Figure 4.18: The linear correlation analysis of FFM measurements between (a) developed MFBIA and QuadScan® 4000, (b) developed MFBIA and BIA450, for males (red) and females (blue). ... 138 Figure 4.19: The Bland-Altman plot of mean differences of FFM measurements between (a) developed MFBIA and QuadScan® 4000, (b) developed MFBIA and BIA450, for males (red) and females (blue). ... 139 Figure 4.20: The linear correlation analysis of FM measurements between (a) developed MFBIA and QuadScan® 4000, (b) developed MFBIA and BIA450, for males (red) and females (blue). ... 141 Figure 4.21: The Bland-Altman plot of mean differences of FM measurements between (a) developed MFBIA and QuadScan® 4000, (b) developed MFBIA and BIA450, for males (red) and females (blue). ... 142 Figure 4.22: The linear correlation analysis of TBW measurements between (a) developed MFBIA and QuadScan® 4000, (b) developed MFBIA and BIA450, for males (red) and females (blue). ... 143 Figure 4.23: The Bland-Altman plot of mean differences of TBW measurements between (a) developed MFBIA and QuadScan® 4000, (b) developed MFBIA and BIA450, for males (red) and females (blue). ... 144

University

of Malaya

Figure 4.24: The linear correlation analysis of ECF measurements between (a) developed MFBIA and QuadScan® 4000, (b) developed MFBIA and BIA450, for males (red) and females (blue). ... 145 Figure 4.25: The Bland-Altman plot of mean differences of ECF measurements between (a) developed MFBIA and QuadScan® 4000, (b) developed MFBIA and BIA450, for males (red) and females (blue). ... 146 Figure 4.26: The linear correlation analysis of ICF measurements between (a) developed MFBIA and QuadScan® 4000, (b) developed MFBIA and BIA450, for males (red) and females (blue). ... 147 Figure 4.27: The Bland-Altman plot of mean differences of ICF measurements between (a) developed MFBIA and QuadScan® 4000, (b) developed MFBIA and BIA450, for males (red) and females (blue). ... 148 Figure 4.28: The reference ellipses (50%, 75 % and 95 %) and impedance vector distribution for healthy adults, male (a) and female (b). ... 161 Figure 4.29: The reference ellipses (50%, 75 % and 95 %) and impedance vector distribution for normal elderly, male (a) and female (b). ... 162 Figure 4.30: A 95 % confidence ellipses of normal cholesterol and high cholesterol elderly subjects, based on total and LDL cholesterol level, for male (a) and female (b).

... 163 Figure 4.31: A BMI based 95 % confidence ellipses of Normal cholesterol and high cholesterol elderly subjects, for male (a) and female (b). ... 165 Figure 4.32: The 95 % confidence ellipses of normal adults, elderly and high cholesterol elderly subjects, for male (a) and female (b)... 167 Figure 4.33: A 95 % confidence ellipses for healthy (solid) and Dengue infected (dashed) subjects, for female (a) and male (b) (Khalil et al., 2016). ... 172 Figure 4.34: A 95 % confidence ellipses for non-severe (solid) and severe Dengue infected (dashed) subjects, for female (a) and male (b) (Khalil et al., 2016). ... 175 Figure 4.35: The reference ellipses (50%, 75 %, and 95 %) for healthy subjects, female (a) and male (b) (Khalil et al., 2016). ... 176 Figure 4.36: The ellipses immigration for healthy (sold), non-severe (dashed) and severe (dots) Dengue infected subjects, for female (a) and male (b) (Khalil et al., 2016). ... 181

University

of Malaya

LIST OF TABLES

Table 2.1: Summary of the commercially available bioimpedance analyzers. ... 41 Table 2.2: Applications of bioimpedance analysis in clinical status monitoring and diagnosis of diseases. ... 44 Table 3.1: The design specification of the developed MFBIA. ... 51 Table 3.2: The electrical module impedance sets and the theoretical values of impedance parameters. ... 70 Table 3.3: The utilized prediction equations for FFM, FM, TBW, ECF and ICF. Where (ht) is the body height, (wt) is the body weight, (R50) and (XC(50)) are resistance and reactance at 50 kHz, Z5, Z50, Z100 is bioimpedance at 5, 50 and 100 kHz. ... 75 Table 4.1: The cost analysis of the developed MFBIA. ... 94 Table 4.2: Comparison of cost, dimension, measurement powering and disease management facilities of available commercial bioimpedance analyzers. ... 95 Table 4.3: The design optimization of the developed MFBIA. ... 102 Table 4.4: The coefficient of determination (R²) for oil and saline contents using commercial and developed MFBIA. ... 106 Table 4.5: The demographic and anthropometric data, for body composition validation group. ... 109 Table 4.6: The bioimpedance parameters measured by Quad Scan® and developed MFBIA, for body composition validation group. ... 110 Table 4.7: The body composition parameters measured by DEXA, Quad Scan® and developed MFBIA, for validation group. ... 112 Table 4.8: The demographic and anthropometric data, for body composition evaluation group. ... 131 Table 4.9: The bioimpedance parameters measured by Quad Scan® and developed MFBIA, for body composition evaluation group. ... 132 Table 4.10: The body composition parameters measured by DEXA, Quad Scan® and developed MFBIA, for the evaluation group. ... 135 Table 4.11: The demographic and anthropometric data for healthy and high total cholesterol elderly Malaysian subjects. ... 157

