This PDF file contains the front matter associated with SPIE Proceedings Volume 10311, including the Title Page, Copyright information, Table of Contents, and Conference Committee listing.

Optical imaging and spectroscopy use light emitted into opaque media such as human tissue to determine interior structure and chemical content, respectively, and have broad application to the field of medicine [1]. Few developments have improved medical diagnos- tics as much as the ability to noninvasively peer inside the body, and it is expected that newly developing optical imaging techniques will continue this trend. Optical imaging and spectroscopy, key components of optical tomography, center around the simple idea that light passes through the body in small amounts, emerging bearing clues about tissues through which it passed. Rapid progress over the past decade, made possible by the collective output of multiple laboratories and advancements in the opto-electronics field, have brought optical imaging to the brink of clinical usefulness. In recognition of this impending transition from lab curiosity to medical tool, it was felt that a comprehensive review was in order. This book, written in large part by members of the key laboratories responsible for recent optical advances, aims to fill that need by exploring state-of-the-art methods, hardware, and applications of optical tomography. It is intended to serve as both an introduction for those unacquainted with the field, as well as a reference for those actively involved. It is hoped that the reader will gain an appreciation of the power and breadth of optical tomography, as well as an understanding of the fundamental limitations and unsolved problems that need study before optical imaging becomes clinically viable. Lastly, we hope to communicate the excitement and enthusiasm felt by those studying optical imaging and spectroscopy, and demonstrate why the field is undergoing both an explosion of interest and rapid progress.

Within only a few years after its introduction in 1972, x-ray computed tomography (CT) established its important role in radiological diagnosis. Today, more than 20,000 clinical CT installations are in operation. CT technology is mature and its clinical use can be considered routine, but new scanning procedures and clinical applications are arising furtheron. Scanning of complete volumes in minimal time with spiral (helical) CT is one of the most important examples. Since the early nineties, conventional slice-by-slice scanning is more and more replaced by this new volume scanning technique. We review the principles of scanning and image reconstruction for axial CT and for spiral CT and present their performance characteristics and major new applications. CT started as a two-dimensional slice imaging modality; it now rapidly develops into a three-dimensional volume imaging technique.

Optics is perhaps the most precisely mathematicized of the physical sciences, with a pedigree stretching from Newton to Maxwell. The only comparable field of study with such an elegant and rigorous basis is quantum mechanics, whose history is much shorter. Unlike quantum mechanics, classical optics suffers from none of the phenomenological complications of the "problem of measurement, " yet like quantum mechanics the application of simple principles to complex real situations leads very quickly to mathematical difficulties that are well beyond current analytical and computational methods. For a situation as complex as the interaction of light with living tissue, approximate methods are required. The simplifications made in such approximations may at first sight seem so dramatic as to render the resultant model unrealistic, even though the computational complexity may still be high. It is thus of fundamental importance to compare and contrast such methods and to validate against experimental results.

This paper introduces the general principles involved in inverse problems in medical imaging, and describes the main theoretical principles behind Time-resolved Optical Absorption and Scattering Tomography (TOAST). The problem is viewed as the optimisation of an error-norm derived from correlated statistics of the time-dependent photon intensity at the surface of an object. The field is compared with Electrical Impedance Tomography (EIT). A comparison of inverse methods are made and several regularisation schemes are described.

Light propagation in random media can be described under rather general assumptions by a one-speed linear transport equation (BoLTzmANN-Equation)I,2. A general solution of this equation, in particular for the time-dependent case, has not been obtained yet. Various ap- proximation to this equation have been used in the past to treat the problem of light transport in turbid media for various geometries. Both time- and frequency-domain solutions for infinite media'--4, slab3,5 and cylindrical5,6 geometries have been reported.

