Imaging structures within a turbid medium using Angular Domain Imaging (ADI) employs an angular filter array to separate weakly scattered photons from those that are highly scattered. At high scattering coefficients, ADI contrast declines due to the large fraction of non-uniform background scattered light still within the acceptance angle. This paper demonstrates various methods to enhance the image contrast in ADI. Experiments where a wedge prism was used to deviate the laser source so that scattered photons could be imaged and subtracted from the image obtained by standard ADI provided the greatest improvement in image contrast.
© 2008 Optical Society of America
Optical tissue characteristics are wavelength dependent, which allows physiological and functional information to be gained from wavelength-dependent measurements . For example, blood content and oxygenation, which are respective surrogates for angiogenesis and hypermetabolism, can be deduced by spectral measurements . The basis for this technique is the strong optical absorption of hemoglobin in blood and its associated oxygen-dependent spectral changes . However, optical methods suffer from optical scatter, a serious limitation that prevents features below the surface from being visualized at high resolutions. For instance, confocal microscopy or Optical Coherence Tomography (OCT) fails to provide high resolution images deeper than a few hundred micrometers –. Fluorescence and luminescence techniques have become popular for whole animal scanning of genetic reporters and molecular contrast agents; however, they too are hampered by the degrading effects of optical scatter and cannot provide high resolution images of internal structures.
To overcome the limitations to tissue imaging imposed by optical scatter, several approaches have been investigated, including time domain, spatial filtering and angular domain. Time-domain optical imaging is a technique that was first shown to be applicable to imaging tissues more than a decade ago [6,7]. The time-domain method separates the early arriving (and image forming) ballistic and quasi-ballistic photons from the late arriving scattered photons (non-image forming). This technique has not been adopted widely likely on account of the lengthy acquisition time resultant from hardware inadequacies. Spatial filtering techniques have been studied as a means to reject scattered light and preserve the image forming photons. The technique uses a lens to create the Fourier transform of the spatial distribution of light on the exit surface of a trans-illuminated object and then removes high-order frequencies using a small aperture placed at the focal point of the lens. Several groups have successfully used this method to image resolution targets, but the approach suffers from insufficient rejection of scattered light, loss of resolution due to the small aperture, and has yet to be shown applicable to imaging thick tissues on its own. Our group has been studying angular domain optical imaging (ADI), which uses a micro-machined silicon angular filter (collimator) to filter unwanted scattered photons and pass the image forming quasi-ballistic photons. This technique has been successfully applied to imaging resolution targets as small as 150 µm suspended in turbid solutions of milk and Intralipid™ [8,9,10]. The objective of the present work was to characterize and compare three methods for improving image contrast in ADI. The first method involved the introduction of a wedge prism into the light path of the ADI scanner to quantify the degree of highly scattered light. The second method introduced a pair of polarizers into the ADI light path as a means to quantify the scattered light. The third method combined the wedge and polarizer techniques. In all cases, the images representative of the scattered light were used to correct the quasi-ballistic photon-weighted images.
In preparation for describing image contrast enhancement in ADI, this paper reviews the light-tissue interactions relevant to ADI, the practical implementation of ADI for imaging turbid media, and mechanisms responsible for the degradation of ADI image contrast. The paper then goes on to describe the three approaches to improving image contrast in ADI and their relative performance compared to ADI alone.
2. Light-tissue interaction
The fundamental mechanism of light transport in absorbing and scattering media was adapted from nuclear physics where Chandrasekar calculated the fluence of neutrons through different materials . Hence, in most medical laser applications light is regarded as a particle and not as a wave. The integro-differential equation describing the stationary photon transport for monochromatic light is given by the diffusion theory as follows [12,13]:
where I(x,Ω) is the angular energy flux density at position x and angle Ω in Wm-2sr-1 and S(Ω’ →Ω) denotes the angular distribution of light scattered from angle Ω ’ to Ω. Q(x,Ω) represents the distributed source density. Here, µa in cm-1 is the absorption coefficient, and µs in cm-1 is the scattering coefficient.
S(Ω’ →Ω) also represents the scattering phase function and is normalized to unity when integrated over a 4π solid angle. When the phase function is independent of angle, the scattering is said to be isotropic. For anisotropic scattering, the phase function depends only on θ, the angle between the incident and scattered radiation. The net anisotropy of scattering for an arbitrary phase function is named the g-factor which is summarized as follows.
