1. Introduction

Light has unique properties to enable sensing of functions, physiology, and lesions of human body. The past decade has witnessed a rapid growth of research activities to exploit some of those attributes of light for the development of optical biomedical imaging and biopsy techniques [1]. Although an impressive array of body-imaging techniques, such as x-ray imaging, x-ray computed tomography, magnetic resonance imaging, ultrasound, and radioisotope imaging are currently available to yield useful information [2], there are important limitations of safety, resolution, cost, and lack or limited specificity to key chemicals necessary for functional body monitoring. More importantly, conventional techniques sometimes fail to provide necessary information. For example, x-ray mammography, the current ‘gold standard’ for monitoring breast cancer, is not suitable for imaging dense, young breasts, and may not distinguish between malignant and benign tumors. Optical imaging modalities are being pursued to alleviate most of these limitations and to achieve diagnostic ability through the exploitation of the spectroscopic characteristics of tissues. Consequently, a number of continuous-wave, time-resolved, and frequency-domain optical imaging techniques have evolved over the years [3–16,1]. It should also be noted that any of the above-mentioned imaging techniques provide one type of image only, while by changing the wavelength of light one can obtain multiple images that highlight the various molecular components. In addition, light interacts with biological tissues in a number of different ways providing different means to probe the body.

In this article, we first present a brief overview of the various light-tissue interactions, and the properties of light that enable sensing the interior of body organs, such as breast and prostate. We then illustrate the biomedical sensing potential of light through the time-sliced optical imaging and spectroscopic imaging of normal and cancerous human breast tissues.

2. Background

Shadowgram optical imaging was first attempted in 1929 by Cutler [3] who used white light to illuminate the breast and look for pathology in the transmitted light. Inadequate spatial resolution attainable with the available light sources of that time, and the rapid progress in x-ray mammography eclipsed the development of light-based method, that was also known as transillumination, or, diaphanography [2]. With the advent of broadly tunable near-infrared (NIR) solid-state lasers, such as Ti:sapphire and Cr:forsterite, and the development of charge-coupled device (CCD) and NIR area cameras, there has been a surge in research activities for the development optical body-scanning modalities since the mid-1980’s.

The purpose of an optical body-scanning modality is to identify, locate, and diagnose a ‘lesion’, such as a tumor, inside human body. The procedure involves illuminating the relevant part of the body, such as the breast or the prostate, with bright light of appropriate wavelength and search for indication of pathology in the emergent light. The physical basis for the development of an optical modality is the difference between the interaction of light with the lesion and the surrounding tissues. Major problems that impede the development include attenuation of light due to scattering and absorption by the tissues, and image blurring due to scattering. The nature and extent of the changes in the properties of light that result from its interaction with biological tissues may be different for normal and diseased tissues, as well as for different types of tissues, and thus provide bases for the development of optical imaging and optical biopsy methods.

The high spectral brightness of laser light makes it the most common choice of researchers for optical biomedical imaging. The enabling attributes of laser light include wavelength, coherence, polarization, pulse duration in the case of mode-locked and Q-switched lasers, intensity and directionality. Each of these properties may change as a result of light-tissue interaction that depends on the wavelength of light. A proper choice of wavelength(s) is a crucial consideration for developing an optical imaging modality. Light in the ‘therapeutic window’ of 700-1300 nm is less absorbed by the tissue constituents compared to near-ultraviolet, visible, or infrared light and is commonly used for tissue imaging.

The common linear interactions of a beam of light with biological tissues include scattering, absorption, specular and diffuse reflection, and diffuse transmission. Scattering changes the directionality, intensity, coherence, polarization, and pulse-width, of the incident light. Consequently, light transmitted through a scattering medium breaks up into ballistic, snake, and diffuse components [5,9,1]. The early-arriving, forward-propagating ballistic and snake photons carry image information while the multiply-scattered diffuse light blurs the image, and in extreme cases buries it in the background noise. The difference in transit time and changes in the above-mentioned characteristics are exploited for selecting the image-bearing light to form direct transillumination images, or, ‘shadowgrams’. Since the image-bearing photons change the least, the idea is to devise a gate that will let a significant fraction of the photons with a specific initial property through but block others. Realization of this idea has led to the development of time, coherence, polarization, nonlinear optical, and space gates [4–16,1]. However, for a highly scattering and thick medium the image-bearing light becomes too weak to form a shadow image. One then resorts to inverse methods that make use of the scattered light intensity patterns measured around the object, a mathematical model for description of light propagation in scattering media, and a sophisticated computer algorithm to reconstruct an image of the object [17].