In this paper, we discuss approaches our group has developed for the problem of imaging the interior of dense scattering media [1]. While our principal focus is on potential biomedical applications, we believe our methods are sufficiently general to have applications to other imaging problems as well. We begin our consideration of the imaging problem by assuming that the target medium of interest interacts with the penetrating energy source with sufficient strength to cause intense scattering. We further assume that for essentially all practical schemes, only the multiply scat- tered signal is measurable. One result of multiple scattering is that all the detected photons will have propagated above and below the plane in which the source and de- tector lie. Thus, it becomes necessary to explicitly consider volume functions whose spatial distribution will depend on the properties and geometry of the medium and on the geometry and type of illumination scheme. Measurement schemes which have been suggested include steady-state [2], ultrafast [3-5], and amplitude mod- ulated [6, 7] sources. Other schemes include holographic methods which have the potential advantages of directly yielding an image without the need for numerical reconstruction [8, 9]. In developing approaches to image reconstruction, our group has emphasized the first two of the four methods [10-16].

Most imaging schemes involve characterizing the interaction of an electromagnetic wave with a target medium. This interaction can be generally described by the wave equation: This equation relates the spatial variation (Laplacian) of the field to the electrical permittivity, e, and magnetic permeability, ix, of the medium. The latter quantities account for the induced alignment of the electrical dipole (polarization) and magnetic moment (magnetization) in the material by the propagating field, and they determine c, the speed of light, which is equal to 1/ Eµ . For a time-harmonic source, equation 1 reduces to the Helmholtz equation: where k, the wave number, is equal to wic, and w is the wave frequency in radians-s-1.

With the ever-growing use of optical techniques in medicine, both for diagnostic and for therapeutic uses, there is the need for accurate data on the optical properties of the various tissues concerned. Due to the complexity of biological tissue, the determina-

The optical properties of white matter human brain, canine prostate and pig liver were measured in the wavelength range 330-1100 nm. The measurements were carried out in native as well as in coagulated tissues. We used the double integrating sphere technique to provide reflection and transmission measurements and a special homogenising technique to prepare the tissue. The optical properties were evaluated using an inverse Monte- Carlo simulation, considering the geometry of the experimental set-up. All tissues show characteristic absorption bands at 420 nm and 550 nm, related to the strong absorption of haemoglobin. After coagulation the scattering increases drastically while absorption remains nearly unchanged. The anisotropy factor g increases with increasing wavelength and drops down slightly after coagulation. The wavelength behaviour of tissue scattering has been compared with theoretical calculations (Mie-theory), showing that ideal spheres with an diameter between 0.6 and 0.8 pm fit best to the experimental results.

There is an ever increasing use of optical methods in medicine in the areas of diagnostics, therapy and surgery. Examples of diagnostic use are the monitoring of blood oxygenation and tissue metabolism 1,2,3, laser doppler flow measurement', pulse oximetry5, detection of cancer by fluorescence methods6" etc, and more recently various suggested techniques for optical imaging9,10,11,12. Therapeutic uses include applications in laser surgery, laser angioplasty and ablation of tumours and in Photodynamic Therapy (PDT)". For these applications, there is a need to know the optical properties of the tissues concerned, in order to interpret and quantify diagnostic data, and to predict light distribution and absorbed dose for therapeutic use.

The effective attenuation coefficient of piglet lung was measured in vitro at 632.8 nm. Interstial fibres with isotropic tips were used to measure the fluence rate as a function of the distance from an isotropic light source. In vitro measurements at 632.8 nm on a lung that was insufflated with oxygen from 50 to 150 ml showed that the effective attenuation coefficient decreases as a function of the volume of air in the lung (at 50 ml /Jeff = 0.297 + 0.011 mnf1, at 100 ml lice 0.150 ± 0.007 mm-1, and at 150 ml /Jeff= 0.1136 + 0.015 mm-1). A single in vitro measurement at 790 nm at an insufflated lung volume of 100 ml gave a comparable result (ii ie = 0.175 + 0.004 mm-1). A ff decrease in effective attenuation coefficient with an ncrease in lung volume was explained by Mie-theory. The effective attenuation coefficient, calculated with 11, and g from Mie-theory, showed a deviation < 22% from the measured in vitro values.

