microwave photonics || rf and microwave photonics in biomedical applications

9

RF and Microwave Photonicsin Biomedical Applications

Afshin S. Daryoush

9.1 Introduction

In the last 20 years the field of microwave photonics has evolved due to unique features of

analogue fibre-optic systems and its applications in radio over fibre for telecommunications [1]

and optically controlled phased array antennas [2] for military applications, as has been

discussed in earlier chapters of this book. Recently, microwave photonics techniques have also

been extended to biomedical systems and this chapter presents two distinctive biomedical

imaging applications that employ these techniques. (Optics already lends its application to

laser Doppler anemometry, optical biopsy and optical molecular imaging, and phase micros-

copy.) The first application to be discussed is the design and implementation of optical

hydrophone for calibration of ultrasound transducers for frequencies up to 100MHz,which has

found applications in sub-millimeter wave imaging and therapeutic applications. The second is

the use of broadbandmodulated near infrared (NIR) light waves for quantifying blood flow and

cellular functionality using spectroscopy, which is to be applied to coagulation monitoring and

functional imaging with sub-centimetre spatial resolution using photon density waves. Both

techniques are first discussed in terms of the fundamental physical interaction of lightwaves

with biological tissues and the technical advantages that RF and microwave photonics could

bring to conventional imaging modalities.

9.1.1 Introduction to Optical Hydrophone

Only a decade ago, the highest ultrasound imaging frequencywas of the order of 7MHz. Today,

modern diagnosticmachines, particularly those designed for applications such as dermatology,

ophthalmology and microsurgery, operate at centre frequencies close to 15 or 20MHz [3].

Microwave Photonics: Devices and Applications Edited by Stavros Iezekiel

© 2009 John Wiley & Sons, Ltd. ISBN: 978-0-470-84854-8

Also, a majority of the currently available machines has integrated harmonic imaging

capability that can provide diagnostic images at twice the fundamental frequency and hence

increase the resolution twofold. However, ultrasound probes need to be calibrated in the

measurement bandwidth extending to at least eight times the centre frequency of the imaging

transducer [4, 5]. This frequency limit has been introduced to account for nonlinear propaga-

tion phenomena, which lead to distortion of the pressure–time waveform launched into the

examined tissue. Assuming the centre frequency of the imaging array to be about 12MHz, this

would require the sensitivity of the hydrophone probe to be determined in the vicinity of

100MHz [6] (Figure 9.1).

In order to acquire and reproduce faithfully the pressure–time waveforms, and also to

determine accurately the key acoustic parameters of the characterized acoustic field, the active

element of the probe has to be considerably smaller in comparisonwith the cross sections of the

acoustic beam profiles measured; alternatively, tedious spatial averaging correction models

have to be used [4, 7, 8]. In order to ensure half-wavelength sampling that is needed to eliminate

spatial averaging effects, the pressure-sensitive portion of the probe needs to be about 8 mm to

faithfully reproduce the field at 100MHz (assuming plane waves). Unfortunately, the physical

dimensions of available ultrasound probes typically have diameters of 500 mm [4–6, 9]. The

smallest available probe has a diameter of approximately 80 mm and this is still an order of

magnitude too large for 100MHz measurements.

The development of fibre-optic (FO) probes that are specifically designed to increase

sensitivity to ultrasound waves with a diameter of about 8 microns to minimize the spatial

averaging error up to 100MHz is discussed. In Section 9.2, various principles of sensing

acoustic waves using interaction of acoustic energy with intensity, phase and frequency of

lightwaves in the FOprobes are briefly outlined, and emphasis is given to the intensity detection

technique. Theoretical calculations of the laser source output power and acoustic pressure

amplitude needed to ensure a minimum acceptable signal-to-noise ratio are presented, and an

experimental evaluation of a custom-designed down-tapered gold-coated FO probe, which

meets bandwidths of up to 100MHz, is also explored.

9.1.2 Introduction to Optical Spectroscopy using NIR

Near infrared (NIR) spectroscopy is a new, noninvasive technique to analyse living tissue in

terms of absorption and reduced scattering coefficients, which could provide information about

disease-related functional and structural changes [10–12]. Three main categories of NIR

spectroscopy are time domain, frequency domain and continuous wave (CW) measurement

techniques [13]. The frequency-domain method has attracted interest from the biomedical

Figure 9.1 High-resolution imaging of biological tissues using short ultrasound pulses with a

bandwidth requirement of up to 100MHz

240 Microwave Photonics: Devices and Applications

research field for decades due to its low component cost, ease in separating absorption and

scattering parameters, and potential for real-time imaging.

In frequency-domain photon migration (FDPM) methods, diffuse photon density waves

(PDW)aregeneratedwhen light ismodulatedbyradio frequencysignals,whichpropagateswith

a wavelength of several millimetres to centimetres depending on the modulation frequency.

Amplitudeandphase informationof thediffusephotondensitywavesareused tomap theoptical

absorption and scattering properties of the medium. These optical properties in turn are used to

obtain haemaglobin concentration, blood volume and ‘absolute’ oxygen saturation [14, 15].

There are several advantages to multi-frequency instruments compared with single-fre-

quency instruments. Since most tissues have a layered structure and because photon penetra-

tion depth is less at a higher frequency due to a higher loss (cf. Figure 9.2), by sweeping the

modulation frequency one can have information for all layers in a single measurement. This

approach is very important in clinical measurements, where it is preferable to make a single

measurement and obtain as much information as possible, due to the calibration challenges.

The development of specialized optical systems with modulation capability of up to 3GHz

has also been demonstrated and spectroscopic information is conducted for both solid and

liquid phantoms. Moreover, the accuracy of the broadband extraction process is compared to

the single-frequency extraction for phantom resembling breast tissue, where the results of this

extraction are extended to future clinical imaging.

9.2 Approaches to Fibre-optic Based Acoustic Pressure Sensors

Fibre optics has been used in underwater acoustic sensing for a long time and is now being

extended to biomedical applications. Optical fibres have the following advantages over

Figure 9.2 Concept of diffused photon near-infrared spectroscopy in biological tissue depicting:

(a) optical absorption coefficient of oxygenated and de-oxygenated haemoglobin (Oxy-Hb andDeoxy-Hb)

and water in near-infrared region; (b) impact of modulation frequency on tissue penetration depth and

structure of banana-shaped photon scattering in tissue

RF and Microwave Photonics in Biomedical Applications 241

conventional acoustic sensing/imaging techniques: (i) immunity to electromagnetic interfer-

ence; (ii) small sensing area; (iii) small physical dimensions and low weight; (iv) large

bandwidth,and (v) high resistance to high temperature, corrosion by chemicals and adverse

climatic conditions. This section provides an overview of various fibre-optic sensing schemes

which have been used in acoustic sensing applications.

An electric wave travelling in a medium in the positive z-direction can be expressed by the

basic equation given below:

Eðz; tÞ ¼ E0cosðvt� kzþ u0Þ ð9:1Þwhere E0 is the electric field amplitude in V/m,v is the angular frequency in rad/s, and u0 is theinitial phase in radians.

In fibre-optical sensing, the physical phenomenon being sensed interacts with the fibre and

changes one or more of the above parameters associated with the electromagnetic field in or

around the fibre. Accordingly we can classify acoustic fibre-optic sensors as intensity

modulated sensors, frequency/wavelength modulated sensors and phase modulated sensors.

After a review of the best reported results, the latest ‘hero’ performance results of intensity

detection techniques are discussed by the author in Section 9.3.

9.2.1 Intensity Modulated Sensors

Acoustic pressure induces change in the intensity of light passing through an optical fibre.

These sensors can be reflection type, transmission type or total internal reflection based.

9.2.1.1 Reflection Type

In reflection type sensors, the incident pressure induces a change in the refractive index of the

sensing medium surrounding the fibre (Figure 9.3). This leads to a change in reflectance at the

fibre–water interface. The intensity of reflected light is thus modulated by incident pressure.

Figure 9.3 Reflection type fibre-optic acoustic sensor with intensity detection [16]. Reprinted from J. E.

Parsons, C. A. Cain, J. B. Fowlkes, ‘‘Cost-effective assembly of a basic fiber-optic hydrophone for

measurement of high-amplitude therapeutic ultrasound fields,’’ J. Acoust. Soc. Am., vol. 119, pp.

1432–1440, 2006. (� Copyright 2006, American Institute of Physics)


This change in intensity is detected using a photodetector [16] where typically a responsivity of

�302 re 1VmPa and sensitivity of 0.9MPa has been reported.

9.2.1.2 Transmission Type

Another technique is the transmission type approach [17], where two single mode fibres are

placed tip-to-tip at a very small separation. An acoustic wave incident on the tip causes motion

of the fibres and hence changes the optical coupling efficiency (Figure 9.4). The transmitted

light intensity thus varies in accordance with the pressure. A modification of the above

approach is the schlieren technique [18], in which the region in between the two fibres is

occupied by a grating structure to enhance system sensitivity by monitoring changes in

coupling efficiency induced by diaphragm movement. A suitable sensitivity of these hydro-

phones is demonstrated for deep-sea applications over a frequency range of 100Hz to 1 kHz.

9.2.1.3 Frustrated Total Internal Reflection

Another modification to the above transmission sensor type includes the use of frustrated total

internal reflection at the edge of two polished fibres [19]. The fibres are angle polished and

placed at an angle such that all the modes undergo total internal reflection at the fibre–air–fibre

interface. Any vertical displacement of one of the fibres due to the applied acoustic wave

violates the total internal reflection condition and hence changes the amount of light coupled to

the other fibre. This sensor type has a bandwidth of 500Hz.

Intensity modulation schemes are simple and are not as affected by phase noise as

interferometric schemes. However, they are subject to high losses of optical energy since

most is lost in reflection or coupling loss. Thus the sensitivity of these systems is lower. External

amplification has to be provided in order to boost sensitivity.

9.2.2 Phase Modulated Sensors

The phase of incident light varies with acoustic pressure and can be detected using interfero-

metric techniques. When two coherent light waves which are shifted apart in phase are

Figure 9.4 Schematic of transmission type intensity modulated fibre-optic acoustic sensor [17]


superimposed they undergo constructive and destructive interference to form an interference

pattern. In the interferometric sensing method, two arms of fibre are used and compared. One

arm acts as the reference arm while the other is the sensing arm. Any change in the relative

phase or optical path length of the light from the two arms leads to displacement of the

interference fringes. Michelson�s interferometer, Fabry–Perot interferometers and Mach–-

Zehnder interferometers (Figure 9.5) have been employed in the past to detect acoustic signals.

These phase modulated sensors have been extensively studied and can be classified as either

external or internal interferometric phase sensors.

9.2.2.1 External Interferometric Phase Sensors

Michelson and Mach–Zehnder interferometers have been extensively used as phase sensors.

Application of pressure to the sensing arm leads to changes in the refractive index of the sensing

fibrematerial aswell as changes in the physical dimensions of the fibre. This leads to phase shift

and subsequently fringe displacement. Sensitivity of 92 kPa/fringe displacement has been

reported [20]. This technique is, however, subject to fringe displacement related to tempera-

ture-inducedfibre indexvariation,which is of the order of 10 p.p.m./C in silica basedfibres [21].

This temperature-induced phase change leads to uncertainty in the measured pressure value

and is only useful at frequenciesmuch higher than the rate of thermal fluctuations in the sensing

arm. In such sensors, the fibre in the two arms of the interferometer can be free or wrapped

around a mandrel (cylinder) of suitable material. Previous studies indicate that higher

sensitivity can be obtained by using mandrel fibres [20, 22]. In this case, when a mandrel

of low Young�s modulus is subject to pressure, its length changes thus causing stretching or

Figure 9.5 Acoustic pressure sensor using Mach–Zehnder interferometric detection. Reproduced

from [20]


compression of the fibre wound around it. This leads to changes in refractive index as well as

changes in the optical length of the fibre as mentioned above. The sensitivity of such structures

depends onmandrel geometry and the properties of themandrel material. Due to the low photo-

elastic coefficient of silica, a long length of fibre is needed for improved sensitivity. This,

however, leads to poor frequency response. As a solution, embedded single-mode fibre acoustic

sensors have been suggested [23], where the fibre is moulded on to the mandrel by using

material of appropriate Young�s modulus and coating thickness, as depicted in Figure 9.6. In

this case, sensitivity in the range of �328 to �338 dB re 1V/mPa has been reported over a

frequency range of 0.75–10 kHz.