The phase function S(Ω’ →Ω) that describes the experimentally observed distribution of scattering angles of photons is the Heneye-Greenstein (HG) function :
It has been found by Jacques et al.  that the HG function S(θ) also accurately describes the scattering of light in biological tissue. Many common biological tissues have a high degree of forward scattering, with g values ranging from 0.79 to 0.95 in the red and near infrared spectrum range . The scattering coefficient decreases to the reduced scattering coefficient, µ’s, for the so-called “quasi-ballistic or snake photons” (i.e. photons which follow paths close to that of ballistic photons) as
For instance, measurements of skin and underlying tissue at 633 nm  have determined µs = 70.7 cm-1 and g = 0.8. By Eq. (4), µs’ will be about 5 times lower and eqgual to µs’ = 11.4 cm-1. It can be assumed that quasi-ballistic photons are closely confined within a small angle of their initial trajectory (i.e. they have a forward directed distribution) due to the anisotropy factor in tissue. A Monte Carlo model of steady-state light transport in multi-layered tissues (MCML) by Jacques et al.  demonstrated that quasi-ballistic photons exit an ideal isotropic scattering medium with very small angular deviations from their initial trajectories. Extending this work, we used Monte Carlo simulations to show that scattered photons exiting with small angular deviations also follow the shortest path, and hence are quasi-ballistic in nature . Therefore, a detection system that uses angular filtration with a small acceptance angle and an aligned, collimated source of light that passes through a turbid media can be a practical method to reject most of the uniformly distributed scattered light while accepting quasi-ballistic photons with small angular deviations. This is the basis of ADI.
3. Angular domain imaging
3.1 Principle of operation
The ADI method operates by selecting ballistic and quasi-ballistic light closely confined within a small angle of the incoming photon trajectory while filtering out highly scattered light. This follows from the observation that highly scattered light tends to have a nearly uniform angular distribution while image forming quasi-ballistic light is highly directional [9,10]. As shown in Fig. 1 the Angular Filter Array (AFA) employs high-aspect ratio micro-tunnels to create extended apertures through which photons can pass if they arrive within the allowable acceptance angle with respect to the longitudinal axis of each micro-tunnel. Photons that arrive with incident angles beyond the acceptance angle will strike the micro-tunnel sides and be attenuated. The performance of ADI is largely dependent on the design of the array of angular filter micro-tunnels. The AFA must be designed with a high aspect ratio, length (l) over diameter (d), to provide sufficiently strict angular filtering of scattered photons.
The ADI technique works successfully in transmission geometry, where the laser source is aligned to the angular filter to trans-illuminate the turbid medium from front to back, ADI can also be employed using an illumination source on the same side as the angular filter to capture quasi-ballistic photons generated deep from within the scattering medium. In the reflective geometry, collimated light is injected into the turbid medium, thus generating a source of illumination from inside the medium and behind the imaging target object. This newly formed light source in the scattering medium emits non-scattered and scattered light relative to the AFA micro-tunnels. When these back reflected photons pass an imaging target, which can be either absorbing or fluorescent material, and reach the AFA, the relative non-scattered photons are accepted through the micro-tunnels and reach the camera, whereas scattered photons are rejected by the filter. Reflection geometry ADI has been investigated previously .
Recently, we investigated a new design for the angular filter array (see Fig. 2), which consisted of a parallel array of square-shaped micro-tunnels, 60 microns wide and tall along a 1 cm long plate to obtain aspect ratios of approximately 167:1. The square shape with small spacing had higher performance compared to our previous semi-circular micro-tunnels. Compared to the semi-circular geometry, the square geometry employed smaller spacing between micro-tunnels with a larger cross-sectional area for a given nominal acceptance angle; therefore, it was more efficient at accepting the informative quasi-ballistic photons exiting the turbid sample. The aspect ratio and micro-tunnel size for the square geometry reported above was observed to provide optimal image contrast compared to other filter geometries (data will be presented in ). Because the micro-tunnels were square in geometry, there existed an angular acceptance angle variation from 0.34° (wall to wall) to 0.48° (corner to corner). This design was selective enough to collect quasi-ballistic photons that provided at least 200 µm spatial resolution through turbid media. The main part of the AFA is an array of long channels etched into a silicon substrate as shown in Fig. 2, forming the bottom section. For the complete AFA a flat silicon wafer added on top of these etched channels to create enclosed tunnels. The new AFA fabrication decreased the power loss due to the larger opening size compared to the previously tested semi-circular shapes. This feature could be used to reduce the time required to capture one image or reduce the light power for delicate samples.