Fluorescence that results from the absorption of light by a molecular constituent, such as tryptophan, tyrosine, collagen, elastin and flavin, of a tissue is a characteristic of that constituent. Changes in the fluorescence spectra, lifetime, intensity, and intensity ratios for different excitation wavelengths may provide markers for lesions and facilitate their imaging [18]. Incident light may be frequency shifted due to Raman scattering by the vibrational modes of the molecular constituents of a tissue and differences in the Raman spectra of normal and diseased tissue may provide diagnostic ability [19]. Nonlinear optical interactions of two photon absorption-induced fluorescence and second harmonic generation have also been used to generate histopathological maps of biological tissues [15,16].

The organs of human body with potential to be investigated by optical spectroscopic techniques include breast, prostate, brain, bladder, bone, cervix, colon, eye, digestive and gynecological tracts, skin, and teeth [20]. The requirements of resolution, penetration depth, diagnostic ability, and specificity to key chemicals of an optical imaging modality depend on the organ and the lesion being investigated.

The remainder of this article focuses on our NIR time-sliced and spectroscopic imaging of human breast tissues in vitro. Optical imaging of breast is complicated by the extremely low ratio of image-bearing photons to multiply scattered image-blurring photons, and paucity of high contrast targets that are indicative of the presence of cancer and suitable for imaging. Our thrust is to develop a better understanding of light transport through different types of tissues, obtain NIR transillumination images, and explore the feasibility of distinguishing between normal and cancerous tissues from those images.

3. Materials and Methods

We have investigated excised female human breast tissue specimens with normal fibrous and adipose tissues, as well as normal and cancerous tissues. Types of cancer included infiltrating ductal carcinoma, and infiltrating lobular carcinoma. The tissue specimens were placed between two glass plates and compressed to provide a uniform thickness. The specimens were provided to us by National Disease Research Interchange and Memorial Sloan Kettering Cancer Center under an IRB approval from the City College of New York. Here we present representative results of a time-sliced imaging measurement on a specimen with infiltrating ductal carcinoma, and a spectroscopic imaging measurement on a specimen with adipose and fibrous tissues to demonstrate the potential of optical imaging methods.

The time-sliced imaging approach is an extension of the idea of time-resolved early-light imaging except that picosecond-scale slices of the transmitted light pulse are used to record a sequence of two-dimensional (2D) images for different positions of the time gate. The experimental arrangement made use of 800-nm, 130-fs duration, 1 kHz repetition-rate pulses from a Ti:sapphire laser and amplifier system for sample illumination, and an ultrafast gated intensified camera system (UGICS) for recording 2D images. The UGICS comprised a compact time-gated image intensifier unit fiber-optically coupled to a CCD camera. It provided an approximately 80 ps-duration gate whose position with respect to an external trigger pulse could be varied in a minimum step size of 25-ps over a 20 ns range. The average beam power used in the experiment was approximately 200 mW. The beam was expanded and a 3-cm diameter central part of it was selected out using an aperture to illuminate the sample. The time-sliced images were recorded by the CCD camera for different gate positions and displayed on a personal computer.

The experimental arrangement for NIR spectroscopic imaging, detailed elsewhere [21], made use of 1210-1300 nm continuous-wave mode-locked output of a Cr:forsterite laser to illuminate the sample. A Fourier space gate [13] and a polarization gate [11] were used to select the image-bearing photons. A 50 mm focal-length camera lens placed on the optical axis at a distance of 50 mm from the aperture in the Fourier gate collected and collimated the low-spatial-frequency light filtered by the aperture and directed it to the 128x128 pixels sensing element of an InGaAs NIR area camera. Appropriate neutral density filters were used to maintain the average optical power of the incident beam at approximately 35 mW for all the wavelengths used in the imaging experiment. The laser beam was linearly polarized along the horizontal direction.

