We present for the first time experimental evidence that the quantitative accuracy of bioluminescence tomography (BLT) can be significantly improved by incorporating prior spatial distribution of optical properties of heterogeneous media obtained from diffuse optical tomography (DOT). A series of experiments were conducted using a CCD-based scanning system where millimeter-size bioluminescent targets were embedded in a 3×3×5cm optically heterogeneous scattering medium. The results show that the BLT images with the recovered optical property distribution in place are considerably better reconstructed compared to that without such prior information, in terms of the location, size and source strength of the targets.
©2008 Optical Society of America
Bioluminescence imaging (BLI) has emerged as a powerful tool for small animals. BLI uses luciferase as an internal biological light source that can be genetically programmed to noninvasively report the presence or activation of specific biological events. Thus far BLI has found its wide applications for small animal imaging, ranging from tracking tumor cells, stem cells, immune cells, and bacteria to imaging gene expression [1–2]. The main advantage of bioluminescence technique over fluorescence is that there is no inherent background bioluminescence in most tissues, making it an extremely sensitive technique. However, the current BLI technique has two notable limitations: (1) BLI is two-dimensional (2D) and unable to obtain depth information, and (2) It does not allow absolute quantification of target signals.
Bioluminescence tomography (BLT) has been recently developed to overcome the two major limitations mentioned above. BLT is concerned with an inverse problem where the 3D geometry of the object and quantitative spatial distribution of the bioluminescence signals are obtained from the measured full-field-of-view (360°) bioluminescent data at the animal surface using a reconstruction algorithm. We have reported the first experimental BLT images with a finite element based reconstruction algorithm , while Wang et al.  established a mathematical foundation on the uniqueness for reconstruction of a bioluminescence source distribution in BLT. Since these early works, various finite element-based reconstruction algorithms have been thus far implemented and tested/validated using simulations and tissue phantom experiments [5–8]. Initial in vivo results have also been reported where a brain or prostate tumor expressing firefly luciferase were tomographically detected [6, 9].
While in vivo BLT imaging is now possible, its quantitative accuracy is notably degraded especially for small bioluminescent targets, because of the lack of accurate knowledge of tissue optical property distribution in the animal. Extensive simulation studies have shown the significant impact of optical property estimation accuracy on BLT reconstructions . Clearly taking published data about optical properties of animal tissues from the literature is ultimately not acceptable, as tissue optical properties can be very different for individual animals. To cover this obstacle, in this paper we describe a multi-modality approach by combining BLT with diffuse optical tomography (DOT) that is able to provide accurate tissue optical property distribution in the animal/heterogeneous media. Using tissue-like phantom experiments, we show that the quantitative accuracy of BLT can be enhanced dramatically when the optical property distribution is known a priori.
In the experiments, the cubic optically heterogeneous phantom was mounted on a 360° rotation stage. For the BLT data collection, the phantom was rotated three 90°, and luminescence light emitted from each side of the phantom was collected by a thermoelectrically cooled (-70°C), back illuminated CCD array (Princeton Instruments/Acton, Trenton, NJ) coupled with an optical lens (Zoom 7000, Navitar, Rochester, NY). The data collection for DOT was realized by adding a diode laser at 655nm to the CCD system as excitation source coupled with a linear stage for multiple source position scanning. The test geometry is shown in Fig. 1 where “lung”, “liver” and “heart” were embedded in a 3×3×5cm background medium, similar to the anatomy of a mouse body [the optical properties for these “organs” are similar to that from the literature  and shown in the caption of Fig. 1]. The phantom materials used consisted of Intralipid, India ink and Agar powder. We placed one or two small luminescent targets (Φ1×2mm each) containing mixed chemical solution from luminescent light stick (Glowproducts, Victoria, British Columbia, Canada) in positions near the “lung” (see T1, T2 and T3 in Fig. 1(b)). The mixed chemical solution emit red light around 650nm. A total of 660 measurements were used for BLT reconstruction and 4 (side)×45 (source)×81 (detector) measurements were applied for DOT reconstruction.
Our reconstruction algorithms for 3D BLT and DOT have been detailed previously [3, 11]. Briefly, the image formation algorithms cast the inverse problem associated with determining the bioluminescence source distribution or optical characters of the medium (e.g., animal tissue) as a linear (for BLT) or nonlinear (for DOT) parameter estimation where measured optical data are used to find a “best fit” of the bioluminescence source distribution or tissue optical parameters needed to reproduce the known measured information.
3. Results and discussion
Figure 2 presents the recovered 3D absorption and scattering images of the phantom at selected transverse planes. We see that the “lung”, “heart” and “liver” are clearly reconstructed in terms of position and size. The optical properties recovered for these “organs” are also quantitatively accurate (see the caption of Fig. 1 for the actual absorption and reduced scattering coefficients of the “organs”).