It can also be noted that the above phase modulation techniques are subject to random phase

fluctuation due to temperature drifts or environmental conditions and hence are limited in

performance by phase noise. These temperature drifts can be reduced by using a push–pull

configuration and tuneable resonant cavity techniques [24], where a feedback control circuit is

employed tomaintain the stability of the operating point on the interference pattern.However, a

large sensing area of the mandrel fibre sensors leads to poor spatial resolution making them

unsuitable for imaging applications.

9.2.2.2 Internal Interferometric Phase Sensors

In this case the interferometer is embedded within the fibre itself. Two distinct approaches are

discussed next.

Fabry–Perot (FP) InterferometersThe FP interferometer is one type of structure used for internal interferemetric phase sensors.

Previous work includes formation of an FP cavity by using 25 mm thick Parylene film deposited

on the edge of a straight-cleaved fibre from which the final 1 mm of jacket has been

removed [25]. Partially reflectingmirrors of the resonant structure consist of reflective coatings

of aluminum, where experimentally system sensitivity in terms of noise equivalent power

(NEP) of 10 kPa over a 25MHz bandwidth is observed. Fabry–Perot cavity resonators

consisting of 10 mm Parylene film surrounded by gold mirrors have been reported [27]. These

structures are fabricated at the tip of straight-cleaved and tapered fibres. The reflectivity can be

Figure 9.6 Schematic of an optical sensor structure wound on the mandrel hydrophone as a pressure

sensor arm of an interferometer. Reproduced from [22]


controlled by the gold-coating thickness. Incident acoustic pressure modulates the optical

length of the Parylene film as well as the reflectivity of the gold creating an interference pattern

(Figure 9.7). The system has a nonflat frequency response of up to 50MHz and extrapolated

sensitivity of 2 kPa.

Multi-layer structuresThese are constructed based on the principle of periodic micro-interferometers as shown in

Figure 9.8 [27]. Two high-reflection subsystems, both consisting of several l/4 layers, are

connected by a central l/2 spacer layer. The sensor is made of 19 dielectric layers with

alternating high–low refractive indices of n¼ 2.3 (Nb2O5) and n¼ 1.48 (SiO2). The sensor

element has an overall thickness of d¼ 1.9 mm. The principle of ultrasound measurement is

based on the elastic deformation of the multi-layer system by an incident acoustic pressure and

the detection of the induced change in optical reflectance, DR. The system is operated at a

maximum sensitivity point (the point where the slope of change of reflectance with pressure is

N+1

(N+1)/2

L L L

0 2 (N+1)/2

Fibre

L L L

nidi 2nidi

I in

Iout

Acoustic pressure

Figure 9.8 Fibre-optic multilayer hydrophone, i¼ 1. . ..N, L¼ low index layer, H¼ high index layer,

p¼ pressure pulse [27]

Figure 9.7 Plane-cleaved fibre-optic sensor based on the principle of the FP interferometric technique.

Reproduced from [26]


highest). The transfer function of this sensor shows a resonant peak at a frequency of 24MHz,

which is attributed to possible diffraction effects.

In intrinsic FP cavity structures, the operating point has to be stabilized at the maximum

slope point on the interference transfer function. Any drift in this operating point due to

temperature or other factors can lead to increased noise or a decrease in sensitivity. Due to

periodicity, the accurate measurement of differential phase change in phase modulated sensors

is difficult and hence such systems are not suited to most applications.

9.2.3 Frequency/Wavelength Modulated Sensors

Frequency or wavelength sensing mechanisms are an extension of phase sensitive fibre-optic

sensors and a few examples are discussed next.

9.2.3.1 External Bragg Cell Sensor

Ameasuring technique based on acoustically induced frequency modulation of light has been

reported in fibre sensors over a frequency range of 100–1200Hz [28]. This arrangement makes

use of an external Bragg cell which shifts the frequency (wavelength) of the laser source by

11MHz. The unshifted frequency component is incident on the sensing fibre which reacts to

acoustically induced strain and change in refractive index. This induces a change in the phase of

the optical signal and results in the frequency modulation of the signal. This frequency

modulated signal from the fibre sensor and frequency shifted component from the Bragg cell

are combined at the frequency discriminator, the output of which is proportional to the incident

acoustic pressure.

9.2.3.2 Fibre Bragg Grating Sensor

The most common type of wavelength modulated sensor makes use of the fibre Bragg grating

(FBG). Detection is based on an acoustically induced change in Bragg wavelength or intensity

modulation of light from the grating.When an acoustic wave is incident on an optical fibrewith

an FBG, both the refractive index and the FBG grating period undergo changes due to elasto-

optic effects (Figure 9.9) [29]. This change in refractive index in turn causes a shift in the FBG

wavelength resulting in wavelength modulation. The bandwidth of these sensors ranges from

0.1–5MHz.

Figure 9.9 Schematic of a fibre Bragg grating sensor with a required interaction length of at least

600 mm for sufficient sensitivity. Reproduced from [29]


9.2.3.3 Distributed Grating Reflector

Another type of sensing technique based on an internal distributed Bragg reflector (DBR) has

been reported recently [31]. Detection of this fibre sensor is based on the modulation principle

of birefringence induced by the incident acoustic pressure in the fibre laser. Two 1550 nm

Er–Yb doped grating structures each of length 10mm and 3mm respectively are written inside

a single mode fibre (Figure 9.10). The separation distance between the two gratings is about

10mm and results in a dual polarized signal. Awavelength divisionmultiplexer, alongwith the

photodetector, monitors beat frequency between both the polarization modes. In the presence

of an acoustic signal, the change in birefringence modulates the beat frequency and additional

sidebands are obtained at the output. The beat frequency serves as an indication of the incident

acoustic pressure, while its amplitude indicates the excitation voltage. The minimum detect-

able pressure level was calculated to be 164 dB re mPa and 158 dB re mPa at 10 and 20MHz,

respectively. Such a sensor can detect ultrasound up to 40MHz though its sensitivity is

relatively poor and its signal-to-noise ratio versus frequency is nonuniform.

These types of sensors are localized and have smaller sensing areas as compared to phase

modulated schemes. They are also insensitive to random amplitude fluctuations. However,

frequency or phase drift due to temperature can affect the performance of such systems. Also,

the sensing aperture dimension is of the order of a fewmillimeters. Hence, they cannot be used

for high spatial resolution measurements and also have limited bandwidth.

9.3 Intensity Sensing Principle, Design and Realization

Intensity-detection based optical hydrophone uses measurement of the Fresnel reflectance

caused by the change in refractive index between the fibre tip and the surrounding fibre

medium. The index of refraction of water depends on the acoustic pressure as reported earlier

and a resultant reflectance versus acoustic pressure is plotted in Figure 9.11 [26]. Sensitivity of

the fibre-optic ultrasound hydrophone probe, S, is calculated from the expression S¼V0/p,

where, V0 is the output voltage of the ultrasound hydrophone probe for a given acoustic

pressure p. With a light source power of 50mW, the theoretical sensitivity of the fibre-optic

probe is calculated to be 4.3mV/MPa for uncoated fibres. Figure 9.12 shows a typical APD

Figure 9.10 Schematic of an optical hydrophone employing a distributed Bragg reflector fibre sensor.

Reproduced from [30] (� 2005 IEEE)


output power at different acoustic pressure amplitudes and laser source power levels. The

�60 dBm plane corresponds to the calculated system noise floor, where it can be seen that at

100 kPa, to achieve an output signal-to-noise ratio of unity, a light source power level of 65mW

is needed. Also, with the level of 200mW, the minimum detectable acoustic pressure is about

34 kPa.

Figure 9.12 The photodetected output power from an APD as a function of different acoustic pressure

amplitudes and light source power levels as reported [26]. The �60 dBm plateau corresponds to the

calculated system RIN-dominated noise floor

-75

-70

-65

-60

-55

-50

-45

-40

1010.10.01

Acoustic pressure (MPa)

∆R

(d

B)

Figure 9.11 Reflectance change versus acoustic pressure, where a linear relationship is depicted


As indicated in this reported analysis the achieved response of this fibre hydrophone sensor is

not competitive with commercial hydrophones, such as the needle hydrophone with a

sensitivity of 40mV/MPa (i.e. �262 dB re V/mPa). System sensitivity improvements are

achieved by coating the fibre tip with gold. Design of optimum gold-coating thickness at the

fibre tip is identified throughmaximization of reflectance sensitivity to acoustic pressure and is

described next.

9.3.1 Transmission Line Modelling of Coated Fibres

A simplified transmission line model was used to improve the understanding of the impact of

fibre coating on the overall sensitivity of the optical hydrophone. Based on this model, the

complex reflection coefficient (r) of coated fibre can be expressed as

r ¼ ndðnc � nwÞþ ðnwnc � n2dÞtanh gdndðnc þ nwÞþ ðnwnc þ n2dÞtanh gd

ð9:2Þ

where d is the coating thickness, nd is the complex refractive index of the coating, nw is the

refractive index of water, and nc is the refractive index of the fibre core. The complex

propagation constant in a lossy gold layer, g, is described in terms of the complex index of

refraction of gold, nd, and wavelength, l, as:

g ¼ jð2pðn� jkÞl

Þ: ð9:3Þ

Note that n is the real part of the complex refractive index andk is the extinction coefficient. The

effective Fresnel reflectance (R) is given by R¼ r.r�, and the dependence of reflectance on

pressure can be found by differentiating the reflectance expression with respect to pressure:

dR

dP¼ r:

dr*

dpþ r*

dr

dpð9:4Þ

where

dr

dp¼ 2ndðdnw=dpÞ½nwtanhgd þ ndðtanhgdÞ2 � ncnd � ncnwtanhgd�

½ndðnc þ nwÞþ ðncnw þ nd2Þtanhgd �2: ð9:5Þ

The analysis presented here can be extended to uncoated fibre by considering the coating

thickness to be zero in Equation (9.2). Reflectance (R) and variation of the reflection coefficient

with pressure, are now given by

R ¼ nc � nwð Þ2nc þ nwð Þ2 ð9:6Þ

and

dr

dp¼ � 2ncðdnw=dpÞ

ðnc þ nwÞ2: ð9:7Þ


A comparison of the sensitivity of uncoated and coated fibre is presented in Figure 9.13 as a

function of coating thickness. In this calculation the complex index of refraction of gold has

been assumed to be 0.18� j2.21. Note that the calculation based on this simplified model

demonstrates an improvement of as much as 15 dB for a thickness of about 50 nm. Naturally a

more accurate model is required for down-tapered coated and uncoated fibres.

9.3.2 Finite Element Model (FEM) for Various Fibre Sensor Designs

In order to predict the performance of various fibre tip geometries, FEM simulations have been

performed using COMSOL. The physical dimensions of 0.1mm length and core diameter of

10 mm are considered for straight-cleaved optical fibre. Core and cladding refractive indices of

1.4456 and 1.4378 respectively are used for all fibres. A small tapering angle of 6� is consideredfor the etched fibre sensor to properly represent anisotropic etching. The FEM modelling of

gold-coated fibre is based on a gold-coating thickness of 100 nm with complex index of

refraction of 0.18� j0.31. The hybrid mode of HE11 is considered as the dominant mode and

the simulated power density profile of the field along the fibre length is depicted in Figure 9.14.

All the sensor tips have been immersed in water. The region outside the cladding is considered

as an absorbing boundary in order to reduce computation complexity. Simulation results for

straight cleaved, down-tapered uncoated and coated fibres have indicated power densities of

�46.7,�51.5 and�36.3 dBm/mm2 respectively, with amarked improvement in reflectance for

the coated fibre case.

