• We are sorry, but NCBI web applications do not support your browser and may not function properly. More information
Logo of pnasPNASInfo for AuthorsSubscriptionsAboutThis Article
Proc Natl Acad Sci U S A. Dec 20, 2005; 102(51): 18252–18257.
Published online Dec 12, 2005. doi:  10.1073/pnas.0504628102
PMCID: PMC1317905
Applied Physical Sciences

Volumetric tomography of fluorescent proteins through small animals in vivo

Abstract

Volumetric detection and accurate quantification of fluorescent proteins in entire animals would greatly enhance our ability to monitor biological processes in vivo. Here we present a quantitative tomographic technique for visualization of superficial and deep-seated (>2–3 mm) fluorescent protein activity in vivo. We demonstrate noninvasive imaging of lung tumor progression in a murine model, as well as imaging of gene delivery using a herpes virus vector. This technology can significantly improve imaging capacity over the current state of the art and should find wide in vivo imaging applications in drug discovery, immunology, and cancer research.

Keywords: fluorescence, whole-body imaging, imaging gene expression, gene transfer, multispectral imaging

Fluorescent proteins (FP) have become essential reporter molecules for the elucidation of the function of proteins within cells, the biodistribution of immune and stem cells, and evaluation of drug candidates in vivo (1, 2). Confocal and multiphoton microscopy have emerged as powerful methods for imaging superficial locations down to a few hundred micrometers and in limited field of views (typically 1–2 mm) (3). Alternative technologies to date use macroscopic epi-illumination imaging using sensitive color cameras that obtain “photographs” of fluorescence activity emanating from up to a few millimeters under the skin surface of animals (4). Such images are often biologically informative, but they are surface-weighted, i.e., they preferentially image superficial over deep-seated activity because the intensity of the fluorescence signal recorded drops with a strong nonlinear dependence as a function of depth. In addition, photographic imaging yields a single projection that cannot resolve depth; signal quantification is complicated with varying tissue optical properties, for example because of angiogenesis; and superficial fluorescence signals or skin autofluorescence can reduce or even shield deep-seated contrast.

To improve on imaging of FPs in whole animals, it is important to develop methods that account for the nonlinear propagation effects of photons into tissues and resolve depth. Such technology has been recently reported for tomographic imaging of extrinsically administered fluorescence probes in the near-infrared (NIR) (5, 6). However, the FPs currently available absorb in the visible (<600 nm) and many also emit in the visible (7). The translation of tomographic methods for operation in the 450–600 nm spectral region comes with some unique challenges. These challenges arise because tissue offers significantly higher photon attenuation in the visible necessitating different instrumentation than implementations that have appeared so far for imaging in the NIR. Similarly, adept theoretical models are required to model the corresponding change of photon propagation patterns in highly absorbing media (8). We have recently developed a system and method for tomographic imaging in the visible (9) achieving 400-μm resolution at 6-mm depth in homogenous slabs of 1.2-cm thickness and the optical properties of mice at the green, and ≈7.5 pmol detection sensitivity of the FITC dye placed at ≈6-mm depth of euthanized nude mice compressed to 1.2-cm thickness. The system uses noncontact technology, i.e., delivers and detects light onto tissue without the use of fibers. As such, it offers great flexibility in dynamically implementing complex illumination patterns and varying fields of view. In addition, data of high spatial sampling can be collected by using a highly sensitive charge-coupled device (CCD) camera.

In this work, we demonstrate the application of this methodology to fluorescence protein tomography (FPT) for noninvasive three-dimensional (3D) imaging of superficial and deep-seated FP activity in small animals in vivo. We show how the proposed method can significantly improve the current state of the art of whole-body fluorescence protein imaging by (i) offering 3D volumetric imaging; (ii) enabling true quantification that is independent of depth, tissue optical properties, and heterogeneity; and (iii) reducing sensitivity to skin autofluorescence and improving contrast. Finally, the imaging potential and limitations of this technology is discussed.

Materials and Methods

FPT Scanner. The proposed method for FPT used a highly sensitive CCD camera (VersArray, Princeton Instruments, Trenton, NJ), similar to the one used for bioluminescence imaging. It offers high detection sensitivity and low noise characteristics based on –120°C cryogenic cooling and using 334 × 324 pixels after a 4 × 4 hardware pixel binning. The light source was an Argon-Ion laser (Melles Griot Laser Corp., Carlsbad, CA) delivering ≈40 mW of light power onto the animals investigated, at 488 and 514 nm, which is appropriate for imaging the green and yellow FPs. Light was delivered through a 100-μm multimode fiber (Thorlabs, Newton, NJ) to the input collimator of a custom made optical scanning device (Nutfield Technology, Windham, NH) that offers multiprojection noncontact photon delivery using a set of two galvanometer-controlled mirrors that scan a focused laser spot (300 μm diameter) through a telecentric lens onto the animal. This previously undescribed noncontact configuration enables efficient light delivery and high spatial sampling, which can significantly improve imaging performance over fiber-based systems (10). The major advantages are the elimination of light-optical fiber coupling losses common in fiber-based systems, the flexibility for a versatile scanning area and source pattern that can be dynamically adapted to a specific experiment, and the ability to operate over a wide wavelength area (for example, all visible and NIR) by choosing the appropriate optics. A detailed description of this system is given in ref. 9, and a schematic representation of the experimental setup is given in Fig. 6, which is published as supporting information on the PNAS web site.

