This is an author-created, un-copyedited version of an article accepted for publication in Physiological Measurement. IOP Publishing Ltd is not responsible for any errors or omissions in this version of the manuscript or any version derived from it. The definitive publisher authenticated version is available online at http://dx.doi.org/10.1088/0967-3334/32/1/009 .

Effect of Blood Content on the Optical and Dielectric Skin Properties. P Zakharov, F Dewarrat, A Caduff and M S Talary Solianis Monitoring AG, Leutschenbachstrasse 46, CH-8050 Zürich, Switzerland Email: [email protected] Abstract. A wearable system incorporating sensors for dielectric and optical spectroscopy was used to study skin properties and their dependence on the cutaneous blood content (CBC). Simultaneous measurements with both modalities were carried out on the upper arm during blood perfusion-provoking exercises performed by 4 subjects in 4 separate sets of experiments. By relating changes in the attenuation of green (central wavelength c =568 nm)

and infrared ( c =798 nm) light the ratio of mean pathlengths travelled by photons in the skin blood plexus was obtained. The pathlength for infrared light is found to be 3.85 times larger than for green. Combining signals of two wavelengths and accounting for pathlength difference we quantitatively characterise the CBC as a cumulative optical thickness of red blood cells in the skin plexus. The dielectric spectra of skin in the MHz range were fitted with the Cole-Cole model and the changes of parameters were quantitatively related to the opticallyderived changes in CBC using a linear regression analysis. The positive correlation with CBC is obtained for the dispersion exponent ( R  0.68 ), and the negative 2

for the dispersion time ( R  0.40 ). Thus dielectric dispersion of the skin gets broader and shifts towards lower frequencies with an increase of CBC. Keywords: non-invasive glucose monitoring; cutaneous blood content; diffuse reflectance spectroscopy; dielectric spectroscopy 2

Optical and Dielectric Skin Properties

1

2

Introduction Non-invasive glucose monitoring (NIGM) could potentially provide a great relief to millions of patients with diabetes mellitus (DM). A majority of the approaches considered in recent years by different industrial and academic research groups are based on exploring the characteristics of the optical or dielectric properties of skin [1,2]. In a typical arrangement, the skin is probed with the electromagnetic (EM) radiation of the range specific to glucose or glucose-induced changes and the response serves as a signal for NIGM. It has been repeatedly shown that under specific conditions, typically referred to as “controlled conditions”, various NIGM technologies have demonstrated the ability to monitor changes in glucose levels in the human body. However, as soon as these conditions become less favourable (towards uncontrolled conditions that more closely represent those expected in daily life use), NIGM technologies have failed to demonstrate the same level of functionality [2]. This can be attributed to the effects of the various perturbations, with the cutaneous perfusion having profound effect [3]. It is known that skin is actively employed in the body’s thermoregulation by adjusting the amount of the blood in microvascular structures of the skin. Using this mechanism, it is able to adapt heat exchange with the environment according to the difference in internal and external temperatures [4,5]. Changes in cutaneous blood content (CBC) can also be caused by such factors as the position of the whole body and its parts, physical and mental activities, food, drugs or chronological cycles [6]. Blood perfusion of the tissue can affect the propagation of EM radiation in different ways, including the direct absorption of radiation by blood proteins (as for visible light, water absorption) scattering by red blood cells (RBC) (in the case of infrared radiation), membrane polarisation for radio frequencies, and molecular polarisation of GHz frequency ranges [3]. Additionally, blood perfusion has an effect on a number of factors that leads to changes in the morphology of the skin, temperature and ionic balance in the skin tissue and on the skin surface. These factors can also indirectly influence the propagation of EM radiation in the skin. From a dielectric perspective, blood is the most conductive matter in the skin and one of the most conductive in the body [7-10] and variations of it’s content in the tissue is a significant perturbing factor for non-invasive glucose monitoring based on dielectric measurements [3,11] where changes in the tissue are not related to glucose changes. In order to calculate estimated glucose concentrations from non-invasive measurements with a sufficient reliability in real-life settings, one has to account for variations of blood content in the specific probed volume. This can be done by probing the tissue volume using EM radiation in the spectral region which has the high sensitivity to the blood content. Such a spectral region can be located outside of the range of the primary glucose sensor and thus CBC monitoring can be performed using a separate sensor for this function. This is a basic idea behind the concept of the multisensor glucose monitoring system (MGMS) which is under development by Solianis Monitoring AG [12], The MGMS incorporates dielectric sensors, for the extraction of glucose related information, along with optical sensors dedicated for the measurement of blood perfusion, as well as sensors for measuring temperature, skin moisture and body movements. The compensation of the blood-related perturbations can be performed if a quantitative link between the optical and dielectric signals can be established. Here we present the theoretical basis and an experimental procedure which allows such a relation to be established.

Optical and Dielectric Skin Properties

2 2.1

3

Theory Skin structure

Figure 1 Simplified scheme of the skin structure with distribution of blood vessels.

In order to investigate the effects of blood relocation on the optical and dielectric properties of the skin, one has to take into account the complex structure of the skin’s microvascular system. In general, skin can be subdivided into several idealised horizontal layers as shown in Figure 1 [13]. The uppermost viable skin layer, the epidermis, does not have its own vasculature and relies for its support on the diffusive transport of nutrients from the dermal blood capillaries. As simplistically depicted in Figure 1, blood is mainly localised in two geometrically separated layers of the dermis – the sub-papillary or upper vascular plexus (UVP) and the cutaneous or deep vascular plexus (DVP), which is located just above the subcutis. The arterioles in the UVP form the papillary loops in the bumpy structure of the dermal-epidermal junction, which increases the interface area with the epidermis for enhanced exchange of metabolites. The UVP contains the largest part of the skin microvasculature [14,15] and is of primary importance for the thermoregulatory changes in the CBC [16]. As discussed above, changes of the blood content in the vascular plexuses affect the optical and dielectric properties of the skin volume probed by non-invasive sensors. In order to explain the relationship of the optical and dielectric properties of skin a brief review of measurement and analysis methods is included below. 2.2 Characterisation of cutaneous blood content A number of non-invasive techniques can be used for the characterisation of the tissue blood content [17,18], including various optical, electrical, thermal and acoustic approaches. Amongst these techniques, the light reflectance spectroscopy in the visible and near-infrared ranges stands out as reliable and suitable for a wearable application thanks to high specificity and cost-effectiveness [17,19,20]. The unique spectroscopic signature of blood in the optical spectral region is defined by meta-protein haemoglobin (Hb) which is contained in the RBCs. Hb plays the role of the major blood oxygen-carrier. Previously we have reported on a diffuse reflectance spectroscopy (DRS) system for wearable application which was optimised for the monitoring of blood content in the upper