9.3.3 Fabrication of Fibre Sensors

The probe sensitivity can be enhanced by increasing the reflected signal power. The fibre tip is

etched to the size of 7–10microns before it is coated with a thin layer of metallic material, such

10-1 100 101 102 1030

2

4

6

8

10

12

14

16

18

20

X: 52Y: 16.15

Impr

ovem

ent i

n se

nsiti

vity

, (dB

)

Coating thickness, (nm)

Figure 9.13 Improvement in sensitivity of a straight cleaved fibre with a thin gold coating. (The

classical appproach formodelling that is presented here is not totally accurate for thin films, but it is a good

approximation for a general understanding of this problem)


Figure 9.14 FEM simulation of power density distribution along the direction of propagation of HE11

for various sensor designs using COMSOL; (a) straight cleaved fibre with mm cross-sectional diameter,

(b) tapered fibre with 7mm cross-sectional diameter, (c) tapered fibre with gold coating

as gold. The fibre tip is etched to a smaller diameter by wet chemical etching of the fibre, using

HF (50% by volume) solution. In this process the fibre being etched is connected to a light

source and detector, and the fibre tip is dipped in the HF solution for tip etching, while the

back reflected signal is continuously monitored. The etched fibre is then coated with a


Figure 9.15 Sensing fibre tip images with 10�magnification: (a) down-tapered gold coated fibre and

(b) down-tapered uncoated fibre tip cross-sectional diameter, 7 mm

semitransparent film of gold. The thin gold layer was sputtered on the fibre tip with a

Cressington 108 sputtering machine, and the approximate thickness of the gold layer is

50–150 nm for sputtering times of 5–20 s. An optical image of the fabricated coated and

uncoated fibres is shown in Figure 9.15.

9.3.4 Experimental Set-up and Results

The fibre-optic hydrophone system is constructed with commercially available components

using singlemode FC/APC connectors. The system block diagram is shown in Figure 9.16, and

the system is composed of optical source, optical sensor, acoustic source and optical receiver

assemblies, as reported in [31]. The optical source is the 1550 nm distributed feedback (DFB)

laser (NEC NX8563LB) with an output power of �2 dBm for Ib¼ 30mA. The source is

coupled to a 10 dB optical coupler and the output from the 10% coupled arm is sent to the

erbium-doped fibre amplifier (NuPhotonics NP2000CORSV303500FCA1) which has an

optical gain of 40 dB and output power of up to 30 dBm. The output from the EDFA is

divided equally using a 2� 2 coupler with a 3 dB coupling factor. One of the optical outputs is

immersed in water as reference, while the other output is connected to the optical sensor. The

optical sensor is immersed in a water tank and placed at the focal point of a focused acoustic

transducer.

The acoustic transducer is a one-element transducer (Sonic Concepts H110AS/N 01) with

dual band operation at frequencies of 1.6MHz and 5.0MHz. It requires a radio frequency (RF)

impedance matching network and 50W over dual bands of 1.41–1.98MHz and 5.0–5.7MHz.

The transducer has an active diameter of 20 mm and it has a focal length of 34.52mm. An RF

power amplifier provides a maximum pulsed power level of 100Wwith 25% duty cycles. The

position of the acoustic source and optical hydrophone are controlled by a precision scanning

system from Onda Corporation. The system provides precise six-axis positioning and data

acquisition from any sensor in thewater tank for accurate measurements of acoustic fields. The

precision of each axis is repeatable within 12 mm and absolute accuracy of 25 mm over 30 cm.

Experiments have been performed using the various optical sensors discussed in Section 9.3.3.

The reflected optical energy is collected in a wide band amplified InGaAs detector (Thor Labs

PDA 10CF) with a responsivity of 0.95A/Wat 1500 nm and signal bandwidth of 150MHz. It

has a transimpedance gain of 5 kW and noise equivalent power of 12 pW/(Hz)1/2. Comparison


Chi

p-ho

pe3

dB c

oupl

er

SMSC

A22

3RP

5005

FA

Wat

er ta

nk

OU

T

Ref

eren

ce

fibre

Sens

or fi

bre

Pow

eram

plifi

er a

nd

mat

chin

g ci

rcui

t

Ultr

asou

nd

trans

duce

r

Thor

labs

PIN

ph

oto-

dete

ctor

PD

A10

CF

Agi

lent

spec

trum

an

alys

er

E840

8A

50%

OU

T

50%

OU

T

ININ

NEC

DFB

La

ser 1

550n

m

NX

856

3LB

10%

OU

T

90%

OU

T

IN

10dB

cou

pler

SMSC

A22

3RP1

005F

A

Nu

phot

onic

s ED

FA

NP2

000C

OR

SB30

3500

FCA

1

Opt

ical

isol

ator

ISIL

PD55

SS9

0.76

m

0.38m

0.46 m

Not

con

nect

ed a

t pr

esen

t

Sign

al

gene

rato

rag

ilent

3325

A

Figure

9.16

System

block

diagram

[31]


of fibre performance using three designed sensors depicted in Figure 9.14 has typically

indicated an 11 dB and 13 dB improvement for gold 5 second (about 50 nm thick) coated fibre

over straight cleaved and uncoated down-tapered etched fibre sensors, which corroborates the

simulated predictions from FEMmodelling. Themeasured reflected optical signals at different

EDFA power levels indicate a 2 dB variation for every dB variation of optical power. The noise

floor level also increases at the same rate, which indicates that the receiver noise is dominated

by amplified relative intensity noise (RIN) of laser source. Using common mode rejection of a

balun as a power combiner, the amplified RIN noise is cancelled by 14 dB leading to a shot-

noise dominated noise floor of the detection system [32]. The measured optimum gain

responsivity for 25 dBmoptical power ismeasured as�245 re 1V/mPawith average sensitivityas low as 150 Pa.Comparison of time domain pressure response of this optical hydrophonewith

commercially available needle hydrophones is depicted in Figure 9.17, where a modulating

tone burst at frequency of 1.5MHz is employed. The small active sensing area of <10 mm of

these sensors avoids spatial averaging leading to high spatial resolution required for diagnostic

applications. As depicted in Figure 9.18, the sensitivity of this optical hydrophone is flat at least

up to 60MHz, which is significantly better than the results achieved for the commercially

available membrane (polyvinylidenfluorid (PVDF)) hydrophone, even after spatial averaging

corrections [7, 8, 33, 34]. Comparison of the best reported data in each category of fibre-optic

hydrophone (i.e. intensity, phase, wavelength and frequency methods) is summarized in

Table 9.1. The reported experimental results (circa 2008) show that the designed gold coated

fiber provides the best performance in terms sensitivity, bandwidth, and responsivity.

8 10 12 14 16 18 20-3

-2

-1

0

1

2

3

4

Time, (µs)

Vol

tage

, (V

)

needle hydrophone

gold coated FOPH

Figure 9.17 Comparison of the measured pressure versus time response of the 5 seconds gold-coated

fibre-optic hydrophone (in solid line) against a commercially available needle hydrophone (in dash line)

to a 1.5MHz tone burst at acoustic pressure of 1 MPa


100 MHz hydrophone calibration

-310

-300

-290

-280

-270

-260

-250

-240

-230

1009080 70 605040302010 0 Frequency (MHz)

PVDF hydrophone with TDS method

PVDF hydrophone with TGFA method

PVDF hydrophone with nonlinear model method

Current gold coated fibre optic probe hydrophone

Hyd

roph

one

sens

itiv

ity

(dB

re

1V/µ

Pa)

Figure 9.18 Comparison in sensitivity response versus frequency for the fibre-optic hydrophone and a

membrane hydrophone. Upper trace data for the 5 second (50 nm) gold-coated fibre-optic hydrophone up

to 60MHz in solid line and estimated response up to 100MHz in dashed line. Lower traces and data points

are for a commercially available 0.5mm diameter, 25mm thick, coplanar PVDF membrane hydrophone.

Two measurement methods of TDS and TGFA are employed to correct for spatial averaging errors of a

500mm wide hydrophone

Table 9.1 Summary of performance comparison of various acoustic pressure sensors.

Sensing technique Detection technique

Gain

responsivity B.W.

Minimum detect-

able pressure

Acousto-optic phase

change in mandrel

fibre [22]

External interfero-

metric phase

detection

Not reported 0.75–10 kHz 92� 27 kPa

Inrtinsic Fabry–Perot

resonant structure [27]

Intensity detection Not reported 20MHz 5 kPa

Multilayer resonant

structure [28]

Intensity detection �264 dB re

1V/mPa10 MHz Not reported

External Bragg cell

wavelength

modulation [24]

FM detector

frequency

detection

Not reported 100Hz

�1.2 kHz

1 kPa

Wavelength modula-

tion of distributed

Bragg reflector

fibre [30]

Intensity detection

of beat

frequencies

Not reported 20MHz 6.4 kPa

Intensity modulation of

reflected light

Intensity detection

of reflected light

�245 dB re

1V/mPaOver

60MHz

150 Pa in shot-noise

dominated


9.4 Approaches to NIR Imaging

Shadow images created as light passes through the body were first proposed by Cutler in 1929

for medical imaging [35]; however, he found the low resolution of the images limited its

clinical application due to high scattering and absorption. In the past 20 years, significant

advancements in laser and detector technologies in the near-infrared (NIR) electromagnetic

spectrum have been driven by the long haul telecommunication industry; combined with a

better understanding of light propagation in tissue this has now led to renewed interest in optical

imaging of the human body as well as acquiring information about tissue optical and dynamic

properties noninvasively.

In NIR spectroscopy the main aim is to extract the optical properties (absorption and

scattering) of the living tissue. The absorption, ma, and reduced scattering, m0s, parameters of

tissue can provide information on a variety of physiological processes. Absorption information

is used to characterize the concentration of biological chromophores, such as haemoglobin,

which in turn indicates the physiological changes in blood [36]. Scattering information

quantifies the composition, density and organization of tissue structures, such as cells and

subcellular organelles [37, 38]. Therefore NIR techniques could ultimately provide informa-

tion about disease-related functional and structural changes in tissue.

Currently, three main categories of diffuse optical measurements have been developed: (i)

continuous wave (CW), (ii) time domain and (iii) frequency domain measurements. In

continuous wave (CW) systems, light sources emit light continuously at constant amplitude

(or are modulated at frequencies not higher than a few tens of kHz to reject ambient light using

synchronous detection schemes). CW systems measure only the amplitude decay of the

incident light. Time-domain, or time-resolved, systems introduce extremely short (picosecond)

incident pulses of light into tissue, which are broadened and attenuated by the various tissue

layers (e.g. shin, skull, cerebrospinal fluid and brain). A time-domain system detects the

temporal distribution of photons as they leave the tissue, and the shape of this distribution

provides information about tissue absorption and scattering. In frequency-domain systems, the

light source shines continuously but is amplitude-modulated at frequencies at least on the order

of tens of MHz. Information about the absorption and scattering properties of tissue are

obtained by recording the amplitude decay and phase shift (delay) of the detected signal with

respect to the incident signal. These techniques are each discussed next.

9.4.1 Continuous-wave (CW) Method

The absorption spectrumof tissue iswavelength dependent, which ismainly the contribution of

Hb (haemoglobyn), HbO and water, as depicted in Figure 9.2, where the absorption coefficient

is provided for blood concentration of 5% in whole tissue and 100% water. Measuring the

concentration of an absorbing species in a sample is accomplished by applying the Beer–

Lambert law [39], where the absorption of a sample at a given wavelength is directly

proportional to the concentration of the absorbing material, its extinction coefficient and the

path length of light through it. The Beer–Lambert law assumes that the medium is homoge-

neous, the incident light is collimated and reflection and scattering do not contribute to the loss

of the transmitted light. The Beer–Lambert law analytically expresses optical density (i.e.

absorbance) as

lnðI0=IÞ ¼ srd ¼ mad ð9:8aÞ


or

log10ðI0=IÞ ¼ eCd ¼ mad

2:3¼ Optical Density ðODÞ ¼ absorbance ð9:8bÞ

where I0 is the incident intensity, I is the transmitted light intensity, s is the absorption cross

section, r is the number density of the absorbing molecules, C is the concentration of the

absorbing molecules (in mM), d is the path length (in cm), e is the extinction coefficient for asolutionofmolarconcentration(inmolar�1 cm�1),andma is theabsorptioncoefficient (incm

�1).