University

of Malaya

Table 4.12: The bioimpedance parameters for healthy and high cholesterol elderly Malaysian subjects. ... 159 Table 4.13: bioimpedance vector parameters for two BMI categories (Normal and Overweight). ... 164 Table 4.14: Demographic and BIVA parameters for healthy (Group 1) and Dengue- infected subjects (Group 2), (P<0.05) (Khalil et al., 2016). ... 170 Table 4.15: Demographic and BIVA parameters for severe and non-severe Dengue subjects (P<0.05) (Khalil et al., 2016). ... 173 Table 4.16: Analysis of mean (robust Welch’s test) and variance between healthy, non- severe and severe Dengue patients, (P < 0.05) (Khalil et al., 2016). ... 177

University

of Malaya

LIST OF SYMBOLS AND ABBREVIATIONS

AC : Alternative Current

ADC : Analog to Digital Converter

ANOVA : Analysis of Variance

BCM : Body Cell Mass

BIS : Bioimpedance Spectroscopy

BIVA : Bioimpedance Vector Analysis

BMI : Body Mass Index

BPF : Band Pass Filter

C : Capacitance

c : Volume Fraction of Non-Conducting Tissue CMOS : Complementary Metal-Oxide Semiconductor

CMRR : Common Mode Rejection Ratio

DAC : Digital to Analog Converter

Db : Body Density

DC : Direct Current

DDS : Direct Digital Synthesizer

DEXA : Dual-Energy X-ray Absorptiometry

DFT : Discrete Fourier Transform

DSP : Digital Signal Processing

ECF : Exteracellular Fluid

ECF_Den_MFBIA : Exteracellular Fluid using Deurenberg’s Equation ECF_Han_MFBIA : Extracellular Fluid using Hannan’s Equation ECF_Seg_MFBIA : Exteracellular Fluid using Segal’s Equation ECF_Ser_MFBIA : Exteracellular Fluid using Sergi’s Equation

University

of Malaya

FFM : Fat-Free Mass

FFM_Den_MFBIA : Fat-Free Mass using Deurenberg’s Equation FFM_Kot_MFBIA : Fat-Free Mass using Kotler’s Equation FFM_Kyle_MFBIA : Fat-Free Mass using Kyle’s Equation FFM_Sun_MFBIA : Fat-Free Mass using Sun’s Equation FFM_Sun_MFBIA : Fat Mass using Sun’s Equation

FM : Fat Mass

FM_Den_MFBIA : Fat Mass using Deurenberg’s Equation FM_Kot_MFBIA : Fat Mass using Kotler’s Equation FM_Kyle_MFBIA : Fat Mass using Kyle’s Equation

GUI : Graphical User Interface

Ht : Height

IC : Integrated Circuit

ICC : Intraclass Correlation

ICF : Intracellular Fluid

ICF_Den_MFBIA : Interacellular Fluid using Deurenberg’s Equation ICF_Han_MFBIA : Intracellular Fluid using Hannan’s Equation ICF_Seg_MFBIA : Interacellular Fluid using Segal’s Equation IDE : Integrated Development Environment JFET : Junction Gate Field-Effect Transistor

Kb : Dimensionless shape factor

MFBIA : Multiple Frequency Bioimpedance Analyzer NCEP : National Cholesterol Education Program

PCB : Printed Circuit Board

R : Resistance

SCL : Serial Clock Line

University

of Malaya

SDA : Serial Data Line

SEE : Standard Error of the Estimate

SEM : Standard Error Mean

SFBIA : Single Frequency Bioimpedance Analyzer

SMM : Skeletal Muscle Mass

TBW : Tota Body Water

TBW_Den_MFBIA : Total Body Water using Deurenberg’s Equation TBW_Han_MFBIA : Total Body Water using Hannan’s Equation TBW_Kot_MFBIA : Total Body Water using Kotler’s Equation TBW_Kush_MFBIA : Total Body Water using Kushner’s Equation TBW_Kyle_MFBIA : Total Body Water using Kyle’s Equation TBW_Luk_MFBIA : Total Body Water using Lukaski’s Equation TBW_Seg_MFBIA : Total Body Water using Segal’s Equation TBW_Sun_MFBIA : Total Body Water using Sun’s Equation

USB : Universal Serial Bus

WHO : World Health Organization

Wt : Weight

Xc : Reactance

Z : Impedance

ρ : Resistivity

φ

University

of Malaya

LIST OF APPENDICES

APPINDIX A: Component Data sheets……….206

APPINDIX B: Programming Codes……..……….337

APPINDIX C: SPSS Reports……….374

APPINDIX D: Ethics approval………..732

University

of Malaya

CHAPTER 1: INTRODUCTION 1.1 Overview

Bioimpedance analysis is a broadly applied approach used in body composition measurements and healthcare assessment systems. The essential fundamentals of bioimpedance measurement in the human body and a variety of methods are used to interpret the obtained information. Also, there is a broad spectrum of utilization of bioimpedance in healthcare facilities such as disease prognosis and monitoring of vital body status.

The electrical properties of biological tissues are currently categorized based on the source of the electricity, i.e. active and passive response. Active response (bioelectricity) occurs when biological tissue provokes electricity from ionic activities inside cells, as in electrocardiograph (ECG) signals from the heart and electroencephalograph (EEG) signals from the brain. Passive response occurs when biological tissues are simulated through an external electrical current source (Kyle et al., 2004b). Bioimpedance or biological impedance defined as the ability of the biological tissue to impede electric current (Martinsen et al., 2011).

Studies on the electrical properties of biological tissues have been going on since the late 18th century (Kyle et al., 2004a). Thomasset (1962) explored the utilization of bioimpedance measurement in total body water (TBW) estimation using needle electrodes. Nyboer (1970), applied quad surface electrode readings for bioimpedance measurements to estimate the lean mass of the human body. Hoffer (1969) introduced the association between total body impedance and body water content reference to tritium dilution techniques.