Optical and Dielectric Skin Properties

4

skin layers [19]. We have introduced the difference in attenuation of 568 nm green and 798 nm infrared light as a parameter characterising the CBC. While the green light provides the sensitivity to the haemoglobin absorption, infrared allows for the compensation of the variations in the light coupling to the skin and to other wavelength-independent factors. In order to be able to relate this parameter to physiologically relevant parameters, such as CBC or concentration of RBCs, one has to quantitatively link the pathlengths of green and infrared light in the blood plexus of the skin. This can be achieved by provoking changes in the CBC while probing the tissue with both wavelengths of light. In the same experiment one can also establish a correlation between the changes in optical and dielectric properties of the measured skin tissue when the measurements are made simultaneously and the same measurement tissue volume is probed. Analysis of these correlations can help establish the method to compensate the blood-induced disturbances of dielectric measurements in order to improve reliability of the glucose estimate that is calculated from these measured signals. In the current study we use an experimental protocol which mimics the real-life challenges expected during daily use of the wearable MGMS device, which can be induced by the activity of the person. These challenges consist of the local orthostatic manoeuvres of lowering and raising of the arm with the sensor attached to the lateral side of the upper forearm and bending the arm at the elbow. Bending induces additional applied pressure of the sensor to the skin. These perfusion-provoking exercises carried out in a controlled setting have a number of advantages. One of the most important is the simplicity of the experiment which is favoured both by the subjects and clinical personnel. It does not require additional equipment such as a pressure cuff to impede blood flow in the vasculature. Experimental conditions are more reproducible in comparison to pressure cuff occlusion since the effect of the latter depends on the exact position of the cuff on the arm and the timing of the cuff’s inflation and deflation. The effect of the orthostatic stress on the microcirculation in the upper limbs has been well studied and described in the literature [21-23]. The placement of the arm below the heart level increases the vascular resistance due to the veno-arteriolar reflex and myogenic responses. It has been observed by Maeda et al. [23] as the DC component of the photoplethysmogram and Engelhart et al. [21] using the laser Doppler flowmetry (LDF) and 133Xenon washout techniques. Here lowering the arm to 40 cm below the heart level results in a similar response as a venous occlusion with 40 mmHg pressure. Additionally orthostatic manoeuvres can provide an insight on the expected ranges of CBC changes in the uncontrolled home use settings. 2.3 Diffuse reflectance spectroscopy In an idealised reflectance measurement of a homogeneous medium, light of known intensity I 0 enters the sample at source location and the intensity I of light diffusely reflected from the sample volume is measured at a point of detection. The wavelengthdependent light attenuation A    ln I   / I 0   is used to characterise the concentration changes of substances. In the skin the measured attenuation AM   is composed of the attenuations of blood AB   and bloodless tissue AT   and a component G , associated with the loss due to coupling of light to the sample and the properties of the optical system [24]: (1) AM    AB    AT    G , Attenuation of blood can be further decomposed to the attenuation of RBCs ( ARBC ) which make up about 99% of the blood cells and attenuation of plasma. In the visible and shortwavelength infrared (SWIR) region, impact of the plasma is negligible and thus we can write

5

Optical and Dielectric Skin Properties

AB  ARBC , where RBC attenuation is defined by the absorption of haemoglobin molecules contained within RBCs and scattering by the RBCs themselves. c  CtHb / Ht is relatively stable (31-36 The mean cell haemoglobin concentration CtHb g/dL). At the same time the proportion of blood volume occupied by the RBCs known as the hematocrit ( Ht  CRBC CB ) varies significantly even for healthy adult subjects (typical range is 0.36 – 0.53) and is also unstable in time for a single subject [5]. Also the scattering properties of blood are defined by the scattering of RBCs. Thus it makes sense to quantitatively relate light attenuation to the RBCs content, while the link to the blood content is dependent on hematocrit. The total attenuation of light produced by the RBCs can be expressed using a Beer-Lambert law in its modified form [19,24-27]: ARBC       eff  LRBC

,

(2)

where LRBC is the nominal path length in the total intracellular space of RBCs or the effective optical thickness of the layer of RBCs,    is a effective wavelength-dependent pathlength factor (PF), and eff    3 a   a     ' s   is an effective attenuation

coefficient [28] with  a   being the absorption coefficient of RBCs and  's   – the reduced scattering coefficient of the volume of RBCs (or blood with Ht  1 ). The product of nominal pathlength and wavelength-dependent PF   LRBC determines the path which is taken by light within the RBC intracellular space on it’s way from the light source to the detector. The intracellular path depends on the total pathlength of light in the tissue L  and concentration of RBCs CRBC :

  LRBC  CRBC L  .

(3)

Thus the optical thickness can serve as a measure of the RBC concentration in the tissue if we assume L  remains constant during the monitoring. The selection of nominal pathlength is arbitrary, thus we can set it to the pathlength with the highest attenuation, which also corresponds to the shortest path. In this case     1 . In the SWIR and visible spectral ranges the RBC absorption is defined by absorption of haemoglobins: c c  a    CrHb  arHb    CoHb  aoHb   ,

(4)

c c where C oHb and C rHb are the mean corpuscular concentrations of the oxygenated and c c c  CoHb  CrHb reduced haemoglobins, correspondingly (we assume that CtHb ). For the case of the isosbestic wavelength i , where the absorption of both haemoglobins is

c  atHb i  . equal:  arHb i    aoHb i    atHb i  we can write  a i   CtHb By choosing a pair of wavelengths 1 and 2 with a significantly different attenuation of light by the haemoglobin as compared with the attenuation by bloodless tissue: AB 1   AB 2   AT 1   AT 2  one can increase the signal specificity to the RBCs by taking the difference of the attenuation on those wavelengths according to eq. (1) and (2): AM 1   AM 2   ARBC 1   ARBC 2    1  eff 1    2  eff 2 LRBC . Thus from the measured attenuations we can obtain the effective optical thickness of RBCs as

6

Optical and Dielectric Skin Properties

LRBC 

AM 1   AM 2   1  eff 1    2  eff 2 

(5) ,

Using the Eq. (5) we can characterise the effective amount of blood content in the tissue in relation to the optical thickness of RBCs. The two attenuations are measured and effective attenuation coefficients can be derived from the literature. The only parameter, which has to be determined, is the relation of pathlength factor for the two wavelengths. This relationship can be estimated using a Monte-Carlo simulation [19] or experimentally. Experimental approaches can involve sophisticated measurement methods, such as time-resolved [24], phase-resolved (Chance et al. 1998) measurements or usage of the water absorption spectral features [30]. However, if one can induce the attenuation changes specific to a certain chromophore, the pathlength factor can be derived by relating the attenuations, as shown in the following section. 2.4 Pathlength factor According to eq. (5), the optical thickness can be calculated using two measured attenuations and attenuation coefficients that can be derived from the literature. The only parameter, which has to be determined is the relation of PF for two wavelengths. This can be done in the experiment with the physiological procedure inducing the specific changes in the cutaneous blood content and correspondingly the RBC content CRBC . The resulting change in the measured attenuation can be written as following, according to Eq. (1) and (2):