The Beer–Lambert relation holds true when specular reflection or scattering does not

contribute to the loss of transmitted light. This is clearly not the case in tissue. When the

scattering length is shorter than, or comparable to, the absorption length, the optical properties

cannot be accurately determined using the Beer–Lambert law. The first attempts at diagnostic

imaging using optical radiation revealed that multiple scattering occurs when visible to

near-infrared light propagates through tissue and blurs features below the surface. As a

consequence, any measurement of the transmitted intensity through more than a few milli-

meters of tissue is dominated by scattered light. The scattering characteristic of tissues is

commonly expressed in terms of the transport (or reduced) scattering coefficient (correspond-

ing to isotropic scattering),

m0s ¼ msð1� gÞ ð9:9Þ

where ms is the scattering coefficient and g is the anisotropy factor of scattering equal to the

average cosine of the single-scattering phase function [40]. In order to correct for the multiple-

scattering effect in the tissue, a modified Beer–Lambert law is introduced,

OD ¼ � log10I

I0¼

Xi

eiCiLBþG ð9:10Þ

where L is the path length (in cm), B is a path-length factor, which accounts for increases in the

photon path length caused by tissue scattering, G is the measurement geometry factor, and

index ‘‘i ’’ represents the ith chromophore. Parameters e and L remain constant, andB andG are

assumed to be constant. Therefore the change in optical density is given by

DOD ¼ � log10Ifinal

Iinitial¼

Xi

eiDCiLB: ð9:11Þ

By considering the contribution of only two chromophores, Hb andHbO, the above equation

becomes:

DODl ¼ ðelHbOD½HbO� þ elHbD½Hb�ÞLBl ð9:12Þwhere [HbO] and [Hb] are themolar concentrations of oxy- and deoxy-haemoglobin for a l thatindicates a particular optical wavelength. The changes in oxy- and deoxy-Hb concentrations

(and therefore the change in total Hb concentration) are assumed to be wavelength indepen-

dent. This assumption could be invalid, if different wavelengths sample different volumes of

tissue with different haemoglobin concentrations.

By measuring DOD at two wavelengths (l1 and l2) and using the known extinction

coefficients of oxy-haemoglobin (eHbO) and deoxy-hemoglobin (eHb) at those wavelengths,


we can then determine the concentration changes of oxyhaemoglobin and deoxyhaemoglo-

bin [41],

D Hb½ � ¼ el2Hbo DODl1

Bl1 � el1Hbo DODl2

Bl2

ðel1Hbel2HbO � el2Hbel1HbOÞLð9:13aÞ

D HbO½ � ¼ el1Hb DODl2

Bl2 � el2Hb DODl1

Bl1

ðel1Hbel2HbO � el2Hbel1HbOÞL: ð9:13bÞ

In addition to the Beer–Lambert law and the model presented here, a more rigorous theory

for the migration of photons through tissue has been developed based on the radiative transport

equation [42]. This approach recognizes that near-infrared photons in tissue essentially

undergo a random walk because the scattering probability is much greater than the absorption

probability, and therefore their propagation through tissue can be described by a diffusion

equation. The photon diffusion equation is [13, 42, 43]:

1

v

@Fðr; tÞ@t

�Dr2Fðr; tÞþmaFðr; tÞ ¼ Sðr; tÞ; ð9:14Þ

whereF(r, t) is the photon fluence at position r and time t and the photon fluence is proportional

to the optical intensity. S(r, t) is the source distribution of photons. D¼ 1/[3(ma þ m0s)] is the

photon diffusion coefficient, m0s is the reduced scattering coefficient, ma is the absorption

coefficient and v is the speedof light in themedium.Note that the absorption coefficient is related

to the extinction coefficient and the concentration as ma¼ eC. For a combination of the

haemoglobin chromophores,

ma ¼ eHbO½HbO� þ eHb½Hb�: ð9:15ÞEquation (9.14) accurately models the migration of light through highly scattering media

provided that the probability of scattering is much greater than the absorption probability. Note

that all factors in (9.14) are wavelength-dependent. Solutions of the photon diffusion equation

can be used to predict the photon fluence (or intensity) detected for typical diffuse measure-

ments. Assuming that concentration changes are both global and small, the solution of the

photon diffusion equation for a semi-infinite medium is

DOD¼�logFFinall

FInitial

¼1

2

3m0s

mInitiala

� �1=2

1� 1

ð1þLð3m0Initials mInitial

a Þ1=2" #

ðeHbOD HbO½ �þeHbD Hb½ �ÞL

ð9:16ÞThe solution of the photon diffusion equation for representative tissue geometry (Equa-

tion (9.16)) tells us that the modified Beer–Lambert law is reasonable for tissues with spatially

uniform optical properties when the chromophore concentration does not change significantly

(i.e. D[X]/[X]� 1). The path length factor B in Equation (9.12) is given by

B ¼ 1

2

3m0s

mInitiala

� �1=2

1� 1

ð1þ Lð3m0Initials mInitial

a Þ1=2" #

ð9:17Þ


for a semi-infinite medium. This shows that B depends on tissue scattering, the initial

chromophore concentration, the extinction coefficient (and thus B is wavelength dependent)

and the optode separation. In practice, the validity of the assumption thatB is independent ofma

and L has often been ignored sinceB is in general empirically determined and the changes inma

are typically small.

The quantity of oxygen in blood is often expressed as the haemoglobin oxygen saturation (S),

which is defined as

S ¼ ½HbO�½HbO� þ ½Hb� � 100% ¼ ½HbO�

½HbT� � 100%: ð9:18Þ

This expresses the percentage of the total oxygenated haemoglobin.

A typical CW imager is shown in Figure 9.19, where light sources are driven by the drive

circuitry and emit near-infrared light into tissue. The diffused and attenuated light is collected

and converted to an analogue electrical signal by the photodetector. Finally, the amplified

signal goes through theA/D converter so that a computer can be used to process and display the

data. As shown in Figure 9.19, the light source is at the second stage of the open loop, so that the

quality of the light source is vital to the whole system. Generally, there are three choices for

light sources inCWimagers:white light (such as tungsten light bulbs), lasers and light-emitting

diodes (LED). Light spectrum purity and light intensity output are two important parameters.

White light has been extensively used with interference filters at 760 nm and 850 nm to detect

blood volume and deoxygenation changes [39]. Lasers are ideal light sources for many

applications due to their excellent spectral purity and collimation. The linewidth of the isolated

wavelength is less than 1 nm.A laser beam focuses all the light energy into a very small area and

over a very small wavelength bandwidth, with potential for tissue damage even though its

power is much less than that of a white light source. This is why its power is limited to less than

0.1mWby the Food andDrugAdministration�s (FDA�s) law (type I) when laser light is applied

to humans. Thus, it will be difficult to satisfy the light intensity requirement of a CW imager.

LED spectral purity is about 30 nm and it is good for a CW imager. More light intensity can be

utilized since LEDs illuminate the tissue more diffusely than a laser but more like a point light

source than white light, and with less heat. Stability of light intensity is another important

Figure 9.19 Typical block diagram of a single channel CW imager [42].


requirement of a CW imager. Both white light and LEDs have a drift of light power and thus

require 2–3 minutes for warming up, but the laser diode operates more stably.

A typical experimental result by an LED imager [42] is shown in Figure 9.20. The probewas

placed on the lateral side of the lower right-hand side of the leg. Seated baseline measurements

were made for 1 minute. Seated exercise began by the subject doing 60 repetitions of toe

extension (pointing the toe as far as possible). This seated exercise recruited the extensor

muscles to a greater extent than the flexors which are more activated during plantar flexion or

walking.

Figure 9.20 Blood volume and deoxygenation changes during cycling exercise are presented during

physical exercise and rest periods. A grey scale coding (black for increase, light grey color is for moderate

decrease, and dark grey for significant decrease) is employed to provide approximate changes in blood

volume and deoxyngenation levels (in unit aM) compared to the initial condition represented in grey

colour [42] (Reprintedwith permission fromY. Lin, et al., ‘‘Noninvasive, low-noise, fast imaging of blood

volume and deoxygenation changes inmuscles using light-emitting diode continuous-wave imager,’’ Rev.

Sci. Instrum. 73(8), 3065–3074 (2002). Copyright 2002, American Institute of Physics)


9.4.2 Pulsed-time or Time-resolved Method

Time-resolved reflectance spectroscopy (TRS) is a novel nondestructive method for the

complete optical characterization of highly diffusive media, that is for the evaluation of the

absorption coefficient ma and the reduced scattering coefficient m0s. TRS is gaining acceptance

in biomedicine for the noninvasive investigation of biological tissues [44–46] since a short light

pulse injected into a turbid medium experiences absorption and scattering during photon

propagation. Moreover, the diffusely reflected pulse is attenuated, broadened and delayed.

Consequently the best fit of its time distribution with a theoretical model of light propagation

allows the simultaneous evaluation of both optical coefficientsma andm0s by probing bulk rather

than superficial properties. Furthermore, useful information on internal quality of tissue can be

gathered.

When a narrow collimated pulsed light beam is normally incident on the surface of a semi-

infinite or finite homogeneous tissue slab, the diffuse photon fluence rate F(r, t) satisfies thediffusion equation (Equation (9.9)). The fluence rate can be accurately calculated using

Equation (9.14) if ma � m0s and if the point of interest is far from sources or boundaries. For

a short pulse from an isotropic point source that is represented by a delta function of

s(r,t)¼ d(0,0), it may be shown that in an infinite medium the solution of Equation (9.14) is

Fðr:tÞ ¼ vð4pDvtÞ� 3=2exp � r2

4Dvt�mavt

� �: ð9:19Þ

One can use this Green�s function to solve the semi-infinite problem by making two further

assumptions. First, assume that all the incident photons are initially scattered at a depth

z0 ¼ ðm0sÞ� 1

so that the actual source term becomes the simple delta function described above.

The second assumption is that F(r,t)¼ 0 on the physical boundary z¼ 0. As discussed by

Eason et al. [47], this boundary condition can be met by adding a negative or image source of

photons to the infinite medium. The fluence rate per incident photon can then be written in

cylindrical coordinates as the sum of contributions from the two sources:

Fðr; z; tÞ ¼ cð4pDctÞ� 3=2expð�mactÞ exp � ðz� z0Þ2 þ r2

4Dct

" #� exp � ðzþ z0Þ2 þ r2

4Dct

" #( ):

ð9:20ÞThe number of photons reaching the surface per unit area per unit time, jJ(,0,t)j, can be

calculated from Fick�s law as:

jJðr; 0; tÞj ¼ �DrFðr; z; tÞjz¼0 ð9:21Þwhich leads to the final expression for the reflectance R(r, t):

Rðr; tÞ ¼ jJðr; 0; tÞj ¼ ð4pDcÞ� 3=2z0t

� 5=2expð�mactÞexp � r2 þ z204Dct

� �: ð9:22Þ

For the case where r2 � z20, also note that

d

dtlnRðr; tÞ ¼ � 5

2t�macþ

r2

4Dct2: ð9:23Þ


The observation that

limt!¥

d

dtlnRðr; tÞ ¼ �mac ð9:24Þ

leads to the suggestion that the absorption coefficient of the tissue can be determined from the

asymptotic slope of the ln R(, t) versus t curve. The transport scattering coefficient m0s can also

be determined from the lnR(, t) versus t curve by noting that the slope is zero at tmax, the time of

maximum detected signal. Solving Equation (9.23) yields the expression:

m0s ¼ 1

3r24mac

2t2max þ 10ctmax

� ��ma ð9:25Þ

Therefore, both optical properties of a semi-infinite slab of tissue could, in principle, be

obtained from Equations (9.24) and (9.25) by measuring the diffusely reflected light some

distance from the source as a function of time. A superior signal-to-noisewould be obtained by

integrating the reflected light over some larger area.

The TRS instrumentation consists of both hardware and signal processing software. The

laser pulse scanning mammography, as shown in Figure 9.21 and developed by PTB [48],

measures time-resolved transmittance through the female breast, which is gently com-

pressed between two parallel glass plates. When the source fibre and the detector fibre

bundle are scanned in tandem across the breast, optical properties can be extracted from the

measured photon density versus time. The proposed mammograph is equipped with two

excitation channels at 670 nm and 785 nm and one detection channel. The output pulse

trains of two picosecond laser diodes are multiplexed in time, and optical mammograms are

simultaneously recorded by detecting the transmitted photons by a fast photomultiplier.

Distributions in the times of flight are recorded for 100ms at each scan position by high-

throughput time-correlated single photon counting electronics at count rates of up to 1MHz.

The optical mammograms are recorded along cranio-caudal and medio-lateral projections

within 3 to 5 minutes each as 1000–2000 scan positions are typically sampled at a step size

of 2.5mm.

Mammograms were generated from a variety of parameters derived from recorded

distributions of times of flight, such as photon counts in selected time windows. By

analysing photon counts in a late time window, changes in absorption can be imaged

qualitatively, whereas photon counts in an early time window are most sensitive to changes

in scattering. In Figure 9.22 the tumour shows up as reduced transmittance in the late time

window, whereas the cyst is clearly seen in the image representing photon counts in the first

time window.