Forward Model for Tomography in the Visible. To model photon propagation in tissues at visible wavelengths, we used a modified solution of the diffusion equation to account for both high scattering and high absorption of tissue in the visible range. The major difference with solutions derived for the NIR is the use of a diffusion coefficient with a weighted dependence on the absorption coefficient, which can be written as equation M1. The use of the modified coefficient results in a modified wavenumber equation M2 (11) for the propagating photon wave. This seemingly simple modification is the result of an efficient treatment of the radiative transfer equation (11) and recently has been found valid for modeling photon propagation in highly absorbing media (8). Importantly, this simplification or approximation can offer a practical and computationally efficient inversion scheme for tomography in the visible that resembles solutions derived for the NIR.

Animal Handling and Image Acquisition Procedures. Animal studies were performed according to procedures approved by the Massachusetts General Hospital Review Board. All animals imaged were placed on top of the custom-made optical scanner and slightly compressed to 1.2 cm. The chamber then was filled with an intralipid-ink solution with optical properties that yield similar attenuation of the mouse boundary. This matching fluid was used to simplify theoretical constraints associated with boundary modeling, although practical schemes for optical tomography without the use of fluids have been recently reported (12). Epi-illumination images then were acquired at the excitation and emission wavelengths for coregistration with the reconstructed tomographic images. Epi-illumination was implemented by illuminating the animal from the top (see Fig. 6) after beam expansion with a set of lenses over a field of view with 5-cm radius of homogeneous illumination (±5% spatial intensity variation). To spectrally separate the signals, we used three-cavity band-pass interference filters (Andover, Salem, NH) centered on 488 ± 1.5, 510 ± 5, and 570 ± 5 nm for the excitation, GFP, and yellow FP (YFP) measurements, respectively.

After planar imaging was completed, the laser light was directed from the front illuminating path to the optical scanner, and a laser spot (≈300 μm diameter) was sequentially scanned on a 33-position grid of equidistant positions at the bottom-side of the animal covering a 23 × 16-mm scan area. Each laser spot position implemented several different projections in “transillumination” through the animal, because photons do not follow straight paths but propagate along different angled directions in a diffusive pattern. This approach is analogous to tomosynthesis techniques used in x-ray imaging; here, however, photons are highly scattered, and therefore different inversion methods are used.

Two types of measurements were acquired during the experiments for each illumination point: fluorescence (emission) measurements with CCD exposure time of 30 s and intrinsic (excitation) light measurements with exposure time of 0.5 s using the same filters used for planar imaging. These measurements were used to compose the normalized Born field equation M3 given by

equation M4
[1]

The use of the Born field eliminates any position-dependent contributions and significantly minimizes the sensitivity of the reconstruction to background heterogeneities (13). At the end of the imaging acquisition the mouse was removed from the chamber, and a baseline measurement was acquired at the excitation wavelength for each of the 33 sources through the matching fluid. This data set was used to determine the location of the sources. The total acquisition time of a complete data set was ≈25 min.

FPT Image Reconstruction. To reconstruct the 3D fluorochrome distribution using the normalized born field, the volume of interest was discretized into volume elements according to the size of the field of view of each experiment (typically 23 axial, 23 sagittal, and 17 coronal layers with voxel size of 1 × 1 × 0.7 mm3). The forward model was constructed by using the described algorithm and inverted by using algebraic reconstruction techniques with positive restriction (13, 14). The inversion used a set of 25 × 25 virtual detectors equidistantly spaced over the 2.3 × 2.3-cm field of view defined by the mesh selected. The virtual detectors were calculated by binning CCD pixels together around the center that corresponds to each individual virtual detector using a radius of 0.1 cm. This binning operation is necessary to improve the signal-to-noise ratio per detector and reduce the overall size of data available. It has been shown that such operations significantly reduce the computation time without significantly limiting the information obtained and the imaging performance (15). The reconstructions were obtained within 5 min of computation on a 2 GHz Pentium 4 personal computer. Reconstructed images display the fluorochrome concentration in three dimensions, albeit with reduced resolution along the axis perpendicular to the camera imaging plane, because the geometry implemented herein allows for limited angle projections through tissue. A detailed description of the methods used for the acquisition and reconstruction of the images is in ref. 9.