The one dimensional linear array of micro-tunnels has a limited field of view as shown in Fig. 3. This can create two limiting issues. First, light from all illuminated regions of the sample that are not in the area being imaged will be scattered into the region of the AFA and a small portion will be within the acceptance angle of the channels, thus reaching the sensor. This result in a background signals at each pixel location and reduces image contrast. This effect can be suppressed somewhat by reducing the amount of unnecessary illumination of the sample using a collimated line source coincident with the AFA . Second, the limited vertical field of view of the AFA necessitates a scanning system for capture of 2D images. We have employed a computer-controlled z-axis stage (with 0.05 µm repeatability) to incrementally raise the scattering sample between scans. One horizontal line image of the sample is taken through the angular filter at each step and a final 2D image is assembled from the stacked line images. Hence, an entire region of the sample can be passed through the angular filter’s field of view and imaged.
Angular domain imaging possesses multiple benefits as an optical imaging method. These include: i) wavelength independence, which allows for broadband or multi-spectral light sources ; ii) non-coherent source compatibility, which implies that ADI can be performed with inexpensive light sources, and iii) multimodal capability, that is ADI is useful not only for detection of absorbing targets in a turbid medium, but also for fluorescence targets, which has promise for imaging genetic reporters and molecular contrast agents in biological tissue.
3.2 Mechanisms of loss of image contrast in ADI
As noted one important limitation of ADI relates to a loss of image contrast as the scattering level of the sample increases. The fraction of the highly scattered photons increases and creates background noise in the image. Inherently, highly scattered light has a nearly uniform angular distribution; hence, a small fraction of the scattered light happens to exit the sample within the micro-tunnel acceptance angle, which biases the ballistic and quasi-ballistic photons accepted by the filter. This results in background scattered light that reduces the effective image contrast. Fig. 4 shows ADI scans using an 808 nm laser diode at three levels of scattering. As the scattering level of the medium is increased from modest (Fig. 4(b)) to high scattering levels (Fig. 4(c)), the fraction of light that is ballistic or quasi-ballistic decreases significantly, while the background scattered light increases. This results in a decrease in contrast between the target and the background. Image contrast decreases as the scattering level increases to the point where the object patterns are obscured by the background scattered light. That is, ADI is unable to distinguish between filtered non-scattered light from background scattered light.
The image exposure time by the detector is dependent on the scattering level, laser power, and camera exposure time. For instance, in order to form an ADI scan in water (Fig. 4(a)) each scan line was acquired with a laser diode power of 2.4 mW and an exposure time equal to 0.012 ms (the full 2D image was formed from 330 line scans). At 0.6% Intralipid™ (Fig. 4(b)), the laser power was increased to 350 mW and the camera exposure time for each scan line was lengthened to 10 ms. However, in the practical implementation, several other processes (data communication, computer software and stage translation) increase the time for one line scan to approximately one second; therefore, the whole scan shown in Fig. 4(b) took approximately 5 min. At 0.75% Intralipid™ as shown in Fig. 4(c), each line scan was captured at a laser power equal to 350 mW with a 25 ms camera exposure time. The full 2D image was acquired in approximately 5 min.