In the clinical trials, mammograms are recorded with the transmitting fibre and receiving

fibre bundle facing each other (on-axis geometry). Whereas lesions can be localized laterally

with sufficient precision in optical mammograms recorded in this way (localization of lesions

in two dimensions), the position of lesions along the compression direction cannot be inferred

from measurements taken using the on-axis arrangement. A promising approach to gain

information about the location of the lesion under investigation along the compression

direction (three-dimensional localization) is to record optical mammograms at several lateral

offsets between the transmitting fibre and detecting fibre bundle (off-axis geometry) and to

analyse the shifts of features in mammograms.


9.4.3 Intensity-modulated or Frequency-domain Method

Light propagation in scattering media can be described by the Boltzmann transport equation.

Under specific approximations and limitations, the Boltzmann transport equation is simplified

to the time-dependent standard diffusion equation (SDE) as shown in Equation (9.14).

Analytical solutions for the SDE have been described for a variety of boundary and initial

conditions. In the case of a sinusoidal point source modulated at a frequency of f¼v/2p, theinfinite medium solution is given by [13, 48]:

fðr; tÞ ¼ Adc

4pD

expð� r=dÞr

þ Aac

4pD

expð� krealrÞr

� exp � iðkimagr�vtÞ� � ð9:26Þ

where Adc and Aac are the DC and RF components of the source respectively, d is the DC

penetration depth, and kreal and kimag are the real and imaginary components of the photon

density wave (PDW) complex wave number. By convention, kreal governs PDW amplitude

Mode-locked laser 785nm

Mode-locked laser 670nm

Photodetector (photomultiplier)

Time-correlated single-photon counting

Signal processor (computer)

~ 400ps

~ 100ps

Optical fibre

X-Y Scanner

x

z

y

Fibre bundle

~ 2.4ns

Dispersed short optical pulses after

flight through breast

Figure 9.21 Schematic diagram of the initial PTB time-domain scanning optical mammography, where

two trains of optical short pulses are combined before injection into the breast tissue with a black circle

representing the tumour/cyst. Reproduced from [48]


attenuation and kimag describes PDW phase propagation. The complex wave vector, I¼ krealkimag, is dependent on the absorption and reduced scattering parameters, as well as the source

modulation frequency and the velocity of light in the medium:

kreal ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffi3

2mam

0s

r1þ v

cma

� �2" #1=2

þ 1

8<:

9=;

1=2

ð9:27Þ

kimag ¼ffiffiffiffiffiffiffiffiffiffiffiffiffiffi3

2mam

0s

r1þ v

cma

� �2" #1=2

� 1

8<:

9=;

1=2

: ð9:28Þ

For reflectance or transmittance measurements performed in infinite media, PDW phase lag

and amplitude attenuation relative to the source are

Qlagðr;vÞ ¼ kimagr ð9:29Þ

Aattðr;vÞ ¼ exp½ � krealðvÞr�4pDr

: ð9:30Þ

For semi-infinite geometries, in reflectance mode and accounting only for the fluence term,

PDW phase lag and amplitude attenuation are given as follows:

Figure 9.22 TRS generated mammograms of a breast (cranio-caudal projection, l¼ 785 nm) contain-

ing a tumour (invasive ductal carcinoma) and a cyst as a function of scan distance of x in cm: (a) reciprocal

photon count in N8, the eighth of 10 consecutive time windows, (b) photon counts in N1 the first time

window. Reproduced from [48]


Qlagðr;vÞ ¼ kimagðvÞr0 � arctanIMAG

REAL

� �; ð9:31Þ

Aattðr;vÞ ¼ Air

4pDðREAL2 þ IMAG2Þ1=2; ð9:32Þ

where real and imaginary parts of the received signal are analytically related to operating

frequency and physical dimensions by:

REAL ¼ exp½ � krealðvÞr0�r0

� cos kimagðvÞðr0b � r0Þ� � exp½ � krealðvÞr0b�

r0bð9:33Þ

IMAG ¼ sin kimagðvÞðr0b � r0Þ� � exp½ � krealðvÞr0b�

r0b: ð9:34Þ

Note that distances r0 and r0b are expressed by:

r0 ¼ ½ðm0sÞ� 2 þ r2�1=2; ð9:35Þ

r0b ¼ 4

3m0s

1þReff

1�Reff

þ 1

m0s

� �2

þ r2

" #1=2

: ð9:36Þ

Air is the net amplitude response of the instrument (due to source power, detector gain, etc.) and

Reff is the effective reflection coefficient.Reff represents the fraction of light that has reached the

surface and is reflected back into the medium.

The frequency-domainmeasurement technique provides both optical tissue parameters such

as TRS and is therefore more advantageous than the CW imaging system. From the three

techniques discussed so far, the frequency domain is most compatible with advances in radio

technologies and hence ismore cost effective than TRS. There are several approaches to realize

frequency-domain detection, such as the I/Q system, the phased array system and the

broadband frequency-domain system.

9.4.3.1 I/Q System

Figure 9.23 depicts the basic form of an I/Q system [49, 50], where two in-phase and quadrature

phase signals are employed for the detection of optical parameters. The phase difference and

amplitude attenuation between the reference oscillator and the signal pathway is detected. The

phase-shifted path involves a laser diode, an optical detector, an appropriate amplifier and a

narrowband filter. Thus the outputs of the I/Q detector are the sine and cosine components of

phase and amplitude and thus require trigonometric computation by nonlinear analogue

circuitry or by conversion to the digital domain and the use of a look-up table or other similar

method. Quadrature imbalance in the I/Q detector is typically 0.3� in phase and 0.5 dB in

magnitude, but anyunwanted signalwill be detected and cause variableDCoffsets in the output

that need to be distinguished from the DC sine/cosine output.


9.4.3.2 Phased-array System

To achieve high sensitivity in the detection of small objects embedded in a scattering medium,

dual interfering sources in a phased-array configuration have been explored experimentally and

theoretically [51, 52]. The analytical solution of the dual-source system can be derived by

applying the superposition principle of a linear system by summing over the solutions to the

single source system.Complex terms ofA1exp(�jvt) andA2exp(�jvt þ Df) are employed to

represent the pair of sources. The total field is equal to the superposition of the two independent

solutions of those terms as represented by:

Ftotalðr; tÞ ¼ F1ðrs1 ; tÞþF2ðrs2 ; tÞ: ð9:37ÞPhoton-density waves generated from a pair ofmodulated optical sources that are excited by

180� out-of-phase RF signals, have been shown byKnuttel et al. [53] to interfere destructively.A null exists along the line of symmetry in the phased-array geometry, where a detector

placement along the null-line detects a very small perturbation to the symmetric environment.

The goal of this initial application was to desensitize the detector at the excitation surface (in a

reflection mode measurement) to effects near the surface. It was shown that the phase

measurement provided greater sensitivity for absorbers at larger depths than for the single-

source case (cf. Figure 9.24). In this way, the approach showed promise for localization by

providing improved depth information with a reflection measurement on the planar source

boundary.

A block diagram of a frequency-domain phased-array system is shown in Figure 9.25. A 50

MHz oscillator is modulated in the single sideband (SSB) mode by a 2 kHz sine wave and the

upper sideband (USB) from the transmitter is selected and split into 0� and 180� by an RF

splitter. The RF signals then modulate the two laser diodes respectively with an optical

modulation near 90%. The optical signal after passing through the diffused medium and

experiencing scattering due to the presence of an object is detected by a photomultiplier tube

(PMT). The output of the PMT goes through two separate channels for detection of amplitude

and phase. The automatic gain control (AGC) voltage of the receiver is used to indicate the

amplitude of the RF signal and the phase information is obtained from the phase meter.

A dual-source scanner that collects optical parameters of cylindrical absorbers using a

phased array system is depicted in Figure 9.26(a). Two sizes of objects of 1 cm and 0.1 cm

Turbid medium

under test

Laser diode

Opticalsplitter

Single RF tone

Photo- detector

IQ demodulator

Signal processor

Figure 9.23 Block diagram of I/Q system based on the detection of optical parameters (reduced

scattering and absorption) of a medium under test. Reproduced from [50]


Phase perturbation

0.01

0.1

1

10

100

0.10.010.001Contrast - delta µa (cm-1)

Ph

ase

shif

t (º

)

Simulation-SS

Simulation-DS

Detection noise

Experiment - SS

Experiment - DS

Figure 9.24 Phase perturbation effect in a phased-array system due to a contrast agent used in an intra-

lipid tank, where experimental and simulation results of a signal source (SS) and dual sources (DS) are

compared as the absorption parameter ma is changed. Reproduced from [51]

50MHz oscillator

SSB modulator

SP

LIT

2kHz phase detector

2kHz oscillator

2kHz

filter

ADC

50MHz SSB receiver

ADC

50MHz SSB receiver

x

z

y

1Hz

0°

180°

Detector PMT

S1

S2

Object

1Hz

USB

Phase detection

AGC control

Amplitude detection

Figure 9.25 Block diagram of a phase-sensitive phased-array detection system for the optical

parameters of an object in an intra-lipid tank. Reproduced from [51]


diameters are used in this experiment. The two-dimensional scanning of the homogeneous

intra-lipid loaded by the cylindrical object is collected using a measurement system similar to

Figure 9.25 in terms of amplitude and phase. However, modern automatic network analysers

could be employed for accurate amplitude and phase measurements.

The experimental results are depicted in Figure 9.26(b) and (c). Consider the homogeneous

case first. When the dual source is positioned toward either side of the domain, the boundary

selectively absorbs photons from the nearest source, disrupting the interference line. As the

dual source approaches the centre of the input scan line, the mutual absorbing effects of the

boundaries begin to balance, allowing equivalent contributions from both sources that are out

of phase to reach the output plane, manifested as a sharp amplitude null and a 180� shift inphase. When a strong absorber is introduced at the centre of the region (1 cm in diameter), an

additional balancing of absorbed light occurs between each boundary and the absorber.

Therefore, at a point between the boundary and the absorber, the loss to the boundary and

the heterogeneity are somewhat balanced and a smaller null line is detected. Likewise, when

the sources are centred about the heterogeneity, a large-amplitude null and phase step are

detected.

As explained in [52], the positions and sizes of the nulls in the magnitude are related to the

nature of the absorber and the proximity of the boundaries. The larger absorber has the

narrowest central null and satellite nulls that are closer to the outer boundary; all nulls are

Figure 9.26 Experimental results of a dual source phased array: (a) typical phased-array geometry,

(b) amplitude and (c) phase response, when two-dimensional scanning is performed over a cylindrical

absorber. Reproduced from [52]


sharper owing to the larger effective absorption by the heterogeneity. The phase responses also

reflect the size of the absorber. All three curves have a 180� phase step in the centre. Over theremainder of the domain, the phase is smooth. The other abrupt phase changes are artificial

phase wraps (as phase plots go through phases larger than 180� or smaller than –180� in

automatic network analysers) and could be corrected with modulo of 2p. Notice that the

magnitudes and phases for the cases with the absorbers asymptotically approach the homoge-

neous case. If the domain were infinite in this scan arrangement, one would see a magnitude

minimum when the detection plane was an appreciable distance on either side of the absorber

and alsowhen it was centred at the absorber, as well as a reduction in null depth just off centre.

Finally, note that the magnitude approaches a maximum as the minimum and maximum scan

positions are approached. Data points do not go all the way to the boundary, where light

intensity at the detector would decrease as a result of losses through the boundary. The

experimental results show the power of differential detection, where small perturbations can be

clearly identified as the line of symmetry is destroyed.

9.4.3.3 Broadband Frequency-domain System

The development of phase-sensitive phased arrays can benefit from the accuracy of modern

vector network analysers, which bring in the capability of accurate broadband measurement

using calibration procedures. There are several advantages of broadband characterization of

tissue. For example, most tissues normally have a layered structure and since photon

penetration depth depends on modulation frequency, as frequency-domain photon migration

(FDPM) is evaluated over a broad bandwidth, intensity-modulated light can be used to quantify

multi-layer tissue absorption (ma) and reduced scattering ðm0sÞ parameters at discrete wave-

lengths, as depicted in Figure 9.2 for low (MHz) and high (GHz)RF frequencies. This feature is

extremely useful in clinical experiments where, due to calibration difficulties, it is preferable to

make a single measurement and obtain as much information as possible [10, 14, 54].