AM    l0  C RBC  eff    LRBC   eff  

(6)

where we assume that there are no changes in the attenuation of tissue AT   and the light loss component G . By combining attenuation changes of two isosbestic wavelengths we can find PF as follows:

 2  AM 2   eff 1    1  AM 1   eff 2 

(7)

For a signal wavelength 1 with the maximal light attenuation we can assume that light propagates by a shortest path and thus we can take this pathlength as a reference:  1   1. Thus the PF of the second wavelength  2  can be found from eq. (7). While the absolute pathlengths are dependent on the skin morphology and the exact location of the blood-containing layers of the skin, the pathlength factor is more a function of the measurement geometry and the layers optical properties, which are assumed to be similar. Thus for a given optical sensor one can obtain the  2  from a calibration set of experiments and further use this value to calculate LRBC . In the following sections we present a clinical protocol that can be used to alter CBC and monitor the changes with the spectroscopic system in order to obtain PF and quantify changes of LRBC . 2.5 Dielectric skin properties As a typical experimental setup for performing non-invasive dielectric measurements of the skin, capacitive fringing field electrodes are attached to the skin surface and a complex impedance can be measured using a low alternating current. The impedance is defined in the following way: (8) Z  Z e j  Re Z  j Im Z

Optical and Dielectric Skin Properties

7

The amplitude Z and phase  are measured. The admittance Y (the inverse of the impedance) for a parallel circuit of the conductance and capacitance is given by

Y  Z 1  G  jC ,

(9)

where G is the conductance, C is the capacitance and  is the angular frequency [31]. In impedance spectroscopy, measurements are performed at a number of frequency points and the obtained spectra are fitted with an a priori model. The measured dielectric spectra of the skin can be represented by a function of the frequency. A physical interpretation can be associated with the components of this fitting function. The choice of such a function depends on the width of the frequency spectra that is used and the number of its measured points. The Cole-Cole function, although not appropriate for a description over a broad frequency range, has been used in several studies to characterise biological material [32-34]. The use of the Cole-Cole function also has the advantage of having a reasonable number of free parameters, as shown in equation (10) (unlike the Havriliak Negami or the Raicu functions [35,36]), parameters that need to be established from the limited number of measured frequencies in the spectra:

 * ( )    

   dc 1 1  ( j ) j 0 ,

(10)

Eq. (10) describes the complex dielectric spectrum with   being the permittivity at the limit of infinite frequency,  - the amplitude of the dispersion,  - the time constant of the dispersion (i.e. the centre of the dispersion occurs at   1 , or at f  1 / 2 ),  - a numerical exponent (this can be seen as a measure of a distribution of the relaxation times [31]; if zero, the dispersion is the one of Debye), dc - the direct current conductivity and  0 the dielectric constant of vacuum. From the measured admittances (9) we calculate the intensive quantity, i.e. the complex dielectric relative permittivity, with the help of geometrical factors obtained on reference measurements. The quantities of the fit function are obtained by separable least squares, i.e. the parameters , , and  are determined by a nonlinear optimisation whose cost function is the least square error of a linear search on the parameters   and dc. The spectra measured with the three capacitive electrodes are processed independently. It is then possible to relate the changes of these dielectric parameters to the changes in the blood content within the volume measured with the optical sensor, as described in the previous section. In this way we can establish the quantitative link between the blood content and the skin dielectric properties. This relationship will consequently allow for compensating the effects of the blood perfusion variations in the measured dielectric signal.

8

Optical and Dielectric Skin Properties

3

Methods

3.1

Multisensor system

a)

b)

Figure 2 a) Combination of the optical DRS sensors and electrodes of the dielectric sensor on the substrate of the Solianis multisensor system. b) Attachment of the multisensor to the upper arm with a flexible band.

Figure 3 Schematic drawing of the DRS sensor. Light emitted by the LEDs located in the central compartment illuminates the skin through the glass window. Part of light diffusely reflected by the skin is detected by two signal photo-diodes (PD) located equidistantly from the LED in symmetrical compartments. Part of emitted light is reflected by the sensor frame and detected by two monitoring photo-diodes located in the same compartment.

The dielectric and optical properties of the skin were measured with the multisensor glucose monitoring system (MGMS) developed by Solianis Monitoring AG. It includes dielectric capacitive fringing field sensors with the capability to achieve different penetration depths of the electromagnetic field into the various tissue layers by utilising three electrodes with different characteristic geometries. For convenience, the electrodes associated with the deep, middle and shallow penetration range of the electromagnetic (EM) field are referred to as the long, middle and short electrode, respectively (see Figure 2a). The characteristic distances d to the ground plate are 0.3, 1.5 and 4 mm for the short, middle and long electrodes respectively (see Figure 2). The length of the three electrodes of 23 mm was the same. The magnitude and phase of the impedance is measured at a set of 16 different frequencies spanning from 0.1 to 100 MHz similarly for each of the 3 electrodes. Figure 2(a) shows two identical optical DRS sensors embedded within the long dielectric electrode which allows to perform optical characterisation of the skin within the volume probed by the long electrodes of the dielectric sensor. Each DRS sensor features 3 LEDs located closely to each other with the following optical wavelengths: green (568 nm), red (660 nm) and infrared (798 nm). Light reflected back from the skin is detected by two symmetrical PiN-diodes acting as a photo-detectors (signal diodes) separated from the LEDs

9

Optical and Dielectric Skin Properties

by the distance of 1.7 mm as shown in Figure 3. These source-detector distances have been shown to increase the sensitivity to variation in cutaneous blood content while limiting the impact of the blood flow in underlying tissues [19]. The variations of emitted LEDs intensity are monitored by two reference PiN-diodes (monitoring diodes) located in the proximity of the LEDs. The LEDs are switched on sequentially one at the time and after a stabilisation time of 20 ms the intensity is measured with the signal PiN-diodes for a period of 50 ms followed by the measurements of monitoring PiN-diodes. At the end of each measurement cycle the intensity of ambient light (dark channel) is determined by switching off all LEDs and performing measurements with the signal and monitoring diodes. While the measurements are performed on all 3 wavelengths, for the current study only the measurements on isosbetic points in the green and infrared channels have been used. The skin and housing temperatures are also measured in the multisensor system with embedded thermometers. MGMS sensor measurements are performed sequentially with a periodicity of 20 seconds. Data collected by the MGMS is immediately transferred via a Bluetooth communication protocol to a personal digital assistant (PDA) computer that is used for a basic visualisation and storage of the data. The same PDA is also used by the study clinical staff to record text comments related to the progress of the experimental steps in a log. At the end of the study day, all the collected study data is transferred from the PDA to a personal computer (PC) using a USB connection for further analysis and long-term data storage. 3.2 Study performance The perfusion challenging experiments were performed by 4 patients (2 males and 2 females) participating in an experimental clinical study. The demographical data for the patients is shown in Table 1. Table 1 Demographic data of the patients involved in the study.