To quantify this effect, we employ a geometric analysis. As a first approximation, scattered photons will exit the medium with a uniform distribution of angles. Of these randomly scattered photons, a small proportion will be in the acceptance range of the angular filter and contribute to the gray background in the image. The amount of scattered light leaked through the angular filter is related to the solid angle subtended by the micro-tunnel. The solid angle of a rectangular field of view (symmetrical with respect to the axis of observation) of width a, height b, and length of d is:
In our case, 60µm×60µm square-shaped micro-tunnels with a length of 1cm subtend a solid angle of 3.6×10-5 sr. The solid angle, Ω, can be imagined as the fractional surface area at the opening of the micro-tunnel. Hence, the fraction of light collected by one micro-tunnel of the angular filter is Ω/4π. Finally, the number of total scattered photons for every scattered photon within the solid angle of the micro-tunnel (SRmax) is:
Therefore, two populations of photons contribute to the ADI image: 1) Scattered photons, which are distributed as a sphere of light in all directions. From this scattered light, a small portion passes through the angular filter array and forms the background scattered light component of the image that degrades image contrast. 2) Forward-directed ballistic and quasi-ballistic photons within the acceptance angle of the filter that contribute to the informative component of the image that provides contrast. It is important to note that this second component decreases as SRmax increases.
Another challenge of ADI implemented with an angular filter array is related to the partially reflective nature of the silicon micro-tunnel sidewalls. Ideally, scattered light that collides with the micro-tunnel sidewalls would be fully absorbed and prevented from passing through the micro-tunnel and reaching the detector. However, silicon is a semiconductor with a very wavelength dependent complex index of refraction of n = 3.72+i0.013 reflecting ~34% reflective to normally incident light. At very shallow angles Fresnel reflection substantially increases this. Using the Fresnel reflection formulae for reflection of angled light, the reflectance for parallel and perpendicular polarized light can be calculated for shallow angles of incidence. For a 0.3° (i.e. 89.7° relative to the normal) angle of incidence that is likely to be present in an ADI experiment, the reflectance is expected to be 92.5%–99.4% for silicon, depending on polarization and the wavelength. Therefore, instead of being fully attenuated, a fraction of the scattered light at very shallow angles collides with the micro-tunnel sidewalls is reflected within the micro-tunnel several times and reaches the detector, where it combines with non-scattered light to form the image. In practice as light is absorbed at each reflection only the scattered light with 2 or 3 times the acceptance angle makes it through to the detector. This scattered light acts as an additional source of bias in the image that leads to a loss of image contrast.
We devised three methods to account for the background scattered light in ADI and use the information to improve image contrast. The first approach attempts to better understand the effect of polarization discrimination during ADI. The approach takes advantage of the fact that the orientation of linearly polarized light is randomized by scattering during propagation through turbid media. The experiments were conducted in transmission mode where photons propagated through a cuvette containing Intralipid™ and were detected on the backside of the sample. The second approach employed a wedge prism to deviate the laser source where it entered the medium by an angle slightly larger than the AFA acceptance angle. This created a second image consisting of highly scattered photons with the filtration characteristics of the angular filter, and a pixel by pixel correspondence to the fully scattered illumination emitted from the medium. The third method combined the polarizers with the wedge prism.
4.1 Experimental setup (ADI)
Due to the benefits of the longer wavelengths in NIR range, the experiment setup employed an 808 nm laser diode (Thorlabs, L808P1WJ multi mode laser diode). The laser beam was shaped and collimated into a line of light by an aspheric and cylindrical lens based collimation system (see Fig. 5). The beam divergence for the laser diode was high along the vertical axis while significantly lower along the horizontal axis; therefore, a two-axis collimation system was implemented using one aspheric lens followed by two cylindrical lenses. This lens collimation system produced a line of light with approximately 3 cm wide by 1 mm tall. This beam then passed through an iris diaphragm approximately 8 mm wide to restrict the width of the beam before it illuminated the sample. This was necessary to ensure that the line of light had a uniform light intensity profile before hitting the sample. Previous work at high scattering levels  showed that due to the very small acceptance angle, the exiting light included many internal shallow angle reflections due to the micro-tunnel walls of the AFA. This light exited at higher angles from the micro-tunnels and degraded both the signal to noise ratio and the spatial resolution. At these shallow angles, ~0.48°, absorbing films or anti-reflection coatings are known to be not effective . Therefore, we chose to use a Keplerian lens system to accept the parallel light exiting the micro-tunnels and reject the majority of the diffracted photons caused by internal reflections. In addition, with the lens system in place, an increase in the scanned area compared to previous work was possible.