To appreciate fully the advantages of operation at various frequencies to extract optical

properties of a layered tissue, the amplitude and phase of received scattered signals are

simulated as a function of frequency employing Equations (9.31) and (9.32). Simulations are

provided for a homogenous phantom resembling breast tissue and calculations are conducted

for reduced scattering and absorption parameters. The simulation results are rendered in plots

shown in Figure 9.27, indicating a greater amplitude and phase change for the same optical

parameters at higher frequencies. Note that at the same condition, the higher-frequency signal

attenuates more compared to the low-frequency signal, which indicates that the penetration

depth of the diffuse photon density wave over tissue depends on the modulation frequency and

itwill bemore shallow at frequencies approaching 1GHz as opposed to being deeper into tissue

at frequencies approaching 100MHz. This result affirms that spectroscopic information could

be extracted about various layers of the multi-layer tissue structures. As shown in Figure 9.2, a

receivedmodulated signal at GHz range can yield the upper layer (fat layer) information, while

the photon density wave modulated in MHz range can penetrate both the fat layer and muscle

layer so that it will bring the information combined with the fat layer and muscle layer. Hence,

by taking a multi-frequency measurement, one can identify this multi-layer structure and

extract the optical parameters on both layers.

However, as attenuation increases at the higher frequency, a reduction in signal-to-noise

ratio is experienced, which leads to a greater potential for phase measurement error. This issue


is highlighted in Figure 9.28(a), where measurements of phase have a larger spread as the

amplitude decreases. Moreover, for typical clinical concentrations of absorbing dyes (such as

nigrosin), a higher attenuation is also measured. Therefore to achieve broadband detection, a

flat response over a wide frequency range of laser sources and detectors is crucial. This flat

frequency response allows one to speed up the extraction process without going through the

two-tier calibration processes of the network analyser and multi-wavelength optical link

system. Moreover, accurate measurements of amplitude and phase over various broadband

systems require careful design of high optical sources (i.e. flat laser driver, comparable laser

diode responsivities and reduced laser diode RIN levels) and the optical receiver (i.e. good

responsivity over NIR wavelengths of interest, flat trans-impedance gain and broadband RIN

cancellation techniques). Some of these aspects are well established in the microwave

-6

-5

-4

-3

-2

-1

0

120010008006004002000

Frequency (MHz)

Amplitude attenuation

(dB)

µ ′s = 8/cm µ ′s = 9/cm µ ′s = 10/cm

µ ′s = 8/cm µ ′s = 9/cm µ ′s = 10/cm

(a)

0

10

20

30

40

50

120010008006004002000

Frequency (MHz)

Phase shift (º)

(b)

Figure 9.27 Frequency dependence for different phantom parameters: (a) amplitude change and

(b) phase shift when the reduced scattering coefficient ðm0sÞ changes from 7 cm�1 to 8, 9 and 10 cm�1

respectively. The simulation assumes that the absorption coefficient is constant at 0.04 cm�1. The infinite

boundary condition and source-detector separation of 3 cm are applied


photonics community and relevant examples are to be discussed next as part of a broadband

system realization.

9.5 Design and Realization Challenges of Broadband NIR OpticalSpectroscopy Systems

As was discussed earlier, microwave photonics can also be applied to optical spectroscopy,

where diffused photon waves will suffer different levels of scattering and absorption as a

function of various optical wavelengths for tissue depending on the level of oxygenated and

Figure 9.28 Impact of operation at higher frequency on the phase measurement accuracy. (a)

Comparison of measured (dot) and theoretical (line) phase-delay expectation of a broadband frequen-

cy-domain instrument. (b) Comparison of the measured (dot) versus expected (line) attenuation due to an

absorbing dye (nigrosin) at a modulation frequency of 500MHz and a wavelength of 782 nm [13].

Reprinted with permission from T. H. Pham, O. Coquoz, J. B. Fishkin, E. Anderson, and B.Tromberg,

‘‘Broad Bandwidth Frequency Domain Instrument for Quantitative Tissue Optical Spectroscopy,’’ Rev.

Sci. Inst. 71, 2500–2513 (2000). � 2000, American Institute of Physics


de-oxygenated haemoglobin and tissue structure. Moreover, since scattering increases at a

highermodulation frequency, then the diffused photonwave penetration in the tissue is less and

multi-level data could be collected from a microwave modulation frequency of 1GHz

compared to 100MHz.Microwave photonics techniques are explored here for the development

of custom-designed optical transmitters to be employed for broadband spectroscopic sys-

tems [55, 56].

9.5.1 Broadband Multi-frequency Instrument

The overall system block diagram of the proposed broadband NIR spectroscopy system is

shown in Figure 9.29. Different laser diodes (677 nm, 785 nm, 830 nm and 977 nm) are

sequentially modulated using RF power from a network analyser using an SP4T electrical

switch. A custom-designed laser driver modulates high-power laser diodes with a flat

frequency response up to 1GHz, while an optical switch is used for different source positions

(N positions). Detected light from different detection positions (M positions) on the turbid

medium is collected by a fibre bundle and received by an optical receiver (e.g. APD C5658,

Hamamatsu, Inc.). The network analyser receives the RF signal and forward or backward

scattering parameters are extracted using the measured amplitude and phase at each wave-

length. The first design is based on high-power laser diodes packaged in a transistor outline

Figure 9.29 Simplified block diagram of a broadband multi-wavelength NIR spectroscopy system

using a commercially available (HP 8753ES) vector network analyser. The four-wavelength system is

based on high-power fibre-coupled laser diodes that are driven by a broadband high-current laser driver,

electromechanical optical switches forM source andN detector positions, and a balanced optical receiver.

The source and detector are separated for distances of up to 3 cm in a turbid medium


(TO) can (cf. Figure 9.30). DC biasing of the laser diodes for low- and high-power is achieved

by a power transistor (Darlington, TIP125), where a DC biasing current of up to 1.2 A can be

achieved. The active laser driver is employed to provide an RF driver design requirement of

45mA in the laser junction. This drive RF current to the laser diode is achieved using an HJ-

FET from NEC (NE6510179A). Details of this design are discussed elsewhere [54] and a

photograph of the four wavelength optical sources and broadband high-current laser driver

boards placed in front of the APD optical receiver is shown in Figure 9.30. A pre-emphasis

circuit based onRC speed-up is used to overcome the bandwidth limitation of the transistor and

laser diode at higher frequencies and extend the system operation up to 1GHz. Insertion loss

and phase of the system (S21) in free space is depicted in Figure 9.31. Note a flatness of�1.5 dB

and phase linearity of �5� over a decade bandwidth is observed.

An optical link is established using the 785 nm optical source, free-space optical attenuator

(OD 1), and the optical receiver (APD, C5658, Hamamatsu, Inc.) to evaluate the performance

of the active driver for a TO-can package. The overall optical link performance is shown in

Figure 9.32, where a linear phase response is observed; however, the amplitude is attenuated at

higher frequencies at a rate of 25 dB/decade. This performance indicates that there is a

bandwidth limitation due to the laser diode operating at 785 nm (typically 40 dB/decade

after relaxation oscillation frequency) in addition to package parasitics associated with the

TO can (typically 20 dB/decade after the corner frequency). A similar behaviour is required

for the other three wavelengths to avoid calibration corrections using two-tier de-embedding

of the measured data to extract the optical parameters of the turbid medium under test

properly.

Figure 9.30 Overall view of the four-wavelength laser diode transmitter assembly and optical receiver.

An APD is shown on the left and four laser diodes are mounted on the driver on the right. Custom-made

four-laser diode drivers are in a single compact form


Figure 9.32 Transmission response of the custom-designed optical link using a laser at 785 nm,where a

frequency roll-off is observed due to the bandwidth limitation of a TO-can and high-power laser

Figure 9.31 Transmission response of a custom-designed high-current laser driver. Note the flatness of

the frequency response, especially at high frequencies. This response is obtained by an RC speed-up

circuit, which pumpsmore current at higher frequencies to compensate for the bandwidth limitation of the

high-current transistor and semiconductor laser


The TO-can package parasitics are eliminated when the approach described above is

extended to a fully integrated optical transmitter with a C-submount package. For operation

at higher frequencies of up to 3GHz, the TO can will contribute to significant roll-off, which

cannot be corrected by using RC speed-up circuits. Therefore, a C-submount laser diode has

been chosen because of its good output power capability while satisfying the high-speed

performance. A laser diode at 820 nm available on a C-submount is employed in the active

driver module, where a monitoring photodiode and TE cooler have been assembled to monitor

the output optical power and keep the temperature of the laser diode constant as depicted in

Figure 9.33(a). As seen in this picture of the optical transmitter, each component is properly

Figure 9.33 Depiction of a packaged broadband optical source: (a) photograph of the optical source

module at 820 nm with an active driver circuit; (b) measured frequency response of the laser driver over

the frequency range from 100MHz to 3GHz


labelled and a 50/125 mmmultimode fibre is employed for light coupling. A custom-designed

laser diode driver can provide up to 1ADCbias current for a laser diode and has a flat frequency

response up to 3GHz (cf. Figure 9.33(b)). An optical link is established using the active optical

transmitter, an OD optical attenuator and an optical receiver (APD, C5658, Hamamatsu, Inc.).

The overall optical link performance is shown in Figure 9.34, where a flat frequency response is

observed except around 580MHz. This discrepancy at 580MHz is due to a resonance in the

active driver portion of the optical transmitter. This link performance is used to calibrate the

link performance through biological tissue and the extraction of scattering and absorption

parameters of phantom experiments.

-20

-10

0

10

20

30

40

50

1000900800700600500400300200100

Frequency (MHz)

Am

plitu

de (

dB)

-200

-150

-100

-50

0

50

100

150

200

Pha

se (

º)

Amplitude

Phase

Figure 9.34 The overall optical link performance with flat gain and linear phase at all frequencies,

except for a self-resonance frequency of about 580MHz

Optical transmitter

Optical receiver (APD)

ANA

Intralipid solution

Figure 9.35 Experimental set-up for intra-lipid solution measurements


9.5.2 Intralipid Experiments using a Broadband System

To demonstrate the advantages of a broadband system in terms of a greater spatial resolution

and sensitivity at higher frequencies, results of a number of experiments are discussed here.

These experiments are conducted using a liquid phantom (i.e. intralipid), which is a container

primarily filled withmilk as shown in Figure 9.35. The system amplitude and phase response to

optical parameter changes of the liquid phantom are measured. The experimental set-up is

shown in Figure 9.35. The optical transmitter fibre and the receiver fibre are submerged in the

intralipid solution. The distance between the transmitter fibre tip and the receiver fibre tip is

3 cm. Before adding more intralipid, the amplitude and phase response were measured as a

baseline. The results of three separate experiments are reviewed next.

-20

-18

-16

-14

-12

-10

-8

-6

-4

-2

0

2

1400120010008006004002000

Frequency (MHz)

Amplitude difference (dB)

Amp1

Amp2

Amp3

Amp4

Amp5

Amp6

Amp7

Amp8

Amp9

(a)

0

10

20

30

40

50

60

70

80

90

1400120010008006004002000

Frequency (MHz)

Phase shift (º)

Phase1

Phase2

Phase3

Phase4

Phase5

Phase6

Phase7

Phase8

Phase9

(b)

Figure 9.36 System response to the scattering coefficient changes of a liquid phantom: (a) normalized

amplitude attenuation; (b) normalized phase shift


9.5.2.1 Experiment 1

A small portion of intralipid was added to the solution to change the scattering parameter

and the new amplitude and phase were recorded. Equal amounts of intralipid were added

eight times and the amplitude and phase changes were recorded each time. The amplitude

and phase responses were collected over the frequency range from 100MHz to 1.2 GHz and

the relative performance with respect to a baseline are shown in Figure 9.36, where the rapid

decrease in the amplitude after 1GHz is due to the frequency response of the APD. Also, a

response notch between 558MHz and 761MHz is observed, which is caused by the

frequency response of the custom-designed laser driver. The phase response is hard to read

since the phase difference between each curve is very small compared to the phase shift for

different frequencies.

The normalized data plotted in Figure 9.36 clearly shows that with larger intralipid

concentrations – and hence larger scattering coefficients – there is an increase in amplitude

attenuation and phase shift. From the figure one can also tell that at each intralipid addition,

higher-frequency data always show a larger signal change. The benefit of higher-frequency

measurement is shown more clearly in Experiments 2 and 3.