Subject

I II III IV Mean Std

Age [years]

Heig ht [m]

53 41 52 23 42.3 13.9

1.78 1.73 1.63 1.68 1.71 0.06

Weight [kg] 80 77 61 67 71.3 8.8

BMI [kg/m2] 25.2 25.6 23.0 23.7 24.38 1.23

DM Onset [years] 39 30 38 5 28.0 15.9

HbA 1c [%] 7.8 6.3 7.7 9.3 7.78 1.23

Sex

male male female female

The study was performed in University Hospital in Zurich in accordance with Good Clinical Practice and the Declaration of Helsinki. All patients signed an informed consent agreement, performed an initial screening visit and were then enrolled in the study according to inclusion and exclusion criteria set in the clinical trial study protocol. Patients arrived at the clinical study unit in the morning of the study day. The multisensor was attached to the lateral side of the right upper arm with the flexible band. They remained in a supine position for at least 8 hours as required by the experimental procedure before the start of the perfusion challenging experiments. This time is sufficient for the adaptation of the measured skin volume to the presence of sensor since is much larger than the typical equilibration time of 75 minutes reported for such type of device [12]. At the end of the study visit, patients performed the perfusion changing exercises according to the protocol presented in Table 2. Each patient completed 4 individual runs in total on separate study days.

Optical and Dielectric Skin Properties

10

Table 2 Protocol of the haemodynamic challenges

Step

Name

1

Resting

2

Arm down

3

Arm up

4 5

Resting Bending

6

Resting

Description

Duration (minutes) The right arm is placed parallel to the body 2 resting on the bed. The arm is straight, no bending at the elbow The subject is lowering the arm vertically. The 2 arm is straight. The subject is raising the arm vertically. The 2 arm is straight. Same as Step 1 2 The arm is bent at the elbow, while still in the 2 horizontal position. The upper arm is placed parallel to the body, the forearm is placed on the abdomen in a bent position. Same as Step 1 2

All the activities are performed by the subject with the right arm, while the left arm remains on the bed parallel to the body. Subjects were asked to sequentially switch from one activity to the next at the beginning of the corresponding step according to the timing in Table 2. The exact durations of the experimental steps have been recorded by the assisting clinical staff with the PDA software in the time log and further used in the data analysis to confirm the timings of the orthostatic manoeuvres. One study day had to be excluded from the analysis, since the middle electrode was seen in the data to have clearly detached from the skin due to the excessive movements of a subject. Thus the data from 15 days have been used for this analysis. 3.3 Data analysis The data analysis has been automated and performed using procedures programmed with Matlab (Mathworks, Inc. USA). The timings of the separate experimental steps are determined using the manually written time log. The data collected in 20 seconds intervals at the beginning and at the end of every step was discarded from the analysis to avoid the influence of possible distortions caused by the rapid movements. 3.3.1 Optical data preparation. The signals obtained by photodiodes are processed in several steps in order to compensate for different factors that can affect the wearable system. First, the dark background is subtracted from the intensities of signal and monitoring photodiodes. Next the measurements of signal photodiodes are normalised with the monitoring photodiode in order to account for variations in the intensity emitted by LEDs. Normalised signals are further converted to attenuation using the calibration measurements from a scattering standard (Teflon block) and the averages for two optical sensors are obtained. Attenuation values within each step of the experiment are averaged to obtain a single estimate for each step of the procedure. Measurements of different wavelengths are processed independently. The signals at resting stages (steps 1, 4 and 6) are used in the analysis as the baseline levels and thus signal change is assumed to be zero ( At1   At4   At6   0 , where ti denotes the time point of step i ). For activity steps 2, 3 and 5 the signal changes are obtained as difference between the signal value and linear interpolation of values in the baseline steps. The combination of several baseline measurements allows to account for the drifts in the measurement signals which are not related to the exercise-induced changes in the haemodynamics: such as drifts in tissue glucose concentration, moisture, water distribution in the skin, skin morphology.

Optical and Dielectric Skin Properties

11

The attenuation coefficients are calculated using the haemoglobin absorption spectra provided by S. Prahl [37]. For the scattering properties of RBCs we have taken the results of Roggan et al., who reported scattering coefficient  s 633 =77.3 mm-1 and scattering anisotropy g 633 =0.994 for blood with Ht =0.1 [38]. We used the empirical relation

 s  1.7 to estimate scattering on the wavelengths of interest and normalised them to Ht =1, since we are interested in the optical properties of the RBC volume. The obtained effective attenuation coefficients are  eff 568  = 92.77 mm-1 and  eff 798  = 3.32 mm-1. The attenuation changes are further converted to the pathlength changes and the relative PF is calculated for the infrared light using a linear regression of pathlengths according to Eq. (7). The PF is thus averaged over all study days. Using the obtained value the attenuations of green and infrared light are converted to the changes of the optical thickness of the RBC layer LRBC during the exercises using Eq. (5). 3.3.2 Dielectric data preparation. The spectra of G and C are estimated from the spectra of measured impedance amplitude Z and phase  using Eq. (8). In order to reduce the effect of electrode polarisation and divergence of the skin measurements from the Cole-Cole model [39], we limited the spectral fit to the MHz region (1-100 MHz). The dielectric spectra are fitted with the Eq. 10 individually for the 3 electrodes. Spectra acquired on each measurement cycle are analysed independently. In this way we obtain the time series of the fitted Cole-Cole parameters a ,  ,  dc ,   and  . Baseline values estimated at the resting stages are averaged and used for the evaluation of the quality of fits. Changes from the baseline are calculated with a linear interpolation in the same way as used for the calculation of the optical attenuation. Additionally, the response of the bulk dielectric spectra G  and C  are also analysed in the same manner. Furthermore, we investigated the relation of changes in dielectric parameters (both fitted Cole-Cole parameters and the original C and G) with changes in the RBC layer thickness. The thickness was measured with the optical sensor using a linear regression analysis.

12

Optical and Dielectric Skin Properties

4

Results and discussion Pathlength factor 0.03

0.15

0.02  A(798) [AU]

0.1  A(568) [AU]

10

0.05 0 -0.05 -0.1

 (798)  LRBC [m]

4.1

0.01 0 -0.01

5

0

-5

-0.02

-0.15 Down

Up

(a)

Bend

Down

Up

(b)

Bend

-10 -2

-1

0

1

2

 (568)  LRBC [m]

(c)

Figure 4 Changes in the optical attenuations of a) green (568 nm) and b) infrared (798 nm) light caused by the orthostatic manoeuvres and arm bending. c) Cross plot of the changes of the pathlengths estimated from data of plots a and b. Each symbol shape represents a different patient, colour and interior represents the procedure: red open symbols corresponds to the lowering arm, green crossed symbols – raising arm, blue filled – bending arm.