Spectral Unmixing. Spectral deconvolution of GFP and YFP fluorescence signals was based on a linear unmixing algorithm (16, 17). The algorithm takes into consideration the relative strengths of the two fluorophores at the two spectral bands that detection was made and processes the raw CCD images acquired at the two different spectral bands to calculate the unmixed GFP and YFP images and then feed them into the FPT reconstruction code. The relative fluorescence protein strengths were obtained from transillumination images acquired from a mouse sequentially implanted postmortem with tubes containing equal quantities of GFP- and YFP-expressing cells so that the wavelength-dependent attenuation of the tissue is registered in the measurements. The measured values are reported in Table 1, which is published as supporting information on the PNAS web site. The uncoupled images were then calculated by solving the following linear 2 × 2 system:

equation M5

where I1 and I2 are the 334 × 324 × 33 gray scale raw CCD images obtained from the FPT scanner, sG1, sG2, sY1, and sY2 are the relative strengths of GFP and YFP at λ1 = 510 ± 5 nm and λ2 = 570 ± 5 nm, and CG and CY are the unknown unmixed images for GFP and YFP, respectively.

Postmortem Studies. To examine whether FPs could be resolved and quantified deep in animals, we performed imaging studies of animals implanted with thin (1.5–1.8 mm diameter) glass tubes containing different amounts of GFP- and YFP-expressing cells suspended in PBS. The tubes were inserted through the esophagus of euthanized animals. The cells used were previously prepared by infecting Gli36Δ5 cells (18, 19) with GFP-or YFP-expressing herpes simplex virus amplicon vectors. The numbers of cells contained in the tubes were 5 × 104, 1 × 105, 5 × 105, and 1 × 106, which corresponded to 0.5-, 1-, 5-, and 10-μl volumes, respectively. We note that blood deoxygenation, due to euthanasia, is expected to slightly increase light attenuation over the in vivo case in the wavelengths of interest (488–540 nm) assuming no significant blood drainage because animals are placed in the horizontal position.

In Vivo Studies. The in vivo studies involved three different animal models.

The first model offered GFP expressing tumors grown superficially in female nude mice (n = 4). The tumors were implanted by injecting GFP-labeled Gli36 glioma cells under the mammary fat pad and allowing them to grow into tumors for ≈20 days.

The second set of in vivo imaging experiments tested imaging of deeper-seated tumors, by implanting GFP expressing 9L rat glioma cells (20) into the right lung of female nude mice (n = 5) using direct needle injection. The injection was done by using a 0.5-ml insulin syringe from the posterior and approximately at the middle of the upper torso area and through the intercostal muscles. The mice were anesthetized before the intrapulmonary implantation. The 9L GFP tumor cells (1 × 106) were suspended in a 10-μl volume of equal amounts of DMEM (Cellgro Mediatech, Herndon, VA) and Matrigel (BD Biosciences, Bedford, MA). The mice were imaged at 10, 14, and 20 days to monitor tumor growth. X-ray computed tomography (CT) imaging was performed on a dedicated animal imager (Gamma Medica X-SPECT, Northridge, CA) after each FPT session for coregistration of anatomical and molecular images. Coregistration was obtained with a semiautomatic homemade code that requires user feedback to visually align the centers in the CT and planar fluorescence images so that the mouse outline overlaps. After imaging studies, the tumors were excised for histological examination. They were sliced with a microtome (CM1900, Leica, Bannockburn, IL), and the histological images were obtained with a fluorescence microscope (Zeiss Axiovert 100TV).

The third animal model tested the ability to simultaneously image tumor and viral delivery in vivo. For this model, 1 × 106 Gli36Δ5 stably expressing YFP cells were implanted under the mammary fat pad of female nude mice (n = 3) and allowed to grow for 20 days. The cells were established by transfection of a YFP expression plasmid (pCAG-YFP-Hyg) followed by hygromycin B selection. Subsequently, 2.5 × 105 plaque-forming units of a GFP-expressing oncolytic herpes simplex virus (Nestin34.5; H.K., E. A. Chiocca, and Y.S., unpublished data) (propagated and titered on VERO cells) were directly injected into the tumors. Imaging was performed 24 h after the virus administration.

Results

Fluorescence Quantification Postmortem. Fig. 1 depicts the postmortem study results. The mean fluorescence values reconstructed were plotted against the actual number of cells and the data were fitted with a linear regression as presented in Fig. 1 a and b for GFP and YFP, respectively. Insets represent one reconstructed coronal slice taken at the position of the maximum intensity (center of the tube) for the cases of 5 × 104 (lowest) and 106 (highest) detected cell numbers. These results collectively demonstrated the feasibility and linear response in reconstructing fluorescent cells within small animals. The fluorescent cells could not be detected in planar epi-illumination mode with this system.