The system amplitude and phase response to absorbers with different absorption coefficients

was measured. The set-up is shown in Figure 9.37, which is similar to the set-up of Experiment

1 except that cylindrical absorbers with different absorption coefficients were placed in the

middle of the distance between the transmitter fibre and the receiver fibre. Three absorbers with

the same diameter of 4.6mm and different absorption parameters (i.e. absorber1

absorber2 < absorber3) were used. The absorber used was India ink, a well-known reference

material.

The amplitude and phasewere again normalizedwith respect to the baseline data and relative

results are plotted in Figure 9.38. The result shows that the one-wavelength NIR spectroscopy

system can easily identify these three absorbers and themulti-frequencymeasurement gives us

Figure 9.37 Experimental set-up for measuring optical parameters of a turbid medium dominated by

different absorption parameters


a clear image that the absorber with higher absorption coefficient causes more amplitude

attenuation and phase shift.

The data in Figure 9.38 also show the benefit of employing higher-frequencymeasurements.

To make this argument more obvious, the amplitude and phase response at 144MHz

(a frequency which has quite often been reported as part of single frequency I/Q receiver)

and 1GHz (a high frequency that is being advocated by the author) are compared in Figure 9.39.

At 144MHz, the amplitude attenuation caused by the three absorbers is less than 0.1 dB and the

phase shift is less than 2.2�, whereas at 1GHz the amplitude attenuation is around 1.5 dB and

phase shifts around 9�, which proves that higher-frequencymeasurement has higher sensitivity

to detect absorption change in a turbid medium.

-1.8

-1.6

-1.4

-1.2

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

1400120010008006004002000

Frequency (MHz)


(dB)

Ph

ase

shif

t (º

)

Absorber1

Absorber2

Absorber3

(a)

(b)

0

2

4

6

8

10

12

1400120010008006004002000

Frequency (MHz)

Absorber1

Absorber2

Absorber3

Figure 9.38 System response to different absorbers: (a) normalized amplitude attenuation; (b)

normalized phase shift



The system amplitude and phase response to absorbers (e.g. India ink of a set concentration)

with different sizes of container were measured. The set-up is identical to the set-up shown in

Figure 9.37. This time, two cylindrical absorbers with the same absorption coefficient but of

different diameters (straws of 4.6mm and 5.5mm) were placed in the middle of the distance

between the transmitter fibre and the receiver fibre. The distance between these two fibres was

still 3 cm. The normalized amplitude and phase responses of these two absorbers are shown in

Figure 9.40. As expected, the 5.5mm absorber causes more amplitude attenuation and phase

shift than the 4.6mm absorber. Again, the higher-frequency measurement is more sensitive to

-1.6

-1.4

-1.2

-1

-0.8

-0.6

-0.4

-0.2

0

321

Absorber #


(dB)

144MHz

1GHz

(a)

0

1

2

3

4

5

6

7

8

9

10

321Absorber #

Phase shift (º)

144MHz

1GHz

(b)

Figure 9.39 Sensitivity of the measured signal to the absorption parameter of three different absorbers

at 144MHz and 1GHz: (a) amplitude change and (b) phase shift


the size difference, which indicates that the higher-frequency measurement has better spatial

resolution than the low-frequencymeasurement. The ability to distinguish the 4.6mmdiameter

from the 5.5mm diameter is only observable at frequencies approaching 1000MHz in

amplitude.

9.5.3 Extraction of Phantom Optical Parameters using a Broadband System

A phantom made from a polymer whose optical parameters resemble biological tissue (i.e.

breast) is used in these experiments. The set-up is shown in Figure 9.41. The solid phantom has

-1

-0.8

-0.6

-0.4

-0.2

0

0.2

1400120010008006004002000

Frequency (MHz)


(dB)

Size1: 4.6mm

Size2: 5.5mm

(a)

0

1

2

3

4

5

6

7

1400120010008006004002000

Frequency (MHz)

Phase shift

(º)

Size1: 4.6mm

Size2: 5.5mm

(b)

Figure 9.40 System response to absorbers with different sizes: (a) normalized amplitude attenuation

and (b) normalized phase shift


homogeneous optical parameters, that is an absorption coefficient and a scattering coefficient.

The transmitter fibre and the receiver fibre are placed on the same surface of the solid phantom.

The amplitude and phase responses at two distances (d1¼ 1 cm, and d2¼ 1.5 cm) were

measured.

The amplitude attenuation and phase difference of the two distances have been calculated as

follows

Amplitude attenuation ðdBÞ ¼ Amplituded1ðdBÞ�Amplituded2ðdBÞPhase difference ðdegreeÞ ¼ Phased1ðdegreeÞ� Phased2ðdegreeÞ

and the results are shown in Figure 9.42. To extract the optical parameters, a Matlab program

was developed based on the relationship between the amplitude and the phase of the received

signal and optical parameters of the phantom as shown in Equations (9.31) and (9.32). The

measured data are employed in the program to extract the absorption and reduced scattering

coefficients. Measurements of relative amplitude and phase changes for two distances are

shown in Figure 9.42.

This measurement result could be used for optical parameter extraction. In a single-

frequency instrument (I/Q system), only the amplitude and phase signal related to this specific

frequency can be acquired, and hence one can extract absorption and reduced scattering

coefficients based on only one frequency. We call this kind of extraction single-frequency

extraction. However, with a frequency sweptmode using amulti-frequency instrument, such as

a network analyser, the amplitude and phase signals of multi-frequencies can be measured at

the same time. The measured results can be fitted to a theoretical model and the absorption and

scattering coefficients are extracted; we call this process broadband extraction. The advantage

of this process is that one can not only obtain information from various layers, but the

performance anomaly due to the in-frequency resonance can be averaged out or even removed.

To demonstrate the strength of this broadband frequency domain system over the CW

measurement system, an experiment is conducted using a known phantom resembling

breast tissue. In this experiment, 101 frequency points have been chosen between 100MHz

and 1000MHz and both broadband extraction and single-frequency extraction have been

performed. The optical parameters are extracted by relating the measured amplitude and

Optical transmitter

Optical receiver (APD)

ANA

Solid phantom

Fibres

Figure 9.41 Experimental set-up for solid phantom measurements


phase of the scattered light from the phantom using Equations (9.31) and (9.32). The

single-frequency extraction of absorption and scattering parameters as a function of

frequency is compared and depicted in Figure 9.43 against the manufacturer�s data. Note

that the extracted single-frequency results show an average error of 35.2% for ma and

23.6% for m0s.

This extraction error is typical of the inaccuracy encountered in I/Q demodulators, when

optical loss is significant. The benefits of the broadband measurement system become more

evident when extraction is correlated over a broader bandwidth. For broadband extraction,

0

2

4

6

8

10

12

14

16

18

20

120010008006004002000

Frequency (MHz)


(dB)

(a)

0

5

10

15

20

25

30

35

40

45

50

120010008006004002000

Frequency (MHz)

Phase difference

(º)

(b)

Figure 9.42 Measurements of relative amplitude and phase change as a function of frequency for a solid

phantom at two distances: (a) amplitude change and (b) phase shift


different frequency ranges have been selected to extract the optical coefficients, e.g. 300 MHz

range (100 400MHz, 400 700MHz, 700 1000MHz), 450MHz range (100 550MHz,

550 1000MHz) and 900MHz (100MHz 1GHz) range. All the extracted values are

summarized in Table 9.2. As shown in this table, for a larger frequency range, the extracted

values tend to be closer to the manufacturer�s values. Moreover, the accuracy of broadband

extraction is much better than single-frequency extraction.

0

0.04

0.08

0.12

0.16

0.2

0.24

0.28

900700500300100

Manufactured value

Extracted valueµa

(cm-1)

µs′(cm-1)

(a)

(b)

Frequency (MHz)

0

3

6

9

12

15

18

900700500300100

Frequency (MHz)

Manufactured value

Extracted value

Figure 9.43 Extracted ma (a) and m0s (b) by using single-frequency extraction, which results in an

average error of 35.2% for ma and 23.6% for m0s


9.6 Conclusions

Microwave photonics techniques have become established for telecommunication and radar

systems in the last 20 years; however, the benefits of RF andmicrowave photonic technology in

biomedical applications have not been fully explored. This chapter has presented two separate

applications of microwave photonics to medical imaging. In particular, it is stipulated that a

higher-spatial resolution and broadband ultrasound imaging could be developed using an

optical hydrophone, where development of a fibre-optic hydrophone probe was described.

Although the probe was designed to operate at frequencies up to 100MHz it has not yet been

tested at 100MHz frequency. Once fully developed, the FO probe will constitute an effective

measurement tool allowing the need for spatial averaging corrections to be eliminated.A power

budget calculation of the fibre sensor indicated that a relatively high-power (1000mW) laser

source is essential to achieve a sufficiently high signal-to-noise ratio. Experiments validated the

analytical results and broadband-measured sensitivity of about 500mV/MPa was achieved in

the frequency range up to 60MHz without any spatial averaging correction required. This

performance at the moment supersedes the commercially available needle hydrophones.

Microwave photonics techniques are also applied to the eventual development of a

broadband NIR spectroscopy system to achieve real-time imaging of biological tissues with

a millimeter scale spatial resolution. In particular, an active driver system, which was

developed for a TO-can based laser system operating at wavelengths of 677 nm, 785 nm,

830 nm and 977 nm, was extended to C-submount based optical transmitters. In particular, an

active laser driver using a C-submount laser diode at 850 nm was developed using an active

laser driver for a high-power laser diode. The laser driver is realized using hybrid techniques.

The optical system was designed for a flat frequency response up to 3GHz, where spatial

resolution of a few millimeters is predicted. Both broadband extraction and single-frequency

extraction have been performed to extract optical parameters ma and ms for a phantom that

resembles breast tissue as a turbid media. Results show that the accuracy of broadband

extraction is much better than single-frequency extraction.

Table 9.2 Broadband extraction and percentage error results.

Extracted

values (cm�1)

Manufacturer�svalues (cm�1)

Error

(cm�1)

Error

(%)

Frequency range from 100MHz 400MHz ma 0.0488 0.045 0.0038 8.44

m0s 11.0769 10 1.0769 10.77


m0s 9.0779 10 0.9221 9.22

Frequency range from 700MHz 1GHz ma 0.0436 0.045 0.0014 3.11

m0s 9.4321 10 0.5679 5.68


m0s 10.2672 10 0.2672 2.67


m0s 9.4655 10 0.5345 5.35


m0s 9.6711 10 0.329 3.29


Acknowledgements

Material covered in this chapter is due to the contribution of a number of colleagues, past and

present graduate students. The intellectual influence of, and fruitful discussions with, my

colleagues Prof. Emeritus Britton Chance, Prof. Peter Lewin and Prof. Kambiz Pourrezaei

from the School of Biomedical Engineering, Drexel University, is greatly recognized. Finally,

without the tireless work ofmy students, the required understanding and scientific research that

has now led to the writing of this chapter would not have been feasible. I would like to

acknowledge the contributions of Dr Sumet Umchid, Chenpeng Mu, Rupa Gopinath, Karthik

Srinivasan andDo-YoonKim to this chapter. Finally, support of theNational Institute of Health

(NIH) is greatly appreciated.

References

[1] Papers in Front End Opto-Electronics for Future Radio Communications, Workshop WM1, 2004 IEEE Radio

and Wireless Conference, September 2004.

[2] Papers in Beamforming Techniques for Active Phased Array Antennas Based Communication Satellites, in the

IEEE 1999 International Microwave Symposium, Anaheim, CA, USA, June 1999.

[3] M. Berson, J. M. Gr�egoire, F. Gens, J. Rateau, F. Jamet, L. Vaillant, F. Tranquart and L. Pourcelot, ‘‘High

frequency (20 MHz) ultrasonic devices: advantages and applications’’, European Journal of Ultrasound,

vol. 10(1), pp. 53–63, 1999.

[4] AIUM, ‘‘Acoustic output measurement standard for diagnostic ultrasound equipment’’, Laurel, MD, USA, 1998.

[5] FDA, Revised FDA 510(k) ‘‘Information for manufactures seeking marketing clearance of diagnostic ultrasound

systems and transducers’’, September 1997.

[6] E. G. Radulescu, P. A. Lewin, J. Wojcik and A. Nowicki, ‘‘Calibration of ultrasonic hydrophone probes up to

100MHz using time gating frequency analysis and finite amplitude waves’’, Ultrasonics, vol. 41, pp. 247–254,

2003.