In Figure 4 one can see that lowering the arm below the heart level leads to an increase of the attenuation of both green and infrared light, while rising and bending the arm leads to the attenuation decrease. This indicates an increase in the blood content within the skin in the former case and a decrease in the latter. The attenuation changes can be converted to the pathlengths within the RBCs according to Eq. (6) and the results are shown in Figure 4c as a cross-plot of the changes in the green and infrared pathlengths. From these results we can estimate the ratio  798/  568 or equivalently  798 if we set the multiplier of green light path as a baseline  568  1 . Using the linear regression we have obtained  798 =3.85 with the R 2 =0.817 for the current data. The pathlengths of green and infrared light in the upper vascular plexus are relatively close, since it can be reached by both wavelengths in the similar way. On the contrary, the deeper vascular plexus can be sampled with the infrared but not with green light, for which the pathlength is close to zero, according to the simulation results for this source-detector separation. Thus the DVP accounts for the difference in pathlengths of green and infrared light in the vascular plexus. Using the obtained results one can combine the measurements of skin attenuation on the two wavelengths in the Eq. (7) in order to increase the specificity to blood perfusion changes.

13

Optical and Dielectric Skin Properties

4.2

Changes in the optical thickness of RBC layer

1.5

 LRBC [ m]

1 0.5 0 -0.5 -1 -1.5 Down

Up

Bend

Figure 5 Boxplot of the changes in the optical thickness of RBC layer during the haemodynamic challenges.

According to the theory described above we can quantify the change in the skin blood content as an optical thickness of the RBC layer using Eq. (5). The statistical distribution of the results is shown in Figure 5. As one can see the arm lowering procedure has the most pronounced effect on the blood layer under the sensor with the median of LRBC of 0.64 μm. The observation of a strong effect of the arm lowering qualitatively agrees with the results of Maeda et al. [23] who reported a statistically significant increase in the light absorption in the finger detected by the transmission photoplethysmograph during the arm lowering by 40 cm below the heart level. The median change during the arm raising equals -0.05 μm. However, this effect is less reproducible if compared to the arm lowering procedure (see in Figure 5).Similarly Maeda et al. have reported on a decrease in amplitude of the light absorption that is comparable to the results found in this work associated with the arm lowering procedure. While the exact reason for this decrease in amplitude observed in our experiments is not fully understood, we can suggest, that the differences in the tissue compartments accessed in our and Meada experiments might play a role. While Maeda et al. measure in the whole volume of the finger, our sensor only probes the cutaneous volume on the upper arm. Similarly, Yamamoto and Oberg recorded only minor variations in the skin blood flow during the arm position changes using the LDF while the volumetric impedance and strain-gauge plethysmography have detected significant changes [40]. Pfützner et al. have observed the increase of the blood flow measured with LDF and no systematic change of the skin impedance during an arm elevation procedure [41], which also agrees with our findings. We can speculate that the arm lowering procedure has the comparable effect on the cutaneous and volumetric circulation, whilst the arm raising procedure affects the cutaneous blood flow to the smaller extent than the circulation in the tissue volume. Bending of the arm in the horizontal position produces a slight decrease in the blood content (median LRBC =-0.18 μm) which can be related to the additional pressure exerted by the device on the skin. This increased pressure leads to the increase of the transmural pressure and results in the decreased blood flow [42]. However, the response to these exercises has a high variability between the experiments which can be attributed to the differences in applied

14

Optical and Dielectric Skin Properties

extra pressure to do the variations in the arm circumference, anatomical peculiarities of the subjects and length of the MGMS band. 4.3

Baseline Cole-Cole parameters 1400 0.32

a)

1200

b)

0.3 1000

 [S/m]

0.28

'

800 600

0.26 0.24

400

0.22

200

0.2

0 6 10

7

10

8

10

0.18 6 10

frequency [Hz]

7

10

8

10

frequency [Hz]

Figure 6: Example of relative dielectric parameter (a) and conductivity (b) spectra for the long electrode: measurement (symbols) and fit (solid line). For both presented fits the R2 is above 0.99.

The validity of the fitting procedure of the dielectric spectra can be tested with an analysis of the measurements in the resting stage of the experiments. Figure 6 presents a typical example of the spectra obtained from the skin in one of the study days and the corresponding fitted function obtained using Eq. (10). As one can see, the model described by Eq. (10) appears to adequately describe the measured spectra with a high accuracy and thus we can conclude that the choice of the function to fit is justified and we expect to obtain reliable parameters for the further analysis. Table 3 Baseline dielectric properties obtained from the fit of long electrode averaged over all resting periods of 4 patients (319 spectra).

Electrode Long

d [mm] 4.0

 0.10±0.01

 [ns]

290±30

dc [mS/m] 0.16±0.01





28.9±2.4

4700±800

The Table 3 shows the mean and the standard deviation of the fitted parameters during the baseline period of the procedure calculated for the measurements of all the patients during the three resting periods (see Table 2), i.e. 319 spectra. As previously described, skin consists of the separate layers with different dielectric properties. On the other hand the characteristic size d of the electrodes defines the penetration depth of the EMF where the long electrode measures at the deepest layers. This explains differences in the dielectric properties measured with different electrodes [43]. We briefly compare our fit parameters with values found in the literature [33] for a sensor with 0.85 mm electrode separation and [10,44] for a sensor with at least 7 mm electrode separation. Our time constant (290 ns) lies between the time constants of two dispersions at 80 and 1600 ns, as measured by Gabriel. The one from Tamura is even smaller, 37 ns. Our exponent (0.1) lies between two values obtained from Gabriel, 0.2 for dry skin and 0 for so called “wet skin”. This indicates that the exponent decreases when skin becomes hydrated and skin becomes more homogeneous from dielectric point of view. Our conductivity of the direct current dc is, as for both groups, in the order of mS/m. Tamura obtained a permittivity at high frequency   of 24.5 and amplitude of dispersion  of 860. The amplitudes of the dispersions of Gabriel are around 30000. The differences between the reference values of

15

Optical and Dielectric Skin Properties

these two last quantities with our quantities are also due to the fact that we do not observe strictly the same dispersions (due to the differences of time constants as said above). The differences with the Gabriel results as point measurement can be explained as follows. In our experiment, the sensors were attached to the arm several hours before the experiments. Natural accumulation of sweat under the sensor and the skin occlusion take place, which leads to the skin hydration. This effect leads to exponent values which lay between the estimates for wet and dry skin. 4.4

Relation of optical and dielectric changes 0.6 0.2 0.4 0.2

 C [pF]

 G [mS]

0.1

0

0 -0.2

-0.1

-0.4 -0.2 -0.6 -2

-1

0

1

2

-2

-1

 LRBC [m]

0

1

2

 LRBC [m]

Figure 7 Changes in dielectric properties (conductance and capacitance) measured with long electrode at frequency 15 MHz in relation to changes in the optical thickness of the RBC layer. Coding of symbols is the same as in Figure 4.