Due to the noninvasiveness of bioimpedance analysis systems, numerous researchers have conducted studies on bioimpedance analysis and its applications in body

University

of Malaya

composition estimation and evaluation of clinical conditions. Recently, several studies reviewed the applications of bioimpedance analysis in body composition assessment and monitoring of chronic diseases with a comprehensive listing of the most used equations (Khalil et al., 2014; Mialich et al., 2014). Lukaski (2013) has revised the conceptual modules of bioimpedance analysis for physiological activities assessment and diseases prognosis. The study states that the applied multiple regression approaches and physical modules in bioimpedance analysis have limited utilization in individuals’ measurement.

Bioimpedance analysis in healthcare practice contributes to the estimation of body compartments readings of fat mass (FM) and fat-free mass (FFM) to assess the regular change in nutrition status in inpatients and to monitor nutritional risk in outpatients (Kondrup et al., 2003). Observation of body compartments fluctuation like FFM, FM, and TBW from normal limits considered as key factors to be used in bioimpedance analysis in healthcare application systems. The abnormal loss in lean body mass and unbalance shift in body fluids are the most measured parameters to be used to assess the healthiness of human body. Most of the body composition assessment methods like body mass index (BMI) techniques, skin fold method, and underwater weight measurements is used to estimate FM and FFM. However, bioimpedance analysis can estimate FM and FFM in addition to total and particular body fluids which is much helpful for diseases prognosis (Thibault et al., 2012). Since 1988, National Health and Nutrition Examination Survey program in the United States had decided to include bioimpedance analysis in the third NHANES program. It designed to assess the health and nutritional status of adults and children because of a general frustration with the dependability of skinfold thickness method to estimate FM and FFM, especially in subjects with a higher amount of segmented fat (Kuczmarski, 1996).

University

of Malaya

Disease management system is a coordinated healthcare intervention and communications system for predefined diseased populations. It implements self-care support for individuals to manage and monitor the progression of their diseases (Hunter et al., 1997). Interest in disease management systems is growing due to the cost effectiveness and optimization in healthcare personnel (Goetzel et al., 2005).

Disease management systems widely applied in health care services, to optimize the resources and enhance the quality of delivered services. Studies state that disease management systems have a high economic impact, particularly in minimizing the hospitalizing expenses and emergency rooms occupation (Hamner, 2005; Mattke et al., 2007). Studies stated that disease management system reduces healthcare expenses due to minimizing the hospital admission charges (Sidorov et al., 2002). In Malaysia, the annual total health expenditure (THE) costs estimated at $6.6 billion (WHO, 2010).

Recently with the rapid growth in telecommunication technologies; there are more than 4.7 million new application sectors were established (Broens et al., 2007). It benefits patients where traditional delivery of health services is affected by distance and lack of local specialist clinicians to deliver services. In May 2008 The UK’s Department of Health’s Whole System Demonstrator (WSD) concluded that the disease management system reduces the mortality rate of 45 % (Health, 2009).

Diagnosis system is a subsection of the disease management system that defines as the practice of medical care using the interactive media. The diagnosis system includes the delivery of medical assistance, diagnosis, consultation and treatment (Setyono et al., 2011). Wagner et al. (2001; 2005), introduced a chronic care model (CCM) to apply and evaluate disease management system. This module suggested that the ideal disease management system is consist of six components, including healthcare system, community resources and policies, self-management support, delivery system design,

University

of Malaya

decision support and clinical information systems. The disease management system needs three main components: an input device, processing and communication device; and output device (Norris, 2002).

The essential part of disease management system is an input device. Many medical devices have used for remote patients monitoring to integrate it to diagnosis system. The input device defined as the apparatus used to collect and send the medical data to processing terminal distantly. These devices may include ECG, blood pressure devices, glucose monitoring devices and body composition analyzers. Many studies conclude that these devices should be noninvasive, easy to use and low cost, to be used directly by subjects in point of care area (Perednia et al., 1995).

Analysis of bioimpedance information obtained at 50 kHz electric current is known as Single-frequency bioimpedance analysis (SFBIA). Analysis of bioimpedance that obtained at more than two frequencies is known as Multiple-frequency bioimpedance analysis (MFBIA). Analysis of bioimpedance data obtained using a broad band of frequencies is known as bioimpedance spectroscopy (BIS). SFBIA is the most used and one of the earliest proposed methods for the estimation of body compartments. It is based on the inverse proportion between assessed impedance and water amount inside the human body (Kyle et al., 2004a; Ward et al., 2007). The BIS method based on the determination of resistance at zero frequency (R0) and resistance at infinity frequency (Rinf) that is then used to predict extracellular fluid (ECF) and TBW, respectively.

MFBIA considered more precise than SFBIA in the assessment of internal body composition and more cost effective than BIS method (Olde et al., 1997).

Most of the available commercial MFBIA are bulky, not fully portable, and utilized derivative prediction modules for overall body composition monitoring. Specifically, to

University

of Malaya

date, no disease management module utilized the MFBIA for clinical surveillance and diagnosis of Dengue severity and cholesterol level in elderly.

This thesis presents a non-invasive, inexpensive and portable MFBIA based diseases management system for Dengue severity and cholesterol level in elderly. The developed MFBIA achieved significant results in body composition measurements with referenced to gold standard methods. The developed system had statistically significant outcomes in diagnosing the Dengue infected and the high cholesterol patients as case studies for system validation. The results from this study will have a high impact on medical informatics area of research and diseases management professionals.

1.2 Objectives

The general aim of this thesis is to develop small and portable body composition and diseases management MFBIA analyzer. The following specific objectives have to meet:

i. Design and develop a small and fully portable MFBIA analyzer for body composition monitoring inclusive of testing and validating the design.

ii. Investigate and validate the measurements of body composition prediction equations with the gold standard method.

iii. Compare the developed MFBIA analyzer with commercial bioimpedance analyzers.

iv. Model disease management modules for Dengue severity and cholesterol level in elderly.