4. Results

Time-sliced images of a 5-mm thick breast tissue sample comprising a piece of normal tissue and a cancerous piece with infiltrating ductal carcinoma for two different gate positions are displayed in the left frames of Fig. 1. The zero position was taken to be the time of arrival of the light pulse through a 5-mm thick glass cell filled with water. Corresponding spatial intensity profiles integrated over the same horizontal area for all the images are shown in the frames to the right. The salient feature of the images and intensity profiles is that at early times (25 ps) more light is transmitted through the cancerous region than the normal region, while at a much later time (275 ps) relatively more light arrives through the normal region. Lower scattering or/and higher absorption of light by cancerous tissues may account for the observed temporal behavior. Since there is no significant absorption of 800-nm light by the breast tissues, we attribute the observed time-dependent difference in light transmission to the higher scattering of light by the normal tissue than by the cancerous tissue. Although we did not carry out independent measurements of the light transport characteristics of the tissues, our results are consistent with the literature values [22] of reduced scattering coefficient of 1.3 mm^-1 for normal glandular tissues, and 1 mm-1 for cancerous breast tissue at 800 nm. Images recorded with different time slices of the transmitted light could highlight normal and cancerous regions of an excised breast tissue specimen.

 

Fig. 1. Time-sliced 2D images (left) of a 25mmx9mmx5mm breast tissue sample with normal and cancerous components for gate positions of 25 ps and 275 ps. Spatial intensity profiles integrated along the same horizontal area are shown in the respective frames to the right. A schematic diagram of the sample is shown in the top left frame. The zero position was taken to be the time of arrival of the light pulse through a 5-mm thick glass cell filled with water.

Download Full Size | PDF

The thrust of the spectroscopic imaging experiment was to test if a spectroscopic difference between different types of tissues in a specimen provides any distinguishable signature in the transillumination image. We obtained images of a normal breast tissue sample comprising adipose tissue in the middle and fibrous tissues in the two ends with light of wavelengths in the 1225-1300 nm range. Some of the wavelengths were near-resonant with the adipose optical absorption band around 1203 nm, [23] while others were far removed from that resonance. Measurements were made both at the room temperature of 20 °C and the normal body temperature of 37 °C. The salient features of the images, displayed in Fig. 2, are: (a) adipose tissue appears much darker (less light transmission) than the fibrous tissues for the near-resonant 1225-nm image as compared to the off-resonance 1285-nm image; and (b) transmission of light through the specimen is higher at 37 °C than at 20 °C. Adipose tissue region appeared as a deeper trough in the spatial intensity profile of the 1225-nm image compared to that for the 1285-nm image. For a more quantitative description of the observed behavior, we monitored the image contrast, C(λ,T) = (I_F - I_A)/(I_F - I_A), where I_A(λ,T) is the optimal intensity value at wavelength λ and temperature T on the spatial profile of the image at the adipose tissue location, and I_F(λ,T) is the corresponding intensity at the immediate fibrous tissue region. Values of contrast at 1225 nm are 0.59 and 0.46 for 20 °C and 37 °C, respectively, while the corresponding values for 1285 nm are 0.25 and 0.10. As the laser output was tuned away from the absorption peak, the contrast between the adipose and fibrous regions in the images decreased. These results demonstrate that an appreciable spectroscopic difference may significantly enhance the contrast between different types of breast tissues in the transillumination image. The ratio of and/or difference between images recorded with resonant (or near-resonant) light and non-resonant light may enhance the image contrast even further.

 

Fig. 2. Spectroscopic 2D images (left frames) of a 28mmx12mmx10mm normal human breast tissue comprising fibrous (F) and adipose (A) components recorded with (a) 1225 nm and (b) 1285 nm light. Spatial intensity profiles of the images integrated over the same vertical area are shown in the respective frames to the right.

Download Full Size | PDF

5. Discussion

The results of time-sliced imaging experiment demonstrate that optical imaging using optimal time slices of transmitted light is a promising method for imaging biomedical media. While the early-arriving light highlighted the cancerous tissue, the later-arriving light accentuated the normal component. Different types of tissues scatter light differently, and for specific sample configurations the later-arriving light may be as revealing as the early light. Another novel application of the sequence of time-sliced 2D images that is obtained by this approach is in fast three-dimensional tomographic inverse image reconstruction [24]. The time-sliced imaging approach enables fast acquisition of data over a wide frequency range that is necessary for 3D reconstruction, as compared to frequency-domain schemes that commonly use several discrete frequencies.

Results of the NIR spectroscopic imaging experiments demonstrate the diagnostic potential of optical imaging. Combined time-sliced and spectroscopic imaging method has the potential to provide much more information than even the x-ray techniques.