We can further link the dielectric and optical responses to the exercises. Figure 7 illustrates the relations of dielectric changes and the optical thickness of the RBC layer. One can observe the pronounced correlation of the optical properties and the measured skin admittance. The blood-induced changes in dielectric properties at 15 MHz, which is in the upper end of the Beta dispersion (interfacial polarisation at cellular membranes), can be compared with effect of glucose induced changes of the measurements of the multi sensor system reported by Caduff et. al. [12]. Comparison shows that if the blood content variations due to orthostatic manoeuvres are left unaccounted in the glucose estimation algorithm they can lead to the errors of 0.45 and 1.54 mmol/L in conductance and capacitance based single-sensor measurements, correspondingly. This agrees with previous findings [11] and original assumption that the monitoring of the CBC is necessary for the accurate non-invasive estimation of the glucose content. Table 4 Coefficient of a linear regression and coefficient of determination R 2 of changes in dielectric measurements at a frequency of 15 MHz on changes in the optical thickness of the RBC layer (the long electrode data presented in Figure 7)

Electrode

Short Middle Long

G  Coef [mS/m] 1.4·10-3 40.6·10-3 90.7·10-3

C  R2 0.00 0.33 0.66

Coef [pF/m] 0.065 0.276 0.260

R2  0.01 0.63 0.76

16

Optical and Dielectric Skin Properties

In general we observe the higher sensitivity of the longer electrodes which is shown in Table 4, where the R 2 increases with an increase in the electrodes separation, reaching a maximum for the long electrode with an R 2 =0.66 for G and R 2 =0.76 for C . The same is true for the regression coefficients – the response is larger for the electrodes with a larger separation. As expected, electrodes with the largest separation designed to probe the deeper skin layers have highest sensitivity to the perfusion changes as observed both in the fit quality and absolute coefficients for conductance and capacitance. The short electrode characterises the most superficial skin layer – epidermis which does not have the vasculature and thus is not affected by the blood content variations induced by the exercises, as we can see in Table 4. Long electrode probes the larger skin volume, which includes part of the dermis and its’ vasculature and thus demonstrates statistically significant correlations of the measured dielectric parameters with the blood content. Middle electrode shows intermediate sensitivity. This demonstrates that such system can perform the depth profiling of the skin layers.

6

(a)

R2=0.678

2

  [ns]

   103

1 0.5 0 -0.5

0 -2

-1

(c)

R2=0.359

1 0 -1 -2

R2=0.398 -1

0

1

-6 -2

2

-1

 LRBC [m] 0.2

2

-4

-1.5 -2

(b)

4

  dc [ S/m]

1.5

0

1

2

-2

-1

 LRBC [m]

(d)

R2=0.451

30

0

1

2

 LRBC [m]

(e)

R2=0.117

20 0.1



 

10 0

0 -10

-0.1 -20 -0.2 -2

-1

0

1

2

-30 -2

 LRBC [m]

-1

0

1

2

 LRBC [m]

Figure 8: Changes of the fitted Cole-Cole parameters a)  , b)  , c)  dc , d)   , and e)  versus the changes on the optical thickness of the RBC layer. The markers correspond to the four patients of the study. Coding of symbols is the same as in Figure 4. Table 5: Coefficient of determination R2 and coefficient of linear regression of changes in dielectric properties on changes of the optical thickness of RBC layer for 3 electrodes Electrod e Short Middle Long

 Coef [μm-1] 0.13·10-3 0.52·10-3 0.60·10-3

 R2 0.04 0.59 0.68

Coef [ns/μm] 0.00 -0.43 -1.35

 dc R2 0.00 0.13 0.40

Coef [Ω/μm] -0.02 -0.03 0.66

  

  R2 0.05 0.01 0.35

Coef [pF/μm] -0.032 -0.023 0.056

R2 0.06 0.07 0.45

Coef [pF/μm] 0.63 0.06 13.6

R2 0.00 0.00 0.12

Optical and Dielectric Skin Properties

17

Changes of the fitted dielectric parameters are presented in Figure 8 for the long electrode and the results of a linear regression fits to the LRBC are presented in Table 5 for all 3 electrodes. Here one can generally observe that measurements with the larger electrodes produce a higher correlation with the RBC layer thickness for most of the Cole-Cole parameters, which are also comparable to the response of the C and G discussed above. The highest correlation is observed for the exponent  , with R2=0.59 for middle and R2=0.68 for the long electrodes. As it was mentioned above the exponent characterises the width of the dispersion, which reflects heterogeneity of the measured tissue. Thus when blood content increases in the tissue, we observe a broader dispersion which indicates higher heterogeneity of skin dielectric properties. This also confirms the in vitro observation of Fomekong [45] who reported the dispersion width increase with an increase in the volume concentration of RBCs (haematocrit). He attributed this increase to the increased interactions of RBCs. Other authors highlighted the practical importance of the  -parameter for the characterisation of biological tissues [31,46,47]. Water is the major component of skin and occupies around 70% by mass in the epidermis [48,49] while the bulk blood content of the skin is relatively smaller under normal physiological conditions [3]. Nevertheless, due to the high variability in the skin blood content and its’ strong conductive and capacitive properties ,of blood can influence the resulting dielectric properties of the skin and thus affect the Cole-Cole parameters. Blood has a pronounced dielectric dispersion (Beta dispersion, interfacial polarisation) in the kHz-MHz spectral range with a time constant of around 130 ns [9,10] or 265 ns in [50]. The time constant of 290 ns that we obtain for the skin for a baseline measurements (Table 3) is larger than the value for blood. With an increase of the blood content in the skin we observe a negative change of  as seen in the Figure 8 which shifts the dispersion towards the frequencies of the pure blood. The positive slope of the conductivity dc coincides with the expectation that an increase in the amount of conductive blood in the skin leads to the increase of the measured effective conductivity of the measured tissue volume. A similar argumentation can be derived for the positive slope of the permittivity at the limit of infinite frequency   . Since the blood (due to its water content) has a larger high frequency permittivity than the dry material,   will increasing with an increase of the blood amount. Note that dc and   have similar behaviour (same sign of slope) as the measured bulk conductance and capacitance (Figure 8). For the , which is proportional to the relaxation strength, the coefficient of determination is too small even for the long electrode (R2=0.11) to allow for the qualitative discussion.. 5

Conclusions We have repetitively investigated the response of the dielectric and optical sensors of the Solianis Monitoring wearable multisensor system during local blood perfusion challenging experiments on a population of 4 patients with diabetes mellitus type 1. We have observed that the highest impact on the measured signals both optical and dielectric is produced by the arm lowering below the heart level. Thus orthosatic manoeuvre produced changes as measured by long electrode, which were equal in magnitude to the change produced by a glucose decrease of up to 1.54 mmol/L. This demonstrates that the blood content alterations are a significant factor for NIGM based on the dielectric characterisation of skin and underlying tissue. Thus it becomes apparent that CBC has to be monitored in order to allow for compensation of such effects in the dielectric signal, which provides a further support to the multisensor concept of the NIGM. Bending of the arm and the resulting additional pressure applied by the sensor to the skin results in smaller effect of the opposite sign. The effect of arm rising above the heart level is less reproducible which we attribute to the different cutaneous blood regulatory mechanisms involved in this case.