The measurement of tissue optical properties is often required for proper design of therapeutic or diagnostic uses of light in medicine. The ability of light to spread into a tissue and the rate of light absorption by the tissue are two related but distinct processes. The two independent optical parameters which affect these processes are absorption and scattering. To understand light propagation, two independent optical measurements must be made. In this paper, we discuss the measurements of total diffuse reflectance and lateral spread of light in response to a point source of irradiance. In this paper, we call this technique "video reflectometry".

Considerable time and effort has been devoted to predictions of the extent of thermal damage owing to laser photocoagulations. To establish a suitable radiation dose requires a complete understanding of the optical-thermal response of tissue to laser radiation and an accurate assessment of the optical and thermal properties of tissue. This is a formidable and mostly unsolved problem. Even the best predictions and algorithms are no better than the accuracy of measurement of parameters used in the governing equations. One problem is the tissue optical properties. Typically these properties are determined from in vitro samples. From the wide range of values of absorption and scattering coefficients reported in the literaturel, it is difficult to judge the average optical properties of a particular in vivo tissue, much less extract values for a particular region that will be radiated. One solution eliminates dosimetry uncertainty by monitoring the progression of photocoagulation during laser radiation and stopping irradiation when a specified endpoint has been reached.

The model of the skin as a multy layered biotissue with varying optical properties for each layer such as absorption, scattering, and scattering anisotropy factor is considered. The Monte—Carlo method was used to evaluate the effects of anisotropic scattering and refractive index mismatch at the boundaries of the sample with regards to light distribution inside the skin during laser irradiation. Calculations were performed for some models of human skin and adjacent tissues for visible and UV ranges of wavelengths. These calculations are especially important for laser percutaneous irradiation of blood and for laser PUVA therapy.

In a vast amount of medical diseases the biochemical and physiological changes of soft tissues are hardly detectable by conventional techniques of diagnostic imaging (x- ray, ultrasound, computer tomography, and MRI). The detectivity is low and the technical efforts are tremendous. On the other hand these pathologic variations induce significant changes of the optical tissue parameters which can be detected. The corresponding variations of the scattered light can most easily be detected and evaluated by infrared diaphanoscopy, even on optical thick tissue slices.

The inspection of parts of the human body with the help of optical instruments has been used in medical diagnosis for more than a century, but gained limited applications only. In recent years however extended knowledge of tissue optics in particular /10 (absorption coefficient), As (scattering coefficient) and g (anisotropy factor), and improved sensor technology offered new solutions to this problem.

Visual lights have been used to detect breast lesions since the 1929 report by M. Cutler', who observed breast transilluinination in a darkened room. Other papers' followed in the 1930s. In 19515, work started again using hard copies of transmitted near-infrared(NIR) shadows.

Optical spectroscopy techniques are providing essential ways to characterize physical and chemical properties in living tissues and cells and then to monitor their functional changes occurring inside them. Thus they offer exciting possibilities for developing spectroscopic computed tomography in the optical region, especially in the red and near infrared regions, to acquire functional/physiological information through their unique and valuable reconstructed images. Nevertheless, its development and application in living tissues and systems are complicated and prevented practically by the diffuse nature of the image quality due to strong multiple scattering of light. In the present paper, we will review and discuss our recent studies on Coherent Detection Imaging (CDI) method which could provide at present one of the most reliable and feasible schemes for achieving the optical/spectroscopic computed tomography using various kinds of lasers for biomedical applications. This method is basically realized with the optical heterodyne detection technique, possessing both the properties of a highly directional antenna and an ultra—sensitive receiver, and the image reconstruction based on the projection slice theorem from sets of line integrals of laser absorption along a large number of rays crossing the object with the parallel beam geometry. It is also postulated that the three fundamental conditions should be satisfied in principle in order to establish the optical absorption computed tomography on the basis of the conventional projection slice theorem. We have demonstrated experimentally the ful— fillment of these fundamental conditions using several biological tissues as well as scattering and diffusive samples by virtue of the optical heterodyne detection technique. Hence the applicability of the CDI method was confirmed in highly scattering and turbid media such as living tissues and bodies in which an object is completely ob— scured from normal visual observation and from conventional direct detection tech— niques. In consequence, we have achieved experimentally the two—dimensional direct (projection) imaging with some test objects immersed in biological tissues and the reconstructive tomographic imaging with various in vitro and in vivo objects such as chicken leg, chicken egg and human tumor specimens together with some plants and new—born mouse head. The collimated optical beam from a cw laser operated usually less than 10 mW of the output power, selected suitably from a set of Ar, He—Ne, Kr, Ti:A1203 and Nd:YAG lasers in our CDI system, enabled us to display these images with the spatial resolution as good as approximately 200 tim to 700 tun in this experi— ment. Several typical results of these measurements are also shown and discussed in this paper.