Figures 3 and 4 show the reconstructed 3D BLT images for a single target at T1 (Fig. 3) or T2 (Fig. 4) position having different source strength without (i.e., assuming homogenous optical background) Figs. 3(a), 3(b), 4(a), and 4(b) and with Figs. 3(c), 3(d), 4(c), and 4 (d) DOT reconstruction. While the target is detected for both cases with homogeneous background assumption, we note that the target size is considerably enlarged and that the target shape is significantly distorted especially for T2 position (Fig. 4(a)). We also see that the target position is shifted in x-y plane relative to the exact target position for T2 position (Figs. 4(a) and 4(b)). When the target is located between the two “lungs” having high absorption and scattering coefficients relative to the background (i.e., T1 position), the target dimension in z-direction is significantly overestimated (Fig. 3(b)). In particular, for both cases the source distribution is recovered with large error compared to the exact source strength of 30nW/mm3 for T1 position (Figs. 3(a) and 3(b)) and of 61nW/mm3 for T2 position (Figs. 4(a) and 4(b)).
When the optical properties obtained from DOT reconstruction are used in BLT reconstruction, we see that both targets are significantly better and quantitatively resolved in terms of the target position, size and, especially, the source strength (Figs. 3(c), 3(d), 4(c), and 4(d)). The improvement with DOT reconstruction is particularly remarkable for the two-target case as shown in Fig. 5. Here we see that the two targets cannot be resolved without DOT guidance (Fig. 5(a)). We also see significant shift of target position and distortion of target shape as well as significant underestimation of the source strength when the optical properties are assumed constant (Figs. 5(a) and 5(b)). With prior optical distribution obtained from DOT reconstruction, the two targets are clearly resolvable (Fig. 5(c)) and the target position and size are recovered with reasonable accuracy. In addition, the recovered source strength is close to the actual value of 60nW/mm3. While we demonstrate clear improvement with DOT assisted BLT reconstruction, we also note that the recovered sizes for all the cases are generally overestimated due to the scattering limited resolution of DOT and BLT. The target shape for T1 position is notably distorted along x-direction (Figs. 3(c) and 3(d) which could be caused by the relatively poor reconstruction of the scattering images of the two “lungs” (see Fig. 2(b)). We suspect that these inaccuracies in DOT reconstruction are most likely caused by the insufficient number of excitation light source positions and detector positions used in the experiments. We expect that an optimized number of excitation source and detector arrangement coupled with advanced mesh schemes  will be able to overcome the problem and to provide improved accuracy for DOT/BLT reconstructions.
In summary, we experimentally demonstrate that BLT reconstruction can be enhanced quantitatively when the distribution of optical property is known as a priori from DOT. This ability of combining DOT with BLT will allow us to obtain truly quantitative images of bioluminescent targets in small animals where optical property distribution is highly heterogeneous.
References and links
1. S. Thorne and C. Contag, “Using in vivo bioluminescence imaging to shed light on cancer biology,” Proc. of the IEEE 93, 750–762 (2005). [CrossRef]
2. J. Park and S. Gambhir, “Multimodality radionuclide, fluorescence, and bioluminescence small-animal imaging,” Proc. of the IEEE 93, 771–783 (2005). [CrossRef]
5. W. Cong, G. Wang, D. Kumar, Y. Liu, M. Jiang, L. V. Wang, E. A. Hoffman, G. McLennan, P. B. McCray, J. Zabner, and A. Cong, “Practical reconstruction method for bioluminescence tomography,” Opt. Express 13, 6756–6771 (2005). [CrossRef] [PubMed]
6. A. J. Chaudhari, F. Darvas, J. R. Bading, R. A. Moats, P. S. Conti, D. J. Smith, S. R. Cherry, and R. M. Leahy, “Hyperspectral and multispectral bioluminescence optical tomography for small animal imaging,” Phys. Med. Biol. 50, 5421–5441 (2005). [CrossRef] [PubMed]
7. G. Alexandrakis, F. R. Rannou, and A. F. Chatziioannou, “Tomographic bioluminescence imaging by use of a combined optical-PET (OPET) system: a computer simulation feasibility study,” Phys. Med. Biol. 50, 4225–4241 (2005). [CrossRef] [PubMed]
9. G. Wang, W. Cong, K. Durairaj, X. Qian, H. Shen, P. Sinn, E. Hoffman, G. McLennan, and M. Henry, “In vivo mouse studies with bioluminescence tomography,” Opt. Express 14, 7801–7809 (2006). [CrossRef] [PubMed]
10. G. Alexandrakis, F. R. Rannou, and A. F. Chatziioannou, “Effect of optical property estimation accuracy on tomographic bioluminescence imaging: simulation of a combined optical-PET (OPET) system,” Phys. Med. Biol. 51, 2045–2053 (2006). [CrossRef] [PubMed]
11. H. Jiang, Y. Xu, and N. Iftimia, “Experimental three-dimensional optical image reconstruction of heterogeneous turbid media from continuous-wave data,” Opt. Express 7, 204–209 (2000). [CrossRef] [PubMed]