Optical and Dielectric Skin Properties

18

The relation of the optical pathlengths for probing wavelengths required for the quantitative estimation of the haemoglobin and RBC content using a double-wavelength system [19] has been obtained from the experimental data. The pathlength of 798 nm infrared light in the blood-holding skin layers has been found to be 3.85 times larger than the same for green 568 nm light. The effect of the blood relocation on the dielectric properties has been investigated for the measured conductance and capacitance as well as for the fitted parameters of Cole-Cole model. We have found that the dielectric response to the blood perfusion changes is higher for the electrodes with larger geometrical separation both in absolute scale and in determination coefficient of the linear fit on the change of optical thickness of RBCs layer LRBC . This can be explained by the higher probing depths of the EM field for electrodes with large separation which are able to reach the dermal vascular plexus. The highest correlation was observed for the long electrode with R 2 =0.65 for G and R 2 =0.76 for C . Among the fitted Cole-Cole parameters, the highest sensitivity was demonstrated for the exponent parameter  ( R 2 =0.65) which suggests that an increase in the blood content in the skin leads to the increase of dielectric heterogeneity. At the same time a negative correlation of  ( R 2 =0.40) indicates that the measured skin dispersion shifts towards the blood dispersion with an increase of the blood content in the skin. Similar argument can be made for the increase of the permittivity at high frequency ( R 2 =0.45), which is proportional to the amount of water in the tissue. High correlation of the skin dielectric and optical properties to the local cutaneous perfusion challenges indicates that optical measurements using the DRS sensor can potentially be used for the compensation of dielectric spectra measured from the skin for the variations in blood perfusion under the real-life conditions. The combination of the optical and dielectric measurements in the discussed arrangement can also be used for the other applications, such as characterising parameters of blood and erythrocytes, accessing the vascular function. The separate access to haemoglobin content, concentration of red blood cells and parameters of dielectric dispersion provided by this system can help to develop applications targeting several physiological properties of general medical interest. 6

Acknowledgements Authors wish to thank Prof. Y. Feldman for the helpful comments on the manuscript and the whole Solianis Monitoring team for their work in providing and preparing the devices and data from the experimental clinical study.

Optical and Dielectric Skin Properties

7

19

References

[1]

V.V. Tuchin, ed., Handbook of Optical Sensing of Glucose in Biological Fluids and Tissues (Series in Medical Physics and Biomedical Engineering), Taylor & Francis, 2008.

[2]

O.S. Khalil, “Non-invasive glucose measurement technologies: an update from 1999 to the dawn of the new millennium.,” Diabetes Technol Ther, vol. 6, Oct. 2004, pp. 660-697.

[3]

A. Caduff, M.S. Talary, and P. Zakharov, “Cutaneous blood perfusion as a perturbing factor for noninvasive glucose monitoring.,” Diabetes technology & therapeutics, vol. 12, Jan. 2010, pp. 1-9.

[4]

V.C. Scanlon and T. Sanders, Essentials of Anatomy and Physiology, Philadelphia: F A Davis Company, 2007.

[5]

A. Rhoades and D.R. Bell, Medical Physiology: Principles for Clinical Medicine, 2008.

[6]

A. Mehling and J.W. Fluhr, “Chronobiology: biological clocks and rhythms of the skin.,” Skin pharmacology and physiology, vol. 19, Jan. 2006, pp. 182-9.

[7]

M.B. Beckett, J.A. Hodgdon, W. Woods, and B. Appleton, Skin blood flow and bioelectrical impedance, SAN DIEGO CA: 1994.

[8]

R. Pethig and D.B. Kell, “The passive electrical properties of biological systems: their significance in physiology, biophysics and biotechnology.,” Phys Med Biol, vol. 32, Aug. 1987, pp. 933-970.

[9]

C. Gabriel, S. Gabriel, and E. Corthout, “The dielectric properties of biological tissues: I. Literature survey.,” Phys Med Biol, vol. 41, Nov. 1996, pp. 2231-2249.

[10]

S. Gabriel, R.W. Lau, and C. Gabriel, “The dielectric properties of biological tissues: II. Measurements in the frequency range 10 Hz to 20 GHz.,” Phys Med Biol, vol. 41, Nov. 1996, pp. 2251-2269.

[11]

M.S. Talary, F. Dewarrat, D. Huber, and A. Caduff, “In vivo life sign application of dielectric spectroscopy and non-invasive glucose monitoring,” Journal of NonCrystalline Solids, vol. 353, Dec. 2007, pp. 4515-4517.

[12]

A. Caduff, M.S. Talary, M. Mueller, F. Dewarrat, M. Donath, and W.A. Stahel, “Noninvasive glucose monitoring in patients with Type 1 diabetes: a Multisensor system combining sensors for dielectric and optical characterisation of skin.,” Biosensors and Bioelectronics, vol. 24, May. 2009, pp. 2778-84.

[13]

G.E. Odland, “Structure of the skin,” Physiology, biochemistry, and molecular biology of the skin, L.A. GOLDSMITH, ed., Oxford: Oxford University Press, 1991, pp. 3 62.

Optical and Dielectric Skin Properties

20

[14]

I.M. Braverman, “Anatomy and Physiology of the Cutaneous Microcirculation,” Bioengineering of the skin: cutaneous blood flow and erythema, E. Berardesca, P. Elsner, and H.I. Maibach, eds., CRC Press, 1995, pp. 3 - 22.

[15]

T.J. Ryan, “Cutaneous circulation,” Physiology, biochemistry, and molecular biology of the skin, Oxford: Oxford University Press, 1991, pp. 1019-1084.

[16]

J.A.A. Hunter, J.A. Savin, and M.V. Dahl, Clinical Dermatology, Blackwell Science, 2002.

[17]

J.W. McMurdy, G.D. Jay, S. Suner, and G. Crawford, “Noninvasive optical, electrical, and acoustic methods of total hemoglobin determination.,” Clin Chem, vol. 54, Feb. 2008, pp. 264-272.

[18]

A. Lima and J. Bakker, “Noninvasive monitoring of peripheral perfusion.,” Intensive Care Med, vol. 31, Oct. 2005, pp. 1316-1326.

[19]

P. Zakharov, M.S. Talary, and A. Caduff, “A wearable diffuse reflectance sensor for continuous monitoring of cutaneous blood content.,” Phys Med Biol, vol. 54, Sep. 2009, pp. 5301-5320.

[20]

M.Y. Jaffrin and C. Fournier, “Comparison of optical, electrical, and centrifugation techniques for haematocrit monitoring of dialysed patients.,” Med Biol Eng Comput, vol. 37, Jul. 1999, pp. 433-439.

[21]

M. Engelhart and J.K. Kristensen, “Evaluation of cutaneous blood flow responses by 133Xenon washout and a laser-Doppler flowmeter.,” J Invest Dermatol, vol. 80, Jan. 1983, pp. 12-15.