The intensity of a collimated laser beam transmitted through a highly scattering medium is attenuated considerably due to photon losses caused by absorption and, much more, by scattering. However a small fraction of the collimated laser beam is transmitted through the scattering medium without any scattering event, but he intensity (Ic(x)) of the collimated beam in the depth (x) decreases with an exponential function depending on the sum of the integral absorption (µa) and scattering (Is) coefficients of the irradiated volume of the collimated laser beam: These "collimated", "geometric" or "ballistic" photons are propagating, according to the laws of the geometric optic, as though travelling through a non scattering medium like in pure water.

If the more or less conventional interferometric techniques are applied in vivo, movements of the subject might falsify the measurement. This led us to develop the dual beam interferometry technique which compensates for these movements. First in vivo tomographic images of the human eye are presented. The implications of the application of partial coherence techniques in dispersive media are discussed. Synthesizing tomograms from interfero- metrically obtained scans is rather time-consuming and therefore difficult to apply in vivo. Hence we discuss the application of spectral interferometry to optical coherence tomography. The spectral interferometry technique does not need moving components.

How can we use visible light to obtain a three-dimensional image of thick, living biological tissues such as the eye or skin? Confocal light microscopy can be used to obtain a stack of optical sections which are stored as digital images. These optical sections can then be processed in a digital computer to form three-dimensional reconstructions. Four-dimensional microscopy is defined as the three-dimensional reconstruction of specimens which are changing with time. At specific time intervals a set of optical sections are collected through the use of a confocal microscope. Each set of optical sections are then transformed into a three-dimensional reconstruction which corresponds to the individual stack of optical sections taken at a given time.

Airplanes in clouds, submarines in murky water, and cancerous tumors in breast tissue all have one thing in common, they are objects hidden in highly scattering media. Optical imaging through such media remains one of the most challenging problems in science and engineering, but the advent of ultrafast lasers and detectors, coupled with a range of time-resolved techniques, has led to recent breakthroughs. Seeing through highly scattering media, normally absorbing the vision by image blurring, is desirable in many contexts. Fog, aerosols, and dust can strongly impair the vision in the atmosphere from Rayleigh and Mie scattering. At high particle concentrations, multiple scattering becomes dominant. The infrared spectral region, where scattering by fog and dust becomes much smaller, is used in many viewing systems.

Techniques for time-resolved transillumination imaging are reviewed. Emphasis is put on diode laser techniques using time-correlated single-photon detection. Experiments on tissue phantoms and human breast cancer measurements are given, and data are interpreted aided by theoretical simulations of photon migration in tissue.

Ultrashort laser pulses passing through a random turbid medium are temporally spread into ballistic, snake, and diffuse components. The intensity and the speed of the ballistic pulse are found to depend on the scattering characteristics of the medium. The diffuse component can be approximated by the diffusion theory when the laser pulse propagates through a distance of more than 10 transport mean free path. The early arriving portion of the diffuse pulse, known as the snake component, consists of photons that propagate along zig-zag paths slightly off the straight path. The attenuation of the snake photon depends on the transport mean free path of the random medium and it decreases exponentially as its thickness increases. However, the snake photon attenuates much more slowly than the ballistic photon. Snake photons have to be used for time-resolved imaging through a thick biological tissue. Ultrafast time-resolved detection of ballistic and snake photons are illustrated for locating opaque and translucent objects hidden in highly scattering media. A thin slab of chicken fat tissue (2.5 mm thick) embedded inside a thick chicken breast tissue (40 mm) has been located by detecting snake photons at X.625 nm. Near infrared light is scattered less in tissue and can be used to image through a significantly thicker tissue. The current ultrashort pulse lasers and ultrafast time-resolved detection techniques can be potentially used to image tumors in a breast.