1.3 Scope of work

This study will focus on the design and development of a portable and low-cost MFBIA analyzer assuming that biasing factors in this design, including the electronic components tolerance, has minor effects on the final results. Testing and validation the

University

of Malaya

developed MFBIA using electrical and biological phantom. Calculate body composition estimations using predefined prediction equations from literature. Analyze the body composition estimations with reference to the gold standards. Perform comparison study with the commercial bioimpedance analyzers using a statistically significant sample size of Malaysian subjects. Develop diseases management modules for Dengue infected and high cholesterol diagnosis for elderly.

1.4 Thesis Organization

The thesis consists of five chapters, including the introduction, literature review, methodology, results and discussion, and conclusion and future works. Chapter 1 includes general views of diseases management systems and bioimpedance analysis and also presents the problem statement, the objectives, the hypothesis and the significance of the study.

Chapter 2 contains brief literature reviews concerning the fundamentals of bioimpedance measurement techniques, including SFBIA, MFBIA, BIS, whole body and segmental bioimpedance measurements and the alternative bioimpedance analysis methods. Also, includes the body composition prediction equations of FFM and FM and body fluids. The bioimpedance measurement biasing factors including anthropometric measurements, gender, age, ethnic groups, measurements protocols and posture, electrode shape and measurement error, and some of applications of bioimpedance analysis in clinical status monitoring and diagnosis of diseases.

Chapter 3 describes the design and development of the general purpose MFBIA analyzer system, including design specifications, design testing using electrical and biological modules, design validation and evaluation on healthy human subjects. Also, illustrate the development process of diseases management system.

University

of Malaya

Chapter 4 presents the results and discussions of the system tested using electrical and biological modules and the validation on healthy human subjects. Also, present the results from the diseases management system.

Chapter 5 presents the overall conclusion of the work accomplished to date and a summary of the future work for applying the developed MFBIA analyzer system on healthy subjects and patients.

University

of Malaya

CHAPTER 2: LITERATURE REVIEW 2.1 Introduction

This chapter review of the fundamentals and the applications of bioimpedance analysis. The first section highlights the main bioimpedance measurement approaches using a single frequency, multiple frequencies, and broadband frequency spectrum signals, in addition to applied bioimpedance measurements method across the whole body, through-body segments and another alternative analysis method such as vector bioimpedance analysis and real-time bioimpedance methods. Body composition parameters, which include lean mass and fluid volumes estimation using bioimpedance measurements, are discussed in the second section. Basic factors in bioimpedance measurements, including anthropometric measurements, age, race, protocols and postures, and shape and artifacts of the electrode discussed in the third section. Finally, applications of bioimpedance analysis in diseases prognosis and clinical monitoring systems outlined in the fourth section.

2.2 Fundamentals of Bioimpedance Measurement Techniques

Impedance (Z), from an electrical point of view, is the obstruction to the flow of an alternating current. It depends on the frequency of the applied current, defined in impedance magnitude (|Z|) and phase angle (φ) as shown in Equations (2-1)–(2-3) (Kasap, 1997). Bioimpedance is a complex quantity composed of resistance (R) which is caused by TBW and reactance (Xc) that is due by the capacitance of the cell membrane (Kyle et al., 2004a):

Z = R + jX_c (2-1)

|Z| = √R²+ X_c² (2-2)

University

of Malaya

𝛷 = tan⁻¹(𝑋_𝑐

𝑅) (2-3)

The resistance of an object determined by a shape, that described as length (L) and surface area (A), and material type, that is described by resistivity (ρ), as shown in Equation (2-4), (Kasap, 1997). Reactance (Xc) of an object as shown in Equation (2-5), is defined as resistance to voltage variation across the subject matter and is inversely related to signal frequency (f) and capacitance (C) (Kasap, 1997). In biological systems resistance is caused by total water across the body, and reactance occurs due to the capacitance of the cell membrane (De Lorenzo et al., 1997; Kyle et al., 2004a):

R_(ohm) = ρ_(Ω.m) L_(m) A_(m2)

(2-4)

X_c(ohm) = 1

2π f_(Hz) C_(Farad) (2-5)

Capacitance (C) defined as the ability of the non-conducting object to save electrical charges, that is equal to the ratio between differentiation in voltage across the object (dV/dt) and current that passed through the object (I(t)), as shown in Equation (2-7). In the parallel capacitor module, the capacitance is in direct proportion to the surface area (A) and inversely proportional to distance (d) between the charged plates. Also, it depends on the permittivity constant of vacuum (ε0 ≈ 8.854 × 10⁻¹² F.m^–1) and the relative dielectric permittivity constant (εr). This constant defined based on the material between the plates (for a vacuum space, εr = 1), as shown in Equation (2-6) (Kasap, 1997):

C_(Farad)= ε₀ ε_r A_(m²₎

d_(m) (2-6)

C_(Farad) = dV(t)

dt ⁄I(t) (2-7)

Body composition estimation using bioimpedance measurements based on the

University

of Malaya

From Equation (2-4) that gives the relation between resistance and ratio of length (L) to surface area (A), body volume (Vb) can obtained by substituting the surface area (A) with the numerator and denominator of the length (L), as in Equation (2-8):

V_b (m³) = ρ(Ω.m)

L²_(m²₎

R_(ohm) (2-8)

The human body as a volume is generally composed of FM which considered as a Non-Conductor of electric charge and is equal to the difference between body weight (WtBody) and FFM, as shown in Equation (2-9); and FFM, which considered as the conducting volume that helps the passing of electric current due to the conductivity of electrolytes dissolved in body water. Studies show that water, known as TBW is the major compound of FFM and is equal to 73.2% in normal hydration subjects, as in Equation (2- 10) (Genton et al., 2002):

𝐹M = Wt_Body− FFM

(2-9) TBW = 0.73 FFM

(2-10) In bioimpedance measurements, the human body is divided into five inhomogeneous segments, two for upper limbs, two for lower limbs and one for the trunk. In the five compartment module, the human body is composed of FM, FFM, TBW, ECF and intracellular fluid (ICF) (Kyle et al., 2004a). Figure 2.1, shows the five segments and compartments of the human body.