[7] R. A. Smith, ‘‘Are hydrophones of diameter 0.5 mm small enough to characterize diagnostic ultrasound

equipment?’’ Phys. Med. Biol., vol. 34(11), pp. 1593–1607, 1989.

[8] G. R. Harris, ‘‘Medical ultrasound exposure measurements: update on devices, methods, and problems’’,

Ultrasonics Symposium, pp. 1341–1352 (1999).

[9] J. Staudenraus and W. Eisenmenger, ‘‘Fibre-optic probe hydrophone for ultrasonic and shock-wave measure-

ments in water2, Ultrasonics, vol. 31(4), pp. 267–273, 1993.

[10] E.M. Sevick, B. Chance, J. Leigh and S. Nioka, ‘‘Quantitation of Time-resolved and Frequency-resolved Optical

Spectra for the Determination of Tissue Oxygenation’’, Anal. Biochem., vol. 195(2), pp. 330–351, 1991.

[11] B. Beauvoit, T. Kitai and B. Chance, ‘‘Contribution of the Mitochondrial Compartment to the Optical Properties

of the Rat Liver: A Theoretical and Practical Approach’’, Biophys. J., vol. 67(6), pp. 2501–2510, 1994.

[12] S. Thornsen andD. Tatman, ‘‘Physiological and Pathological Factors ofHumanBreastDisease that Can Influence

Optical Diagnosis2, Ann. N. Y, Acad. Sci., vol. 838, pp. 171–193, 1998.

[13] T. H. Pham, O. Coquoz, J. B. Fishkin, E. Anderson and B. Tromberg, ‘‘Broad Bandwidth Frequency Domain

Instrument for Quantitative Tissue Optical Spectroscopy’’, Rev. Sci. Inst., vol. 71, pp. 2500–2513, 2000.

[14] B. J. Tromberg, B, O. Coquoz, J. B. Fishkin, T. Pham, E. R. Anderson, J. Buffer et al., ‘‘Non-invasive

Measurements of Breast Tissue Optical Properties Using Frequency-domain Photon Migration’’, Philos. Trans.

R. Soc. Lond. B. Biological Sciences, vol. 352(1354), pp. 661–668, 1997.

[15] A. Duncan, T. L. Whitlock, M. Cope and D. T. Delpy, ‘‘A Multi-wavelength, Wideband, Intensity Modulated

Optical Spectrometer for Near Infrared Spectroscopy and Imaging’’, SPIE, vol. 1888, pp. 248–257, 1993.

[16] J. E. Parsons, C. A. Cain and J. B. Fowlkes, ‘‘Cost-effective assembly of a basic fiber-optic hydrophone for

measurementofhigh-amplitudetherapeuticultrasoundfields’’,J.Acoust.Soc.Am.,vol.119, pp.1432–1440,2006.

[17] W. B. Spillman, Jr. and R. L. Gravel, ‘‘Moving fiber-optic hydrophone’’,Optics Letters, vol. 5, No. 1 pp. 30–31,

January 1980.

[18] W. B. Spillman, Jr., ‘‘Multimode fiber-optic hydrophone based on a schlieren technique’’,Appl. Opt., vol. 20, pp.

465–470, 1981.


[19] W. B. Spillman, Jr. and D. H. McMahon, ‘‘Frustrated-total-internal-reflection multimode fiber-optic hydro-

phone’’, Appl. Opt., vol. 19, pp. 113–117, 1980.

[20] G. B. Hocker, ‘‘Fiber-optic sensing of pressure and temperature’’,Appl. Opt., vol. 18, No. 9, pp. 1445–1448, 1978.

[21] A.S. Daryoush, H.W. Li, M. Kaba, G. Bouwmans, D. Decoster, J. Chazelas and F. Deborgies, ‘‘Passively

Temperature Stable Opto-electronic Oscillators Employing Photonic Crystal Fibers?’’, Journal of the European

Microwave Association, vol. 3, Issue 3, pp. 201–209, September 2007.

[22] A. M. Yurek, ‘‘Status of Fiber Optic Acoustic Sensing’’,Optical Fiber Sensors Conference, vol. 8, pp. 338–341,

1992.

[23] N. Lagakos and J.A. Bucaro, ‘‘Linearly Configured Embedded Fiber-optic Acoustic Sensor’’, Journal of

Lightwave Technology, vol. 11, pp. 639–642, April 1993.

[24] G.E.McDearmon, ‘‘TheoreticalAnalysis of a Push-Pull FiberOpticHydrophone’’, J. LightwaveTech., vol. LT-5,

pp. 647–652, 1987.

[25] P. C. Beard,A.M.Hurrell, andT.N.Mills, ‘‘Characterization of a polymer film optical fiber hydrophone for use in

the range 1 to 20 MHz: A comparison with PVDF needle and membrane hydrophones’’, IEEE Transactions on

Ultrasonics, Ferroelectrics, and Frequency Control, vol. 47, pp. 256–264, 2000.

[26] P. Morris, A. Hurrell and P. Beard, ‘‘Development of 50 MHz Fabry-Perot type fiber optic hydrophone for the

characterization of medical ultrasound fields’’, Proceedings of the Institute of Acoutics, vol. 18, pp. 717–725,

2006.

[27] V. Wilkens, C. Koch, and W. Molkenstruck, ‘‘Frequency response of a fiber-optic dielectric multilayer

hydrophone’’, Ultrasonics Symposium, IEEE, vol. 2, pp. 1113–1116, 2000.

[28] J. A. Bucaro and T. R. Hickman, ‘‘Measurement of sensitivity of optical fibers for acoustic detection’’, Applied

Optics, vol. 18, No. 6, pp. 938–940, 1979.

[29] P. Fomitchov and S. Krishnaswamy, ‘‘Response of a Fiber Bragg-Grating Ultrasound Sensor’’, Optical

Engineering, vol. 42, No 4, pp. 956–963, 2003.

[30] Bai-Ou Guan, Hwa-Yaw Tam, Sien-Ting Lau and Helen L. W. Chan, ‘‘Ultrasonic hydrophone Based on

Distributed Bragg Reflector Fiber Laser’’, IEEE Photonics Technology Letters, vol. 17, no. 1, pp. 169–171,

January 2005.

[31] R. Gpoinath, P. Arora, G. Gandhi, L. Bansal, A.S. Daryoush, P.A. Lewin and M. El-Sherif, ‘‘Broadband Fiber

Optic Hydrophone Sensors for Ultrasound Applications’’, MWP 2008 Proc., International Topical Meeting on

Microwave Photonics, Gold Coast, Australia, 2008.

[32] K. Srinivasan,‘‘Noise Cancelled Optical Receivers in Fiber Optic Hydrophone up to 100MHz?’’, M.S. Thesis,

Drexel University, USA, 2007.

[33] Sumet Umchid,‘‘Development of Calibration Techniques for Ultrasonic Hydrophone Probes in the Frequency

Range from 1 to 100 MHz’’, Ph.D. Dissertation, Drexel University, Philadelphia, PA, USA, 2007.

[34] P.A. Lewin, C.Mu, S.Umchid,A.Daryoush andM.El-Sherif, ‘‘Acousto-optic, PointReceiverHydrophoneProbe

for Operation up to 100 MHz’’, Ultrasonics, vol. 43, Issue 10, pp. 815–821, December 2005.

[35] M. Cutler, ‘‘Transillumination of the breast’’, Surg. Gynecol. Obstet., vol. 48, pp. 721–727, 1929.

[36] E. M. Sevick, B. Chance, J. Leigh and S. Nioka, ‘‘Quantitation of time-resolved and frequency-resolved optical

spectra for the determination of tissue oxygenation’’, Anal. Biochem., vol. 195(2), pp. 330–351, 1991.

[37] B. Beauvoit, T. Kitaiand B. Chance, ‘‘Contribution of the mitochondrial compartment to the optical properties of

the rat liver: a theoretical and practical approach’’, Biophys. J., vol. 67(6), pp. 2501–2510, 1994.

[38] S. Thornsen and D. Tatman, ‘‘Physiological and pathological factors of human breast disease that can influence

optical diagnosis’’, Ann. N. Y, Acad. Sci,. vol. 838, pp. 171–193, 1998.

[39] A. Zourabian, A. Siegel et al., ‘‘Trans-abdominal monitoring of fetal arterial blood oxygenation using pulse

oximetry’’, J. Biomed. Opt., vol. 5(4), pp. 391–405, 2000.

[40] S. R. Arridge, ’’Optical tomography in medical imaging’’, Inverse Probl., vol. 15, pp. 41–93, 1999.

[41] D. A. Boas et al., ‘‘The accuracy of near imfrared spectroscopy and imaging during focal changes in cerebral

hemodynamics’’, NeuroImages, vol. 13, pp. 76–90, 2001.

[42] Y. Lin et al., ‘‘Noninvasive, low-noise, fast imaging of blood volume and deoxygenation changes inmuscles using

light-emitting diode continuous-wave imager’’, Rev. Sci. Instrum., vol. 73(8), pp. 3065–3074, 2002.

[43] Y. Yang, H. Liu, X. Li and B. Chance, ‘‘Low-cost frequency-domain photon migration instrument for tissue

spectroscopy, oximetry, and imaging’’, Opt. Eng., vol. 36(5), pp. 1562–1569, 1997.

[44] M. S. Patterson et al., ‘‘Time resolved reflectance and transmittance for the non-invasive measurement of tissue

optical properties’’, Applied Optics, vol. 28, pp. 2331–2336, 1989.


[45] S. L. Jacques, ‘‘Time-resolved reflectance spectroscopy in turbid tissues’’, IEEETrans. Biomed.Eng., vol. 36, pp.

1155–1161, 1989.

[46] B. C. Wilson and S. L. Jacques, ‘‘Optical reflectance and transmittance of tissues: principles and applications’’,

IEEE J. Quantum Electron., vol. 26, pp. 2186–2199, 1990.

[47] G. Eason et al., ‘‘The theory of the backscattering of light by blood’’. J. Phys., vol. 11, pp. 1463–1479, 1978.

[48] D. Grosenick, H. Wabnitz, H. Rinneberg, K. T. Moesta and P. Schlag, ‘‘Development of a time-domain optical

mammograph and first in vivo applications’’, Appl. Opt., vol. 38, pp. 2927–2943, 1999.

[49] B. Chance et al., ‘‘Phase measurement of light absorption and scatter in human tissue’’, Rev. Sci. Instrum.,

vol. 69(10), pp. 3457–3481, 1998.

[50] T. Tu et al., ‘‘Analysis on performance and optimization of frequency-domain near-infrared instruments’’,

J. Biom. Opt., vol. 7(4), pp. 643–649, 2002.

[51] Y. Chen, C. P. Mu, X. Intes and B. Chance, ’’Signal-to-noise analysis for detection sensitivity of small absorbing

heterogeneity in turbidmediawith single-source anddual-interfering-source’’,Opt. Express, vol. 9, pp. 212–224,

2001.

[52] M. Erickson et al., ‘‘Comparison of sensitivity for single-source and dual-interfering-source configurations in

optical diffusion imaging2, J. opt. Soc. Am. A, vol. 14(11), pp. 3083–3092, 1997.

[53] A. Knuttel, J. M. Schmitt and J. R. Knutson, ’’Improvement of spatial resolution in reflectance near-infrared

imaging by laser-beam interference’’, in ‘Time-Resolved Laser Spectroscopy in Biochemistry III’, J. R. Lakowicz

(Ed.), Proc. SPIE vol. 1640, pp. 405–416, 1992.

[54] A. Duncan, T. L. Whitlock, M. Cope and D. T. Delpy, ‘‘A multi-wavelength, wideband, intensity modulated

optical spectrometer for near infrared spectroscopy and imaging’’, SPIE, vol. 1888, pp. 248–257, 1993.

[55] C. Mu, D-Y Kim, U. Sunar, K. Pourrezaei and A. Daryoush, ‘‘Multi-wavelength NIR system for spectroscopy of

biomedical tissues’’, MWP 2003 Proc., International Topical Meeting on Microwave Photonics, Budapest,

Hungary, pp. 275–278, 2003.

[56] DoyoonKim,‘‘Design of laser diode driver for diffused photon near infrared imaging applications’’,M.S. Thesis,

Drexel University, Philadelphia, PA, USA, 2002.


microwave photonics || rf and microwave photonics in biomedical applications

Documents