Techniques for non-invasive biomedical imaging include nuclear magnetic resonance, ultrasound, positron emission tomography, x-ray computed to- mography, and optical transillumination. Each of these methods has dif- ferent advantages and limitations and has found particular applications in medicine. Optical imaging of tissue offers the potential of a non-invasive diagnostic with non-ionizing radiation and the possibility of using spectro- scopic properties to distinguish tissue type and probe metabolic function.

Time resolved imaging involves detecting and measuring the flight times of photons diffusely transmitted through scattering media, and uses the photons with the shortest flight times to generate an image. Recent interest in time resolved methods has been motivated by the desire to develop medical imaging techniques which use harmless doses of visible or near-infrared radiation. In particular, a technique is sought which can provide a safe screening modality for the detection of breast disease. In this paper, the progress towards this objective is summarized, and the current prospects of developing a successful device are discussed.

Over the last few years there has been a vast improvement in our knowledge of light transport in turbid media. Although the basic physics of the transport process was previously well developed, the particular application of the theory to different relevant measurement protocols was not well understood. In this section the basic equations describing light transport in turbid media, using the frequency domain protocol, are reviewed (Sevik et al.). The proposal of using frequency domain methods in the context of photon density waves is expanded and examples of measurements are reported. Two complementary aspects of the frequency domain methods are presented: 1) the capability of the frequency domain technique to detect localized differences of the scattering and absorption coefficients, and 2) the possibility to resolve absorbing objects of different shape. In the photon density wave framework it is shown that objects of different shape scatter the photon density waves with different modalities, raising the possibility of distinguishing of the shape of objects immersed in the turbid medium from their characteristic diffraction pattern. The important issue of the additional advantage of using high-frequency methods to localize absorbing and scattering objects is also discussed. It is shown that the diffraction pattern is not very sharp, even at high frequency (Patterson et al.), but that the particular shape of the wavefront of the scattered photon density wave carries information on the size and shape of the scattering object (Gratton et al.). Finally, the issue of boundary conditions in medically relevant cases is discussed (Chance et al.). It is suggested that by using the substitution method, the escape of light from physical boundaries can be minimized. The consequences of this approach, when applied to the conventional finger oximeter, is discussed.

A physical model of how optical heterogeneities affect time-dependent measurements of photon migration in tissue-like scattering media is presented. Using this model, changes in frequency-domain measurements of phase-shift, 0, and amplitude modulation, M, are predicted in the presence of transparent and perfectly absorbing objects. Two-dimensional Monte Carlo simulations of photon migration and single-pixel measurements of 0 and M confirm the physical model. Recent experimental "images" from multi-pixel measurements of 0 and M are also consistent with the physical model. These results suggest that two dimensional frequency-domain measurements provide direct information for detection and three dimensional localization of optical heterogeneities without the use of computer intensive reconstruction algorithms.

Between 600 and 1300 nm the optical absorption coefficient pa of mammalian soft tissues is quite low', ranging from about 0.1 - 1.0 cml. The scattering coefficient ps, on the other hand, is high - typically 100 - 1000 cm'. This scattering is forward peaked, so that g, the mean cosine of the scattering angle, is about 0.9. Since the net result of several small angle scattering events is similar to that of a single isotropic scatter, it is useful to define a reduced or transport scattering coefficient p. = (1-g) Ns. Even this reduced coefficient, which is usually 5 - 50 cm', is typically 10 -100 times the absorption coefficient.