University

of Malaya

Figure 2.1: Main body segments and compartments (Khalil et al., 2014).

Most of the known prediction methods rely on the relation between water volume and the ratio between square length to resistance (L²/R) (Thomas et al., 1998b). However, the alternation in anatomical and anthropometric features of the whole human body and segments cause variations in estimated volumes. Jaffrin and Morel (2008) reviewed that most TBW estimation equations between 1985 and 1994. Studies showed that TBW were predicted based on values of the H²/R50 (Kyle et al., 2001a).

University

of Malaya

Measurement of bioimpedance is obtained from the whole body and body segments separately, using a single frequency, multiple frequencies, and bioimpedance spectroscopy analysis. Several alternative bioimpedance methods used such as bioimpedance vector analysis and real-time bioimpedance analysis.

2.2.1 Single Frequency Bioimpedance Analysis (SFBIA)

Analysis of bioimpedance information obtained at 50 kHz electric current is known as single-frequency bioimpedance analysis. SFBIA is the most used and is one of the earliest proposed methods for the estimation of body compartments. It based on the inverse proportion between assessed impedance and TBW that represents the conductive path of the electric current (Kyle et al., 2004a; Ward et al., 2007).

SFBIA predicts the volume of TBW that is composed of fluctuating percentages of ECF which is almost equal to 75% of TBW and ICF that represent the rest (Kyle et al., 2004a). SFBIA instruments have been used to assess TBW and FFM using the derived Equations (2-2) and (2-3), respectively, for normally hydrated subjects, although SFBIA is not valid for body conditions with significantly altered hydration (Lukaski et al., 1986).

Studies by Hanai (1968) on mixture theory report that body tissue conductivity is diverse (Kyle et al., 2004a). This studies stated that SFBIA shows limitations in ICF variance prediction, however many of studies show an acceptable correlation in ICF estimation (Olde et al., 1997).

2.2.2 Multiple Frequency Bioimpedance Analysis (MFBIA)

Analysis of bioimpedance that obtained at more than two frequencies is known as multiple frequency bioimpedance analysis. MFBIA based on the finding that the ECF and TBW can assessed by exposing it to low and high-frequency electric currents, respectively. Thomasset (1962) has proposed TBW and ECF estimation using 100 and 1 kHz based on the Cole model (Cole et al., 1941). However, in later years, Jaffrin et al.

University

of Malaya

(2008) stated that technically a bioimpedance analyzer should use frequency range between 5–1000 kHz. Simpson et al. (2001) state that low frequency in MFBIA is less than 20 kHz and high frequency is more than 50 KHz.

Hannan et al. (1995) report that parameters measured using a frequency of less than 5 KHz and more than 200 KHz fluctuate around the actual value and conclude that estimated TBW is more accurate using the MFBIA than the BIS method with the same predicted values of ECF for both approaches. Patel et al. (1996 ) reported that in diseased subjects, TBW prediction using SFBIA gave more precise results than MFBIA. In general, the MFBIA method predicts ECF more precisely than the SFBIA method;

however, in elderly diseased subjects, the MFBIA method shows less sensitivity in detecting fluid shifts between ECF and ICF (Olde et al., 1997).

2.2.3 Bioimpedance Spectroscopy (BIS)

Analysis of bioimpedance data obtained using a broad band of frequencies is known as BIS. The BIS method based on the determination of resistance at zero frequency (R0) and resistance at infinity frequency (Rinf) that is then used to predict ECF and TBW, respectively. The use of 100 and 1 kHz, respectively, was earlier proposed by Thomasset (1963). He applied the basics of Hanai’s mixture theory (Hanai, 1968), and Cole’s module (Cole et al., 1941) as explained by the Cole-Cole plot (Figure 2.2). However, it is complicated to directly measure these values because of the relaxation phenomena of living tissue (Woodrow et al., 1996).

Reference methods for estimating TBW based on radioisotopic dilution of deuterium method. And for ECF estimation they are based on the dilution of bromide (Pastan et al., 1992) and for ICF they are based on the radioactive potassium isotope, ⁴⁰K, both elements which readily diffused in the human body (Ellis et al., 1998; Scanferla et al., 1990).

University

of Malaya

Figure 2.2: Cole-Cole module plot and Cole module parameters (Khalil et al., 2014).

Referenced techniques are invasive, expensive and complicated when compared to bioimpedance methods, although the precision is dependent on the electrical module and body parameter variation (Jaffrin et al., 2008).

Estimation of TBW, ECF and ICF using BIS techniques can be performed using an equation modules approach (De Lorenzo et al., 1997; Jaffrin et al., 2006; Matthie et al., 1998; Matthie, 2005). Moreover, analytically derived equations approach (Cornish et al., 1996). Hanai’s mixture theory shows limitations in some studies (Baarends et al., 1998;

Chertow et al., 1997; Gudivaka et al., 1999). However it showed advantages in other studies (Cox-Reijven et al., 2000; Earthman et al., 2000). Ward et al. (1998) stated that the differences in biological construction among subjects may limit mixture theory as noted in some studies (Deurenberg et al., 1996; Hannan et al., 1995). Scharfetter et al.

University

of Malaya

(1997) report that an accurate module for body fluid allocation and trusted suitable methods are most crucial factors in the BIS method.