Turbid samples with an embedded resolution target rested on a computer-controlled z-axis stage, where they were illuminated by the collimated light source and imaged through the angular filter array by a CMOS camera. Resolution targets could be inserted at any position along the test container, ranging from the front position facing the laser to the back position facing the angular filter and camera as shown in Fig. 5. The AFA was placed in front of a high-resolution (1280 x 1024 pixel) CMOS detector with a square pixel size (5.2 µm × 5.2 µm) that was smaller than the dimensions of each AFA micro-tunnel and was mounted on a 6 degree-of-freedom (DOF) jig that provided alignment to the laser source. As noted earlier, due to the design of the AFA only one horizontal projection through the sample could be imaged with AFA at a time. Thus, the sample was incrementally raised by a computer controlled z-axis stage so that an entire region of the sample passed through the field of view of the AFA. The default vertical step size used was 26 µm, which was equivalent to 5 rows of pixels in the CMOS detector. As the test sample was raised, one horizontal projection of the sample was captured at each step, and a final image was assembled by combining the horizontal projections.
For all experiments, we used an aqueous suspension of Intralipid™. Intralipid™, a phospholipid emulsion used usually as an intravenous nutrient, is a practical phantom medium for light dosimetry studies since, like tissue, it is turbid at visible and near infrared wavelengths. Additionally, Intralipid™ lacks strong absorption bands in the visible and near infrared region of the spectrum . As a scattering medium, the optical properties of Intralipid™ have been well characterized . To test the spatial resolution of our scans, resolution targets were fabricated as L-shaped patterned aluminum thin films on a glass slide. The resolution target was placed either within the scattering medium or at the front (i.e. closest to the laser). In our experiments the resolution target had L-shaped lines and spaces varying from 150 µm, 200 µm, 300 µm, and 400 µm. The L-shape geometry provided a means to test the spatial resolution of our imaging setup in both x and y -axis.
4.3 Estimation of background scattered light using polarization discrimination (PADI & PADI-CS)
In principle, ballistic photons maintain their polarization memory while scattered photons start to become randomly polarized within the sample. Hence, when the axes of two polarizers are parallel, ballistic light will be passed and some scattered light will be attenuated. The experimental arrangement for the polarized angular domain imaging (PADI) measurements is shown in Fig. 5 and described in Table 1. The collimated line beam was passed through a linear polarizer to grant linear polarity to the incoming light to the sample. The second polarizer was set with their polarization axis parallel to the first. The camera was focused on the surface of AFA using the Keplerian lens system. The measurements were performed on an aqueous suspension of Intralipid™.
Alternatively, when the axis of one polarizer is rotated by 90° relative to another (crossed), the ballistic light, which maintains its polarization, will be rejected and the scattered light will be transmitted. Therefore, ADI can be used with a pair of crossed polarizers to estimate the background scattered light. The second image collected with crossed polarizers contains multiply scattered light with pixel by pixel correspondence to the ADI scan with parallel polarity configuration (i.e. PADI). The technique of PADI with crossed subtraction (PADI-CS; Table 1) utilizes digital image subtraction of the PADI scan obtained with the crossed polarizers from the PADI scan collected with the parallel polarizers. This technique provides a way to enhance the image contrast by subtracting the background scattered light and potentially correct for its degrading effect on image contrast in ADI.
4.4 Estimation of background scattered light using deviated light (ADI-WS)
In principle, most scattered photons are assumed to be eliminated while exiting the turbid medium with an angle greater than the acceptance angle of the AFA (e.g. 0.48°). However, as discussed earlier in this paper, a small proportion of scattered photons leak through the AFA giving rise to background scattered light. Experiments show that a slight deviation of the collimated laser beam prior to striking the face of the sample (~2× of angular filter’s acceptance angle) results in a loss of detected ballistic and quasi-ballistic photons, but a preservation of the detected background scattered light. Therefore, the ADI image resultant from the deviated light source provides a pixel by pixel estimate of the background scattered light present in a corresponding ADI image where the light sources is aligned with the AFA. Digital image processing is then used to subtract the background scattered light image from the original ADI image (i.e. wedge subtraction), thus enhancing contrast on a pixel by pixel basis. In our system, we used a wedge prism to deviate the light source slightly beyond the angular filter acceptance angle (~1 degree in our experiments). We define this method as angular domain imaging with wedge subtraction (ADI-WS; Table 1). It is important to note that scattered light is affected by many factors such as the individual micro-tunnel geometry, illumination uniformity, and the test structure pattern. Capture of the background scattered light image by the wedge technique provides a more accurate pixel by pixel estimates of the scattered light distribution than digital image processing techniques applied globally to the original ADI image.
