2.1.

Experimental Setup

To examine planar reflectance and transillumination measurements we employed a scanner similar to the one previously described for small animal tomography operating^{13, 21} in the near IR (NIR). A schematic of the system implementing planar geometry is shown in Fig. 1 . Illumination was provided by a $672\text{-}\mathrm{nm}$ cw laser diode (B&W Tek Newark, Delaware) with adjustable light power reaching a maximum of $100\mathrm{mW}$ , although typical powers employed in all experiments ranged within $0.5\text{to}2\mathrm{mW}$ . Laser light is routed through a $1\times 2$ optical switch (Dicon Fiberoptics Inc., Richmond, California) either to beam expanding optics for reflectance imaging (front illumination) or to a programmable optical switch (Dicon Fiberoptics Inc. Richmond, California) for transillumination. The switch can time-share the light input to 46 source fibers ( $100∕140\mu \mathrm{m}$ core/clad multimode fibers). Animals are positioned into the imaging chamber that is shown in more detail in Fig. 2 . The chamber is made water tight and it utilizes a front glass window that is antireflection coated for the NIR and a moving plate that can gently compress the animal against the glass window while placing the 46 source fibers onto the opposite of the animal-window interface, as shown in Fig. 2. Typical separations between the moving plate and the glass window were $1.3\mathrm{cm}$ , which is typical for small animal imaging. The 46 source fibers are evenly distributed over a field of view of $1.8\times 1.5\mathrm{cm}$ onto the moving plate. Light is collected through the glass window with a CCD camera (D) ( $512\times 512\text{pixels}$ back-illuminated VersArray from Roper Scientific/Princeton Instruments, Trenton, New Jersey) using a $35\text{-}\mathrm{mm}$ $f1.2$ lenses (Nikon Inc., Melville, New York) and bandpass filters for emission and excitation measurements ( $705\pm 5$ and $671\pm 5\mathrm{nm}$ central wavelengths, respectively, Andover Inc., Salem, New Hampshire).

Fig. 1

Experimental setup used in the study implementing concurrent reflectance and transillumination measurements.

Fig. 2

Implementation of reflectance and transillumination in the imaging chamber. (a) In reflectance mode, a beam of light is expanded onto the animal surface and light emitted back from the animal (in excitation and emission wavelengths) is collected with the CCD camera of Fig. 1 from the same side of the animal. (b) In transillumination, a group of fibers establishes an excitation beam on one side of the animal and light emitted through the animal is collected from the other side.

2.2.

Planar Imaging and Transillumination

The implementation of planar reflectance and transillumination geometries is illustrated in Fig. 2. Planar reflectance imaging is realized by front illuminating the imaging chamber through the glass window. Back-emitted intrinsic or excitation light is then captured by the CCD camera, which is focused on the inside surface of the glass window.

Transillumination can be implemented in several different ways. In this work, transillumination is accomplished by superposition of the signals collected through the diffuse medium imaged due to the 46 point sources at the back side of the chamber. This selection of back-illumination using superposition of point sources was herein directed by the concurrent tomographic purpose that the scanner serves. Potentially, other patterns could be employed, including expanded light beams similar to reflectance imaging.

2.3.

Measurements

We investigated normalized measurement methods and compared them to standard uncorrected signals. Typically, six images were obtained for each experiment, i.e., the front-illuminated fluorescence image (reflectance image), the back-illuminated image (transillumination), and the dark current image (noise) at both the excitation (intrinsic) and emission (fluorescence) wavelengths. We then observed the nonnormalized reflectance fluorescence image $I_r$ , i.e.,

Subsequently, we also examined the corresponding images after normalization with the intrinsic images acquired. The corresponding normalized reflectance image $U_r$ can be written as

Similarly, the corresponding normalized transillumination image $U_t$ can be written as

2.4.

Phantoms

Two sets of phantoms were used. The first employed a relatively spatially homogeneous phantom containing two diffusive and fluorescent tubes placed $5\mathrm{mm}$ apart and immersed in a highly scattered fluid contained in the chamber. The tubes were glass capillaries of $\sim 1.5\mathrm{mm}$ diameter, sealed on one end, and contained a 1% intralipid solution, $25\mathrm{ppm}$ of India ink, and $200\mathrm{nM}$ of Cy 5.5 dye. The chamber was filled with the same solution of intralipid and ink, but without the fluorochrome. Subsequently, the two-tube phantom was placed at different depths in the imaging chamber to assess the relative performance of the techniques employed as a function of depth.