The determination of Cole module parameters (R0, Rinf, α, Fc), in Figure 2.2 is done using the BIS method which based on the argument that the human body is composed of a mixture containing conducting and non-conducting compartments (Hanai, 1968). In Equation (2-4), the reference method based on the assumption that the measured resistance (R) represents the total conducting volume of the lean body mass. However, in the BIS method, the measured resistance represents the total conducting and non- conducting part of the lean body mass, so that the non-conducting part is included by multiplying the obtained resistance by body shape factor (Kb) and substituting the surface area (A) by body volume (Vb). Ayllon et al. (2009) report that the estimation of Cole module parameters (R0, Rinf, α, Fc) that obtained by using only resistance achieves slightly better results and there is less standard error based on the non-linear least squares technique as compared to the capacitive and impedance complex components. Ward et al. (2006) conclude that the Cole parameters can be obtained by using four selected frequencies and substituting a fitting technique based on amplitude-impedance values at these frequencies:

R = K_b ρHt²

V_b (2-11)

Where, R is resistance, ρ is resistivity, Ht is the human height, Vb is the body volume.

Kb is a dimensionless shape factor calculated from the length and perimeters of the upper and lower limbs, and the trunk. Taken into consideration the body shape composed of the five cylinders. Van Loan et al. (1987) calculated the shape factor (Kb) from statistical anatomical measurements in adults to be equal to 4.3.

University

of Malaya

2.2.4 Whole Body Bioimpedance Measurement

Measurement of total body bioimpedance is the most commonly used method for estimating whole body compartments. Many of the whole body bioimpedance instruments apply three approaches for impedance measurement. Hand to foot method (Gudivaka et al., 1999; Lukaski et al., 1986). Foot to foot (Jebb et al., 2000; Utter et al., 1999; Xie et al., 1999). And hand to hand method (Deurenberg et al., 2002b; Ghosh et al., 1997). The hand to foot (Figure 2.3 (a)) one is the most commonly used method. It was introduced by Hoffer (1969) and later revised (Nyboer., 1970) to decrease the contact impedance between skin and electrodes, and validated in 140 normal adults (Lukaski et al., 1986).

Figure 2.3: Whole body bioimpedance measurement techniques, (a) hand to foot and (b) foot to foot electrodes positioning (Khalil et al., 2014).

Tetrapolar hand to foot measurements are performed on a supine subject for 15 min.

Placing electrodes filled with gel to minimize gap impedance on the dorsal surfaces of

University

of Malaya

the right hand and foot. Distal (current) ones being respectively proximal to the metacarpal and metatarsal phalangeal joints, by standard tetrapolar electrode placement (Buchholz et al., 2004). Foot to foot measurements (Figure 2.3 (b)) were introduced through the use of a pressure-contact foot-pad electrode (Nuñez et al., 1997). In leg to leg bioimpedance measurements, the subject stood vertically, with uncovered feet, on four stainless steel foot pads electrodes and divided for each foot into a frontal and back portion for current injecting and voltage measurement (Utter et al., 1999). Hand to hand bioimpedance measurements were introduced by performing body composition analyses using a handheld impedance meter in subjects with malnutrition (Ghosh et al., 1997). The device held while both arms were stretched out horizontally in front of the body.

Deurenberg et al.(2002b) validated the hand to hand method on 298 Singaporean subjects and reported that readings obtained using a handheld impedance meter were significantly acceptable for those subjects.

2.2.5 Body Segment Bioimpedance Measurement

Segmental bioimpedance analysis achieves a better estimation of skeletal muscle mass (SMM) than whole body bioimpedance analysis, with a reported standard error of 6.1%

about MRI measurements among 30 male subjects (Tanaka et al., 2007). Baumgartner et al. (1988) stated that multi-frequency segmental bioimpedance analysis enhances and elucidates the relationship between bioimpedance analysis and body compartment estimation after examining the impact of phase angle on body composition prediction among 116 normal subjects.

Segmental bioimpedance analysis detects the fluctuation in ECF due to differences in posture. It is more precise than the ankle foot method (Thomas et al., 1998a), and gives a better estimation of TBW than total body measurements concerning dilution method (Thomas et al., 2003).

University

of Malaya

Segmental or perpendicular bioimpedance analysis defines the measurement method of body segments that mostly treated as five cylinders as in Figure 2.1, (Ward et al., 2007).

It introduced to overcome the disagreement between trunk resistance to upper limbs ratio and trunk resistance to lower limbs ratio of 0.72 and 0.66 respectively (Baumgartner et al., 1988). Earthman et al. (2007) stated that the trunk represents 50% of the body mass.

Kyle et al. (2001a) pointed out that total bioimpedance measurement assesses mainly the upper and lower limb compartments, and shows some limitation to predict water compartments of the trunk.

Measurement of segmental bioimpedance can achieved through four types of protocols. The first approach uses dual current injection electrodes on the proximal area of the right forearm and lower leg, and quad voltage electrodes placed on the right proximal forearm, shoulder, upper thigh and lower leg (Figure 2.4 (a)) (Scheltinga et al., 1991). The second approach is suggested by Zhu et al. (1998), through the sum of segments technique. He uses dual current injection electrodes on the right wrist and foot, and quad voltage electrodes placed on the right wrist, shoulder, upper iliac spine and foot (Figure 2.4 (b)). A third approach was presented by Organ et al. (1994), who suggested the use of dual current injection electrodes on the right wrist and foot, and quad voltage electrodes, two placed on the right wrist and foot, and two on the left wrist and foot (Figure 2.4 (c)). The fourth approach as suggested by Jaffrin et al. (2009), is through the use of quad current injection electrodes located on the right and left wrist and foot, and quad voltage electrodes located at the same place (Figure 2.4 (d)).