Light propagation in very turbid media can be described as a diffusion process.' Light in the near-infrared is multiply scattered in tissues. The scattering coefficient, is on the order of 103 to 104 per mm and is essentially forward scattering.2,3,4,5,6,7 Taking into account the average of the cosine of the scattering angle, g, it is possible to define a reduced scattering coefficient, !Zs= (1-g)µs, which is still on the order of 1 per mm. Typical values of the absorption coefficient, ga, in animal tissues are on the order of 0.01 per mm.7 Under the above conditions, the Boltzman transport equation for photons in tissues can be solved in the diffusion approximation.' Several researchers have experimentally demonstrated the validity of the diffusion approximation in typical tissues.4,9,10,11,12 Once we have achieved a fairly good understanding of the physical nature of light propagation in tissues, the question remains as to how best to determine the parameters that appear in the diffusion approximation solution, i.e., the scattering and the absorption coefficient. In general these values are a function of the location in the tissue and they can vary with time. Ideally, we want to reconstruct a 3-D map of the scattering and absorption coefficients with the highest possible spatial resolution from time-resolved measurements of light intensity performed at the surface of the object being investigated." At this point several questions arise as to the mathematical possibility of the reconstruction of the map of scattering and absorption, the ultimate resolution achievable, the size of differences in scattering and absorption coefficients that can be measured, the best measurement method, and the best reconstruction algorithm to name a few.

The problem of accurate measurement of scattering and absorption of small absorbers has recently been solved by what is termed a "substitution technique" in which a suitable medium surrounds the biological object and affords an essentially infinite and homogeneous photon migration space in which scattering and absorption factors may be precisely determined. This approach finds applications in the quantitation of hemoglobin saturation and important concentration changes. This opens up the possibility of quantifying such data for small absorbers and scatterers. One case of particular interest is the human finger or earlobe in which the pulse oximetry industry is based.

In 1880 an important book with the title The Spectroscope in Medicine was published by MacMunn in London. In 1914, MacMunn published another book with the title, Spectrum Analysis Applied to Biology and Medicine. The color of cells and tissues, and their absorption and fluorescence properties begins with the studies of hemoproteins by MacMunn, continues with the classic work of Keilin on cytochromes, and continues in this century with the modern advances in the development of redox fluorometrv.

This chapter describes redox imaging of cells, tissues, and organs based on these intrinsic fluorescent probes of cellular metabolism." Cellular metabolism may be noninvasively monitored through the "optical method" based on the fluorescence intensity of intrinsic probe molecules. The intrinsic fluorescent probes, which report on cellular metabolism, are the reduced pyridine nucleotides, NAD(P)H, and the oxidized flavoproteins.

The oxygenation of human forearm muscle tissue can be investigated using near infrared spectroscopy (NIRS). Oxy and deoxy hemoglobin changes can be quantified combining attenuation measurements with pathlength data obtained by time resolved spectroscopy. This study reports some applications of NIRS to non- invasive measurements of skeletal muscle oxygenation in untrained volunteers and patients.

The ability to assess the adequacy of oxygen delivery by the circulation to the mitochondria in tissue is essential in surgery, intensive care and anaesthesia. Furthermore it has become increasingly clear that it is no longer sufficient to measure global oxygen delivery and consumption as indicators of metabolic requirements but that knowledge about local oxygenation is needed. This is due to the redistribution of oxygen delivery and consumption between and even within the different organs. The success of the recently introduced tonometry, using splanchnic oxygenation instead of global oxygen parameters, as a guide to therapy in critical care [Fiddian-Green '87, Guterriez '92] is witness to the importance of the need to obtain information about the course of oxygen to tissue at microcirculatory and even mitochondrial level.

A mathematical analysis is presented for the application of dual wavelength excitation and dual wavelength detection for in vivo fluorescence imaging. A dimensionless ratio-image is derived that is independent of optical properties and detection geometry.

In this paper we present a good mathematical model description of excitation and fuorescence of Haematoporphirine (HpD), under saturation conditions. This complex molecule can be described as a three level atom and under special conditions by a two level atom. The interaction between atom and light is well known and can be represented by rate equations which enables us to do calculations on various interactions. This model gives a better understanding of fluorescence of a photosensitizer and shows it is possible to determine several molecular constants of HpD.