The second phantom examined the performance of the planar method examined to image two fluorescent tubes at varying background optical properties. These measurements employed a similar, relatively homogenous background medium, which contained two $3\text{-}\mathrm{mm}$ -diam plastic tubes placed in contact with the chamber window. Both tubes and the imaging chamber contained with the same solution of 1% intralipid solution and $25\mathrm{ppm}$ amount of India ink that gives light attenuation similar to that of mouse boundaries, as determined experimentally in other studies.²¹ The absorption coefficient of the fluid was ${\mu }_a=0.3{\mathrm{cm}}^{-1}$ and the reduced scattering coefficient ${\mu }_s^{\prime }=10{\mathrm{cm}}^{-1}$ . In addition, the left tube contained $400\mathrm{nM}$ of Cy 5.5, whereas the right tube contained $200\mathrm{nM}$ of Cy 5.5. After this phantom was imaged, the absorption of the left tube was increased by adding India ink at $2.5\times$ the original concentration and one imaging session was again performed. This experiment is similar to a previously reported study that assessed the tomographic performance under such conditions.¹³

2.5.

Animal Imaging

Two sets of animal measurements were employed to showcase the relative merits of the techniques examined.

The first animal model consisted of a euthanized nude mouse with a $1.8\text{-}\mathrm{mm}$ -diam glass tube inserted into the esophagus. The tube was filled with 1% intralipid and $400\mathrm{nM}$ of Cy5.5 dye and was positioned at the level of the midtorso of the animal. The second in vivo model was an MMTV/neu transgenic mouse, which exhibits multifocal spontaneous mammary tumorigenesis [FVB/N-TgN(MMTVneu)202Mul]. An MMTV/neu mouse with two spontaneous mammary tumors was injected with $2\mathrm{nmol}$ of the cathepsin-sensitive activatable fluorescent probe ProSense680 (Visen Medical Inc., Woburn, Massachusetts).

3.1.

Imaging Performance as a Function of Depth

Figure 3 shows the measurements collected for the two fluorescent capillaries placed at different depths. The top row shows fluorescence images obtained in reflectance mode when the tubes were 0, 1, 3, 5, and $7\mathrm{mm}$ away from the front glass window. At $0\mathrm{mm}$ , the tubes are fully immersed in the diffusive fluid and physically in contact with the glass window facing the CCD camera. At larger depths, the tubes move away from the camera and the glass window into the diffusive fluid, therefore coming closer to the transillumination sources. The bottom row shows corresponding images obtained in normalized transillumination mode. We note that in Fig. 3, plain reflectance images (not normalized) are shown. This is because the intrinsic normalized reflectance images cannot record the tubes at most depths and they see only the excitation field reflected off the diffuse surface. Therefore, they do not offer correction for deep-seated activity. All images are scaled to their maximum, since signal intensity drops exponentially with depth.

Fig. 3

Reflectance and normalized transillumination performance as a function of depth. Top row: Reflectance images acquired with two fluorescence tubes placed at different depths as indicated on the figure titles. Bottom row: Corresponding normalized transillumination images of the same tubes and depths examined.

The tubes are detected well in all normalized transillumination images, although resolution significantly deteriorates as a function of depth. Conversely, reflectance imaging resolves the tubes better at $0\mathrm{mm}$ than transillumination, but tube detection is overall more challenging as a function of depth. Within a millimeter away from the front window, both methods fail to resolve the two tubes. In addition, there is a characteristic broadening of the tubes, as seen in the transillumination images when moving away from the center where the sources are concentrated, as indicated by the white arrows on the 1- and $7\text{-}\mathrm{mm}$ transillumination images.