Limitations of whole body bioimpedance measurement in evaluating body segment compartments have given rise to the demand for segment localized bioimpedance analysis applications. Studies reported that using segmental (across the waist) localized bioimpedance analysis can significantly estimate abdominal fat with a correlation

University

of Malaya

coefficient of R² = 0.99 (Scharfetter et al., 2001); furthermore Seward et al. (2002), introduced localized bioimpedance analysis as a trending diagnostic tool for neuromuscular disorders. The study applied on 25 neuromuscular patients and 45 normal subjects for control.

Figure 2.4: Segmental bioimpedance analysis techniques. (a) right, side dual current and quad voltage electrodes. (b) right, side dual current and quad voltage electrodes. (c) dual side, dual current and quad voltage electrodes and

(d) dual side, quad current and quad voltage electrodes (Khalil et al., 2014).

Studies report that the segmental bioimpedance analysis method shows some limitations in the estimation of FFM (Jaffrin et al., 2009; Thomas et al., 1992), with estimation power not significantly different from whole body bioimpedance method (Xie et al., 1999). However, Kyle et al. (2001a) concluded that enhancement could achieved through applying the MFBIA method and further studies on electrode types and allocation.

2.2.6 Alternative Bioimpedance Analysis Method

Bioimpedance analysis, as an independent method for the assessment of the human

University

of Malaya

analysis and interpretation. The bioimpedance vector analysis method (BIVA) is a novel approach established essentially by Piccoli et al. (1996; 1994) to estimate the hydration status using height indexed resistance and reactance data (R-Xc graph) from bioimpedance measurements. Using 8,022 normal subjects (3796 female and 4226 male) Piccoli et al. (2002b) formulated 50%, 75%, and 95% tolerance ellipses. It determines to increase and decreasing body mass if the minor vector falls on the left and right half of the 50 % ellipse, along with increasing and decreasing hydration ratio if the major vector falls in the lower and upper half of the 50% ellipse (Figure 2.5).

Figure 2.5: Bioimpedance vector analysis (BIVA) and tolerance ellipses.

Reproduced with permission (Piccoli et al., 2002a).

Evaluation study of the BIVA method by Cox-Reijven et al. (2003), on 70 diseased subjects with gastrointestinal disorders, conclude the high specificity and low sensitivity of BIVA method in classifying patients with extraordinary rates of body fluids. Low values (Xc/H < 27.7 Ω/m and R/H < 563.6 Ω/m) in the BIVA method can consider as predictors of severity among ill children. This result is shown in a study conducted on

University

of Malaya

332 precarious pediatric patients with multiple organ dysfunctions (MODS), acute respiratory distress syndrome (ARDS) and acute lung injury (ALI) (Azevedo et al., 2013).

In (Haas et al., 2012) the BIVA method successfully monitored rapid increases in ECF during short-term recovery (3 weeks). And a dramatic increase in BCM during long-term recovery (3 months) among 47 % of 57 diseased women with anorexia nervosa (Kyle et al., 2004a).

The BIVA method also considered as a valid tool for the estimation of dry weight in 24 hemodialysis patients’ concerning the Bilbrey Index based on the different allocation of values before and after obtrusion (Norman et al., 2012).

Kyle et al. reported that the BIVA method is affected by differences in biological factors and measurement artifacts (Kyle et al., 2004a). Ward and Heitmann state that BIVA is affected by body size and influenced by the cross-sectional area of the body (Ward et al., 2000).

A specific BIVA method proposed by Marini et al. (2012) to neutralize the bias due to to body size. The specific BIVA method used a resistivity-reactivity graph that is constructed using information and results collected from the multiplication of resistance and reactance by the ratio of cross-section area and length (L/A) from Ohm’s law (Equation (2-1)). The cross-section area (A) and length (L) were estimated as follows: A

= (0.45(arm area) + 0.45(calf area) + 0.10(waist area)) in square meter (Bracco et al., 1996; Yanovski et al., 1996). Where segment area = c²/4π and (c) is the circumference in the meter of the arm, waist, and calf, respectively; L = 1.1 (Ht), where Ht is body height in meters.

Another alternative method for analysis is real-time processing of bioimpedance data which currently introduced as a key feature for body health monitoring applications. A

University

of Malaya

logarithmic analysis carried out between 0.01 and 10 Hz with five frequencies needs 276 s to be completed; this includes the calculation time (Sanchez et al., 2012b). Studies (Sanchez et al., 2013a) stated that real-time processing, accuracy and the ability of data retrieval and throughput of a BIS system were the most important features to be applied in health monitoring systems. Sanchez et al. (2011a) introduced a local polynomial based method for impedance-frequency-response estimation. Comparison studies between four different multi-sine periodic broadband excitations broadband for EIS measurements in term of accuracy and speed in frequency and time domain concluded that multi-sine and discrete interval binary sequences (DIBS) enhance SNRZ and have better accuracy than chirp and maximum length binary sequences (MLBS) (Sanchez et al., 2012b).

Use of multi-sine excitation signals in bioimpedance measurements that proposed in (Sanchez et al., 2012a; Sanchez et al., 2011b) helped increase the accuracy of the measured bioimpedance parameters. It has been validated using a set of optimal multi- sine measurements on 2R-1C equivalent electrical circuits, then applied on healthy myocardium tissue. The multi-sine excitation method introduced as a parametric-in-time identification method for electrical bioimpedance measurements with the inclusion of harmonic impedance spectra (HIS). HIS directly identified from noise current and voltage myocardium measurements at the multi-sine measurement frequencies to express periodic changes of impedance, rather than the commonly used method that assumed the measurement changing over time (Sanchez et al., 2013b).