[22]

J.K. Kristensen, “Local regulation of digital blood flow in generalized scleroderma.,” J Invest Dermatol, vol. 72, May. 1979, pp. 235-240.

[23]

J. Maeda, T. Takemiya, M. Shimoda, J. Suzuki, and H. Komiya, “Local Vascular Response Induced by Arm Position Change: The Related Myogenic Mechanism and Veno-arteriolar Reflex in Humans,” Abv. Exerc. Sports Physiol., vol. 3, 1997, pp. 6773.

[24]

D.T. Delpy, M. Cope, P. van Der Zee, S. Arridge, S. Wray, and J. Wyatt, “Estimation of optical pathlength through tissue from direct time of flight measurement.,” Phys Med Biol, vol. 33, Dec. 1988, pp. 1433-1442.

[25]

M. Shimada, Y. Yamada, M. Itoh, and T. Yatagai, “Melanin and blood concentration in human skin studied by multiple regression analysis: experiments.,” Phys Med Biol, vol. 46, Sep. 2001, pp. 2385-2395.

[26]

I.V. Meglinski and S.J. Matcher, “Computer simulation of the skin reflectance spectra.,” Comput Methods Programs Biomed, vol. 70, Feb. 2003, pp. 179-186.

[27]

M. Hiraoka, M. Firbank, M. Essenpreis, M. Cope, S.R. Arridge, P. van der Zee, and D.T. Delpy, “A Monte Carlo investigation of optical pathlength in inhomogeneous tissue and its application to near-infrared spectroscopy.,” Phys Med Biol, vol. 38, Dec. 1993, pp. 1859-1876.

Optical and Dielectric Skin Properties

21

[28]

V.V. Tuchin, Tissue optics: light scattering methods and instruments for medical diagnosis, SPIE, Bellingham, USA, 2007.

[29]

B. Chance, M. Cope, E. Gratton, N. Ramanujam, and B. Tromberg, “Phase measurement of light absorption and scatter in human tissue,” Review of Scientific Instruments, vol. 69, 1998, p. 3457.

[30]

S.J. Matcher, M. Cope, and D.T. Delpy, “Use of the water absorption spectrum to quantify tissue chromophore concentration changes in near-infrared spectroscopy.,” Physics in medicine and biology, vol. 39, Jan. 1994, pp. 177-96.

[31]

K.R. Foster and H.P. Schwan, “Dielectric properties of tissues and biological materials: a critical review.,” Critical reviews in biomedical engineering, vol. 17, Jan. 1989, pp. 25-104.

[32]

S. Gabriel, R.W. Lau, and C. Gabriel, “The dielectric properties of biological tissues: III. Parametric models for the dielectric spectrum of tissues.,” Physics in medicine and biology, vol. 41, Nov. 1996, pp. 2271-93.

[33]

T. Tamura, M. Tenhunen, T. Lahtinen, and T. Repo, “Modelling of the dielectric properties of normal and irradiated skin,” Physics in Medicine and , 1994.

[34]

T. Goto, M. Hashimoto, N. Shinyashiki, S. Yagihara, and Y. Hayashi, “Dielectric study on distribution of water in human skin,” Transactions of the Materials Research Society of Japan, vol. 31, 2006, pp. 771-774.

[35]

K. Asami, “Characterization of heterogeneous systems by dielectric spectroscopy,” Progress in Polymer Science, vol. 27, 2002, pp. 1617-1659.

[36]

V. Raicu, “Dielectric dispersion of biological matter: model combining Debye-type and "universal" responses.,” Physical review. E, Statistical physics, plasmas, fluids, and related interdisciplinary topics, vol. 60, Oct. 1999, pp. 4677-80.

[37]

S.A. Prahl, “Tabulated molar extinction coefficient for hemoglobin in water,” 1998.

[38]

A. Roggan, M. Friebel, K. Dörschel, A. Hahn, and G. Müller, “Optical Properties of Circulating Human Blood in the Wavelength Range 400--2500 nm,” Journal of Biomedical Optics, vol. 4, 1999, pp. 36-46.

[39]

V. Raicu, N. Kitagawa, and A. Irimajiri, “A quantitative approach to the dielectric properties of the skin.,” Physics in medicine and biology, vol. 45, Feb. 2000, pp. L1-4.

[40]

Y. Yamamoto and P.A. Oberg, “Measurement of digital blood flow using the laser Doppler, impedance and strain-gauge methods.,” Medical & biological engineering & computing, vol. 28, Mar. 1990, pp. 113-8.

[41]

A. Pfützner, A. Caduff, M. Larbig, T. Schrepfer, and T. Forst, “Impact of posture and fixation technique on impedance spectroscopy used for continuous and noninvasive glucose monitoring.,” Diabetes technology & therapeutics, vol. 6, Aug. 2004, pp. 43541.

Optical and Dielectric Skin Properties

22

[42]

S. Yamada, “Effects of positive tissue pressure on blood flow of the finger.,” Journal of applied physiology, vol. 6, Feb. 1954, pp. 495-500.

[43]

T. Lahtinen, J. Nuutinen, and E. Alanen, “Dielectric properties of the skin,” Physics in Medicine and Biology, vol. 42, 1997, pp. 1471-1472.

[44]

C. Gabriel, T.Y. Chan, and E.H. Grant, “Admittance models for open ended coaxial probes and their place in dielectric spectroscopy.,” Physics in medicine and biology, vol. 39, Dec. 1994, pp. 2183-200.

[45]

R. Fomekong, “Passive electrical properties of RBC suspensions: changes due to distribution of relaxation times in dependence on the cell volume fraction and medium conductivity,” Bioelectrochemistry and Bioenergetics, vol. 47, 1998, pp. 81-88.

[46]

T. Yamamoto and Y. Yamamoto, “Analysis for the change of skin impedance.,” Medical & biological engineering & computing, vol. 15, May. 1977, pp. 219-27.

[47]

A. Ivorra, M. Genescà, A. Sola, L. Palacios, R. Villa, G. Hotter, and J. Aguiló, “Bioimpedance dispersion width as a parameter to monitor living tissues.,” Physiological measurement, vol. 26, Apr. 2005, pp. S165-73.

[48]

P.J. Caspers, G.W. Lucassen, E. a Carter, H. a Bruining, and G.J. Puppels, “In vivo confocal Raman microspectroscopy of the skin: noninvasive determination of molecular concentration profiles.,” The Journal of investigative dermatology, vol. 116, Mar. 2001, pp. 434-42.

[49]

R.R. Warner, M.C. Myers, and D.A. Taylor, “Electron probe analysis of human skin: determination of the water concentration profile.,” J Invest Dermatol, vol. 90, Feb. 1988, pp. 218-224.

[50]

T. Chelidze, “Dielectric spectroscopy of blood,” Journal of Non-Crystalline Solids, vol. 305, Jul. 2002, pp. 285-294.