This performance of transillumination imaging and key features of the normalization process are better explained in Fig. 4 . Shown in Fig. 4(a) is the intrinsic field collected in transillumination mode [ $I_{\mathrm{et}}$ in Eq. 4]. The field seen maps an elliptical area of $\sim 3\times 2\mathrm{cm}$ on the front window and its intensity exponentially drops outside this area. Figure 4(b) shows the fluorescence image collected with the tubes of Fig. 3 at a $5\text{-}\mathrm{mm}$ depth in transillumination mode. This is the nonnormalized image $I_t$ of Eq. 2. Figure 4(c) is the normalized fluorescence image $U_t$ described by Eq. 4, which better reveals the full length of the tubes compared to Fig. 4(b), beyond the limited area covered homogeneously by the intrinsic field. In principle, one would aim to homogeneously illuminate the complete region of interest, however, this finding indicates how the normalized method works well even at inhomogeneous back-illumination. However, due to the asymmetry of the illumination field, there is an asymmetry at the “shape” resolved at the extremes of the image (indicated with the arrows in Fig. 3).

Fig. 4

Characteristic field involved in normalized transillumination images. The intrinsic image is acquired at the excitation wavelength and the fluorescence image is the corresponding image acquired at the emission wavelength for two fluorescence tubes at $5\mathrm{mm}$ away from the front window. The normalized image, formed according to Eq. 4 described in the text, better outlines the extent of the tubes even outside the homogeneous span of the intrinsic field.

3.2.

Imaging Performance as a Function of Optical Property Variation

Figure 5 depicts the relative performance of normalized reflectance and transillumination for varying background optical properties. This experiment imaged the two $3\text{-}\mathrm{mm}$ -diam tubes immersed in the chamber and placed in contact with the glass window facing the CCD camera. The two tubes and surrounding diffusive fluid are evident in Figs. 5(a) and 5(g), which were collected at the excitation wavelength. The left tube further contained $400\mathrm{nM}$ and the right tube $200\mathrm{nM}$ of Cy5.5, as described in the methods section.

Fig. 5

Investigation of imaging performance as a function of optical property variation. (a) to (f) Appearance of two fluorescence tubes immersed in the diffuse fluid and placed on the glass window. The left tube contains twice the fluorochrome content of the right tube (400 vs. $200\mathrm{nM}$ of the Cy5.5 fluorochrome). The optical properties in the tubes and the surrounding medium match the average optical properties of mouse boundaries in the NIR. All fluorescence imaging methods reveal correctly a 2:1 relation between fluorescence field detected from the left versus the right tube respectively. (g) to (l) Same imaging results when the absorption in the left tube is raised to $2.5\times$ concentration compared to the right tube. The insert text boxes report on the fluorescence strength ratio between the left and the right tubes.

When imaged in reflectance and transillumination both tubes demonstrated a 2:1 $(\pm 5\%)$ ratio in average photon counts between left and right tubes, as calculated over identical regions of interest. Subsequently, India ink was added to the left tube at a concentration of 2.5 times the background concentration. The imaging results, collected after the ink titration, are shown in Figs. 5(g), 5(h), 5(i), 5(j), 5(k), 5(l). The reflectance and transillumination measurements fail now to accurately report the 2:1 ratio of the fluorochrome concentration present in the tubes, because light is more absorbed in the left tube compared to the right tube due to the added absorber. The ratio’s obtained for the four fluorescence imaging approaches are summarized on the corresponding images on the text boxes in Figs. 5(h), 5(i), 5(j), 5(l).

The normalized transillumination image [Fig. 5(l)] shows the most accurate result yielding a ratio of 1.58:1 followed by normalized reflectance at 1.41:1 [Fig. 5(i)]. The worst performance (1.11:1) is seen by standard transillumination [Fig. 5(k)], where the two tubes appear to have virtually identical fluorescence. It is characteristic that even though the absorption change of the left tube is not as evident on the reflectance excitation image as it is on the excitation transillumination image (due to the white plastic tube encasing), it can nevertheless improve the quantification performance of the normalized reflectance image over the standard reflectance image. In a similar experiment,¹³ tomography exhibited a ratio of $\sim 1.8:1$ .

3.3.