2.3 Body Composition Prediction Using Bioimpedance Analysis

Body composition assessment is considered a key factor for the evaluation of general health status of humans. Several methods use different assumptions to estimate body composition based on the number of compartments. This review considers that the human body is composed of two main compartments, FM and body lean mass or FFM. FFM

University

of Malaya

consists of bone minerals and body cell mass (BCM) that includes skeletal muscle mass (SMM). BCM contains proteins and TBW that represents 73% of the lean mass in normally hydrated subjects. TBW is composed of ICF and ECF as illustrated in Figure 2.1. In this section, several predictive equations for both lean and FM, in addition to body fluids, will be discussed.

2.3.1 Fat Mass (FM) and Fat-Free Mass (FFM)

FM and FFM estimations are considered one of the main objectives of body composition assessment techniques. Variations in FM among the reference population are due to several factors but are believed to follow aging factors in addition to gradual changes in lifestyle (Kyle et al., 2001b).

Anthropometric and skin fold thickness measurements are traditional, simple and inexpensive methods for body fat estimation. It used to assess the size of specific subcutaneous fat depots (Roubenoff et al., 1995). Compared with other methods such as underwater weighing, dilution method and dual-energy x-ray absorptiometry (DEXA) that requires a trained practitioner to perform it.

Bioimpedance analysis has shown in recent studies to be more precise for determining lean or FM in humans (Kyle et al., 2000). In comparison with BMI, anthropometric and skinfold methods, BIA offers trustable results in the estimation of fatness across human tissues (Heitmann, 1994). Several studies conducted to establish reference values for FFM based on bioimpedance measurements.

Kyle et al. (2001a) developed a single Equation (2-12) for the prediction of FFM, using 343 normal subjects aged from 22 to 94 years old, with BMI between 17.0 and 33.8 kg/m² in reference to DEXA method:

University

of Malaya

FFM = −4.104 + 0.518 ht²/R₅₀ + 0.231 wt + 0.130 X_c,50) + 4.229 sex ; Sex { 1, Male

0, Female

(2-12)

Where (Ht) is body height, (R50) and (XC, 50) is resistance and reactance at 50 KHz, and (Wt) is body weight. The developed equation achieved a correlation coefficient (R) that is equal to 0.986, standard error of the estimate (SEE) is equal to 1.72 kg and technical error is 1.74 kg.

In (Kyle et al., 2001b; Kyle et al., 2003), FFM assessed in a population of 5,225 white subjects aged from 15 to 98 years old using bioimpedance measurements. Moreover, it concluded that mean FFM was 8.9 kg or 14.8% lower in men older than 85 years than in men 35 to 44 years old and 6.2 kg or 14.3% lower in women older than 85 years than in women 45 to 54 years old. Study (Sun et al., 2003), used a multi-component model based on densitometry, isotope dilution, and dual-energy X-ray absorptiometry to build Equations (2-13) and (2-14) for FFM estimation:

FM = Wt_Body− FFM

(2-13) TBW = 0.73 FFM

(2-14) The mean FFM prediction equations achieved a correlation coefficient R² = 0.90, and 0.83 and root mean square errors of 3.9 and 2.9 kg for males and females, respectively.

Deurenberg et al. (1991a), used densitometry, anthropometry and bioelectrical impedance to formulate FFM Prediction Equation (2-15) using 661 normal adult subjects aged from 7 to 83 years old:

FFM = −12.44 + 0.34 Ht²⁄R₅₀+ 0.1534 Ht + 0.273 Wt

− 0.127 Age + 4.56 Sex ; Sex { 1, Male 0, Female

(2-15)

University

of Malaya

The FFM prediction equations achieved a correlation coefficient R² = 0.93 and standard estimation error (SEE) = 2.63 kg. (Pichard et al., 2000), assessed FFM and FM in a 3,393 white subject population aged from 15 to 64 years old using bioimpedance measurements. He performed a comparison of %FM as determined by BIA with %FM determined using BMI developed by Deurenberg et al. (Deurenberg et al., 1991b).

Moreover, concluded that the FFM ranged of 59.1–61.0 kg for men and 43.3–44.1 kg for women which is 38% greater in men. Heitmann (1990) compared three body composition methods (BMI, skinfolds, and BIA) using 139 healthy subjects aged from 35 to 65 years old:

FM = 14.94 − 0.079 Ht²/R₅₀ + 0.818 wt − 0.231 ht

− 0.064 sex wt + 0.077 Age (2-16) The multiple regression Equation (2-16) for impedance had a higher correlation coefficient (R² = 0.89) and lower standard estimation error (SEE = 3.32 kg) than the multiple regression equations for skin fold (R² = 0.81, SEE = 3.91 kg) or BMI (R² = 0.85, SEE = 3.94 kg).

Studies (Heitmann, 1991) assessed FFM and FM in 2987 out of a 3608 subject Danish population aged from 35 to 65 years old. The obtained data, which estimated from measurements of electrical impedance, concluded that men have an FM of 4.5 kg, an increase of 30% when compared to women that have a 6.9 kg increase of 36% for evaluated sample.

Recently, study (Pichler et al., 2013) assessed FM in 116 subjects (32 healthy subjects and 84 patients) and concluded that the following prediction equation (2-17 and 2-18) overestimated FM by 6.55 ± 3.86 kg:

University

of Malaya

FM_Male = −18.42 + 0.60 Wt − 0.57 Ht²

R_tbw+ 0.62Ht²

R_ecf (2-17)

FM_Female= −9.81 + 0.65 Wt − 0.66 Ht²

R_tbw+ 0.65Ht²

R_ecf (2-18)

Where Recf and Rtbw represent the resistance of ECF and TBW extracted using the Cole module (Cole et al., 1941). In conclusion, all studies state that the men have higher estimated FM as compared to women. Moreover, FFM for both