4.5 Combined approach (PADI-WS)
In this setup, parallel polarization measurements with angular domain imaging (PADI) filters the light from scattering medium based on polarization and small acceptance angle provided by AFA. Estimation of background scattered light is also feasible by employing the wedge prism. We define the combined technique as polarized angular domain imaging with wedge subtraction (PADI-WS; Table 1).
In this technique, light passes through two polarizers with axes parallel to obtain an image which includes quasi-ballistic light and a small fraction of scattered light. Then the wedge prism is inserted, which causes the quasi-ballistic light to be deviated out of the field of view of the AFA, hence only scattered light will be detected by camera through the AFA. Using the same principle as ADI-WS, the second image can be subtracted from the original to correct for the background scattered light. Table 1 summarizes all the methods employed to enhance the image contrast.
5. Results and Discussion
5.1 Detection limit imposed by sample thickness and Intralipid™ concentration
When the concentration of Intralipid™ was increased, the detection limit of ADI depended inversely on the thickness of the sample. For instance, a resolution target was detectable with ADI in a 5 cm thick optical cell filled with 0.14% Intralipid™, a 2 cm thick optical cell filled with 0.3% Intralipid™, and a 1 cm thick optical cell filled with 0.7% Intralipid™. This confirmed that the product of the reduced scatter coefficient (proportional to Intralipid™ concentration) and the thickness was a predictor of the detection performance of ADI. Based on these findings, we chose 0.25% and 0.3% Intralipid™ in a 2 cm optical cell to study the performance of ADI at scattering levels below and in the range of the detection limit. The experiments used the 808 nm laser diode as the source of illumination. According to Mie theory , the reduced scattering coefficient, µ’s for 0.25% and 0.3% Intralipid™ solutions is approximately 2 cm-1 and 2.4 cm-1, respectively. However, the effective scattering level at the angular filter array was higher due to the refractive index mismatch between the Intralipid™-glass, and glass-air boundaries.
5.2 Image contrast enhancement by wedge-based subtraction (ADI-WS)
The results of the ADI-WS image contrast enhancement techniques are shown in Fig. 6(a)–6(c). Without the wedge in place (Fig. 6(a)), the resolution target was discernable, but the contrast was poor. With the wedge prism inserted into the light path, the resolution target could not be visualized (Fig. 6(b)). This was to be expected since the quasi-ballistic photons were directed out of field of view of the AFA, which left only scattered light to leak through the AFA and reach the camera. Digital image subtraction of the two images revealed an image of the resolution target with better contrast compared to ADI without the wedge (i.e. compare Fig. 6(c) to Fig. 6(a)). The superior image contrast of the subtracted image as shown in Fig. 6(c) was due to the removal of the majority of the background scattered light.
As described earlier, the background scattered light becomes more dominant in ADI scans as the scattering level of the sample increases. This was evident in Fig. 6(a) where there was poor contrast between the resolution target and background for the 2 cm thick optical cell filled with 0.3% Intralipid™. Based on our previous work , digital image processing techniques could be used to increase the visibility of the lines and spaces to some degree, but the background scattered light image subtraction method (Fig. 6(c)) shows much better contrast compared to image processing alone.
5.3 Image contrast enhancement by polarization-based subtraction (PADI)
Figures 6(d)–6(e) show the results of the polarized ADI images with the linear polarizers oriented with axes parallel (Fig. 6(d)) and axes perpendicular (Fig. 6(e)). Digital image subtraction of the two images revealed an image of the resolution target with better contrast compared to ADI with the polarizers in the parallel orientation (i.e. compare Fig. 6(f) to Fig. 6(d)). However, it was evident that the contrast enhancement was qualitatively not as good as the ADI-WS technique (i.e. compare Fig. 6(c) to Fig. 6(f)). Although it was assumed that the perpendicular polarization measurement did not include quasi-ballistic light, it was likely that some quasi-ballistic photons suffered random changes in their polarization and leaked through the cross polarizers. This leakage added to the estimated background scattered light and degrade image contrast.