Small Animal Imaging

Figure 6 shows results obtained from the nude mouse implanted with a fluorescent tube after euthanasia. Figures 6(a), 6(b), 6(c) depict the intrinsic, fluorescence and normalized fluorescence image obtained in reflectance mode. Figure 6(a) also shows the approximate position of the tube in the viewing plane, found by measuring the insertion distance. None of the reflectance fluorescent images depict contrast from the implanted tube since, similarly to the experiment of Fig. 3, reflectance imaging is not well suited for resolving objects that are deeper than a few millimeters. Significant autofluorescence appears in Fig. 6(b), especially in the upper torso area, as indicated by the arrow. Conversely, normalization aids in obtaining a more uniform signal with less artifacts, as is apparent in Fig. 6(c), which shows lower background contrast.

Fig. 6

Postmortem imaging of a fluorescence tube inserted in the center of the animal through the esophagus. Top row: Intrinsic, fluorescence, and normalized reflectance images. Middle row: Intrinsic, fluorescence, and normalized transillumination images with the animal immersed in a matching diffusive medium. Bottom row: Intrinsic, fluorescence, and normalized transillumination images in the absence of a matching diffusive medium. The transillumination images are plotted in an inverse gray-scale color map. The fluorescent tube is clearly resolved in the transillumination fluorescence images as indicated by the arrows but not on the reflectance images. The best imaging performance is obtained with the normalized transillumination method.

Figures 6(d), 6(e), 6(f), depict the intrinsic, fluorescence, and normalized transillumination images obtained in transillumination mode. Here the results are shown as negative images, i.e., darker regions indicate stronger signal. Confirming the results of Fig. 3, the tube is well detected and it is better resolved in the normalized transillumination image. The tube size is overestimated but its 2-D location is well resolved compared to its known position. While Figs. 6(d), 6(e), 6(f) show results obtained at the presence of a matching fluid surrounding the animal, Figs. 6(h), 6(i), 6(j) depict similar results in the absence of a matching fluid. The tube can be well observed, and in this case, the mouse outline is also better shown. The use of matching fluids may be useful in transillumination measurements for attenuating intrinsic light close or outside the animal boundaries from directly impinging on the photondetector and saturating the measurement. However, the absence of matching fluids offers significant experimental simplicity. In this case, special care should be taken to spatially attenuate the intrinsic light used to match the animal outline and better interface with the dynamic range of the photon detector used. In the case of Figs. 6(h), 6(i), 6(j), the source distribution covered a smaller area than that covered by the animal body hence no saturation effects were observed.

Figure 7 shows an in vivo imaging experiment chosen from an animal bearing an ellipsoid-shaped mammary tumor of $4\times 3\mathrm{mm}$ (long and short axis dimensions, respectively, as determined by caliper measurements). The tumor is indicated by an arrow in the images. This tumor was highly vascular and highly absorbing, and was therefore seen as darker than surrounding tissue on the intrinsic reflectance image shown in Fig. 7(a). The tumor, however, is not visible in the corresponding fluorescence reflectance image shown in Fig. 7(b). Similarly to the phantom results shown in Fig. 5, increased lesion absorption significantly decreases contrast in nonnormalized images. Figure 7(c) shows the normalized fluorescence reflectance image. Similarly to Fig. 6(c), this image appears to be less affected by background signals than Fig. 7(b). The tumor is also now marginally identifiable based on fluorescence signals. Overall, however, detection ability suffers mainly due to other surface fluorescence activity (auto-fluorescence) of comparable magnitude also observable from the animal in this case. Figures 7(d), 7(e), 7(f) depict the intrinsic, fluorescence, and normalized transillumination images, also plotted in reverse scale, where a darker image indicates a stronger signal. The tumor is not visible in Fig. 7(e), and demonstrates 44% lower signal intensity compared to the stronger of background signals. However, in the normalized transillumination image shown in Fig. 7(f), the tumor can be well seen and demonstrates a 50% fluorescence increase over the stronger of background signals as a result of the normalization method. While reflectance imaging preferentially samples surface activity, transillumination probes tumors volumetrically and yields higher detection sensitivity and contrast in this case after normalization.

Fig. 7

In vivo imaging of a subcutaneous mammary tumor. The location of the tumor is indicated by an arrow. Top row: Intrinsic, fluorescence, and normalized reflectance images. Bottom row: Intrinsic, fluorescence, and normalized reflectance images with the animal immersed in a matching diffusive medium.