5.4 Image contrast enhancement by combined use of wedge and polarization-based subtraction (PADI-WS)
Comparing the image contrast between ADI and PADI, parallel polarization measurements do not appear to increase the image contrast significantly, only a small improvement is observed. This was likely due to some informative quasi-ballistic photons losing their initial polarization followed by attenuation by the second polarizer in a similar manner to randomly scattered light. Our experiment at the lower scattering level (0.25% Intralipid™ in the 2 cm optical cell) showed that with crossed polarized ADI measurements, the L-shape targets were still faintly observable (not discernable in presented image), which indicated that some of the informative signal changed polarization and passed through the cross polarizer. Hence, cross polarized ADI cannot be effective as a pure background scattered light correction method like ADI-WS. The following section will compare all the presented methods in a more quantitative manner.
5.5 Comparison of image contrast enhancement methods
In order to compare the results of wedge, polarization, and combined subtraction procedures in terms of image contrast enhancement, the contrast of the contrast-enhanced images was calculated based on the following formula:
where Imin is the average light intensity corresponding to shadowed region (i.e. due to the shadow of the L-shape target within the phantom) and Imax is the average light intensity corresponding to the unshadowed local area near the structure.
Fig. 7 shows the image contrast computed from Eq. (7) for the results collected at two scattering levels of the sample. For standard ADI, the image contrast declined from 0.4 to 0.08 when the scattering level of the sample increased due to an increase in concentration of Intralipid™ from 0.25% to 0.3%. As Fig. 7 shows, addition of the linear polarizers to ADI (PADI) improved the image contrast by only a minor amount compared to ADI alone. This was likely due to the poor discrimination of the polarizers between quasi-ballistic and scattered photons. On the other hand, estimation of the background scattered light by the wedge prism technique with subsequent image subtraction (ADI-WS) resulted in improved image contrast compared to ADI alone. A similar result was observed for the combination of the wedge subtraction technique with polarized angular domain imaging (PADI-WS). The combined technique provided the highest level of image contrast improvement at the detection limit of ADI (i.e. 0.3% Intralipid™). The subtraction of the cross polarizer measurements from the parallel polarizer measurements (PADI-CS) provided an intermediate level of improvement in image contrast, and was inferior to the ADI-WS and PADI-WS techniques.
Although ADI is proving to be a powerful technique for imaging turbid samples, methods for contrast enhancement in ADI are likely needed if the technique is to be applied to imaging biological systems. Here we reported significant improvements in ADI image contrast by placement of readily available optical components into the light path combined with digital image subtraction. These methods captured pixel by pixel estimates of the background scattered light levels without compromising optical alignment of the setup. This significant side benefit ensured that contrast enhancement during transillumination ADI could be facilitated without disturbing the highly aligned collimated light source and AFA. For turbid samples where the scattering level was low enough for ADI to detect targets easily, the wedge subtraction technique (i.e. ADI-WS) had the greatest impact on improving image contrast with a greater than 2-fold improvement over ADI alone. For turbid samples were the scattering was at a level beyond which ADI could reliably detect embedded objects, the combination of the wedge and polarizer-based subtraction techniques (i.e. PADI-WS) gave the greatest improvement in image contrast. This suggested that although the use of polarizers in ADI was not as good at improving image contrast as the wedge-based ADI method, the combination of the two was superior to either technique alone. Further improvements to ADI image contrast may be possible by replacing the camera with one of higher bit precision and lower noise so that contrast differences below 0.08% can be distinguished. These approaches will likely push the capabilities of ADI into the realm of high resolution imaging of highly scattering (equivalent to 1% Intralipid™ and beyond) biological specimens up to 1 cm in thickness, which would enable multispectral imaging for tissue characterization and functional studies at a spatial resolution exceeding 200 µm. These capabilities could prove useful for histological analysis of thicker tissue specimens over a larger field of view compared to current optical methods. This might find a direct application in optical analysis of entire biopsy specimens rather than selected sections as is currently the case.
References and links
2. A. Roggan, M. Friebel, K. Dorschel, A. Hahn, and G. Muller, “Optical Properties of Circulating Human Blood in the Wavelength Range 400 2500 nm,” J. Biomed. Opt. 4, 36–46 (1999). [CrossRef]
3. T. Wilson, ed., Confocal Microscopy (Academic, London, 1990).
4. A. F. Fercher, “Optical coherence tomography-principles and applications,” Rep. Prog. Phys. 66, 239–303 (2003). [CrossRef]
5. J. G. Fujimoto, “Optical coherence tomography,” C. R. Acad. Sci. Ser. IV Phys. Astrophys. 2, 1099–111 (2001).
8. N. Pfeiffer, P. Chan, G. H. Chapman, F. Vasefi, and B. Kaminska, “Optical imaging of structures within highly scattering material using a lens and aperture to form a spatiofrequency filter,” Proc. SPIE 6854, 68541D (2008). [CrossRef]
9. G. H. Chapman, M. Trinh, N. Pfeiffer, G. Chu, and D. Lee, “Angular Domain Imaging of Objects Within Highly Scattering Media Using Silicon Micromachined Collimating Arrays,” IEEE J. Sel. Top. Quantum Electron. 9, 257–266 (2003). [CrossRef]
10. F. Vasefi, B. Kaminska, P. K. Y. Chan, and G. H. Chapman, “Multi-spectral angular domain optical imaging in biological tissues using diode laser sources,” Opt. Express 16, 14456–14468 (2008). [CrossRef] [PubMed]
11. S. Chandrasekhar, Radiative Transfer, (Oxford University Press, London, 1950).
13. W. F. Cheong, “A review of the optical properties of biological tissues,” IEEE J. Quantum Electron. 26, 2166–2185 (1990). [CrossRef]
14. L. Henyey and J. L. Greenstein, “Diffuse radiation in the galaxy,” Astrophysics. J. 93, 70 (1940). [CrossRef]
15. S. L. Jacques, C. A. Alter, and Prahl, “Angular Dependence of HeNe laser Light Scattering by Human Dermis,” Laser Life Sci. 1, 309–333 (1987).
16. T. Vo-DinhBiomedical Photonics Handbook, (Publisher: CRC, 2003). [CrossRef]
17. V. V. Tuchin, “Fundamentals of low-intensity laser radiation interaction with biotissues: dosimetry and diagnostical aspects Bull.Russ. Acad. Sci. Phys. Ser. 59, 120–143 (1995).
19. F. Vasefi, P. K. Y. Chan, B. Kaminska, G. H. Chapman, and N. Pfeiffer, “An Optical Imaging Technique Using Deep Illumination in the Angular Domain,” IEEE J. Sel. Top. Quantum Electron. 13, 1610–1620 (2007). [CrossRef]
20. F. Vasefi, B. Kaminska, G. H. Chapman, and J. J. L. Carson, “Angular distribution of quasi-ballistic light measured through turbid media using angular domain optical imaging,” submitted to SPIE Photonics West: Optical Interactions with Tissue and Cells XX, (2009).
21. F. Vasefi, B. Kaminska, and G. H. Chapman, “Angular Domain Optical Imaging using a Micromachined Tunnel Array and a Keplerian Lens System” in 30th IEEE EMBS Annual International Conference , 3730–3734 (2008).
22. P. K. Chan, F. Vasefi, G. H. Chapman, B. Kaminska, and N. Pfeiffer, “Angular Domain Optical Tomography in Scattering Media with Multi-spectral Diode,” Proc. SPIE 6435, 64350M (2007). [CrossRef]
23. S. Jacques, Optical properties of “Intralipid™”, an aqueous suspension of lipid droplets, http://omlc.ogi.edu/spectra/intralipid/index.html.
24. H. G. van Staveren, C. J. M. Moes, J. van Marle, S. A. Prahl, and M. J. C. van Gemert, “Light scattering in Intralipid-10% in the wavelength range of 400–1100 nanometers,” Appl. Opt. 30, 4507–4514 (1991). [CrossRef] [PubMed]
25. F. Vasefi, G. H. Chapman, P. K. Y. Chan, B. Kaminska, and N. Pfeiffer, “Enhanced angular domain optical imaging by background scattered light subtraction from a deviated laser source,” Proc. SPIE 6854, 68541E (2008). [CrossRef]