Fig. 1.
Quantification accuracy for reconstructions of GFP-expressing (a) and YFP-expressing (b) tumor cells (human glioma cell line) contained in thin capillary glass tubes inserted in the esophagus of killed nude female mice and obtained postmortem. The black ...

In Vivo Imaging of Superficial Tumors. To examine the imaging characteristics of epi-illumination and transillumination/tomography, we investigated the imaging performance in a mouse with two s.c. grown GFP-expressing tumors as indicated with two arrows on the mouse photograph in Fig. 2a. Typical images obtained in epi-illumination and transillumination can be seen in Fig. 2 b and c, respectively. Here the images are scaled so that the background is visible. Epi-illumination detected brighter signals from the tumors compared with transillumination; however, it was limited by skin autofluorescence. Conversely, the transillumination images were CCD-noise limited and not autofluorescence limited. This difference is observed because of the relative photon propagation characteristics of the illumination selected. In epi-illumination, native skin fluorochromes such as elastin are excited with maximum signal intensity so autofluorescence is evident on the images. In contrast, in transillumination attenuated light reaches the surface on the CCD camera side so that minimal skin autofluorescence is excited. Deeper-seated autofluorescence is also significantly attenuated. We further found that the tumor size is also better defined in transillumination than in the epi-illumination images. Tomographic images can be seen in Fig. 2 df, where three coronal slices from the reconstruction are depicted in color and superimposed on the mouse photograph of Fig. 2a. The reconstructed images come with a color bar calibrated in number of cells and are rendered after a threshold corresponding to 50% of the maximum reconstructed value was applied. Calibration used the results of the postmortem studies where known amounts of cells were imaged. Fig. 2d corresponds to a slice deep in the body where virtually no activity was detected. Fig. 2 e and f correspond to superficial depths reconstructing the tumors with high contrast and good size definition.

Fig. 2.
Imaging of superficially grown GFP-expressing tumors. (a) A white light image of the mouse. The two s.c. tumors are shown, indicated by red arrows. (b) Image obtained with epi-illumination showing bright signals from the tumors and strong skin autofluorescence. ...

Lung Cancer Imaging. Although the above experiments confirm tomographic capacity of FP-expressing cells in vivo, the goal of this study was to examine whether in vivo detection would be feasible deeper in the body. Fig. 3 depicts results of the in vivo lung imaging studies and further renders the corresponding anatomical micro-CT images obtained with the animals under similar placement conditions. The top (ag) and bottom (hn) rows correspond to the images obtained at days 10 and 20 after injection, respectively. Fig. 3 a and h show the white light images (photographs) of the mouse obtained at corresponding days. Fig. 3 b and i show planar epi-illumination images of the mouse. In both days, the lung tumors are not evident on the epi-illumination images, but significant skin autofluorescence is observed. Conversely, the tumors are clearly visible on Fig. 3 c and j, which depict FPT coronal slices (in color) overlaid on the white light image of the mouse after a threshold at 50% of the maximum reconstructed value was applied. Fig. 3 d and k shows corresponding CT coronal slices, anatomically resolving the tumors. The corresponding axial CT and FPT slices are shown respectively, in Fig. 3 e and f for day 10 and Fig. 3 l and m for day 20. Finally Fig. 3 g and n depicts 3D renderings of the CT data, which illustrate the plane from which axial images have been acquired. Histological correlation can be seen in Fig. 7, which is published as supporting information on the PNAS web site. Three-dimensional fused FPT–x-ray CT renderings can be seen in Fig. 8, which is published as supporting information on the PNAS web site.

Fig. 3.
FPT of lung cancer at 10 and 20 days of tumor growth and corresponding x-ray CT images. (ag) Images obtained from the first imaging session 10 days after the injection of the cancer cells. (hn) Corresponding images obtained at the 20th ...

Generally, there was good congruence observed between the location seen on FPT and CT images. FPT was able to record tumor growth for the different imaging sessions that was confirmed by x-ray CT. Plots of the tumor growth assessed by CT and FPT are shown in Fig. 4 as a function of time (days) with good agreement between the two methods. The error bar in the plots represents the uncertainty in each independent volume measurement assuming different thresholds for determining the extent of the volume reconstructed. Some artifacts appear on the FPT image. There is some activity recorded close to the image boundaries, which is a typical effect of noise in the data. In addition, the tumors reconstructed on FPT appear elongated compared with the CT images. This result is an effect of the limited projection tomography that was used here because of the slab geometry implemented. Similarly to x-ray tomosynthesis, this approach offers straightforward tomographic implementation but has compromised resolution along the axis perpendicular to the source and detector planes. Complete projection tomographic schemes recently developed for optical tomography (21) can offer more symmetric resolution along all axes.

Fig. 4.
Tumor growth observed by FPT (a) and x-ray CT (b) studies. The FPT volume calculations were obtained by integrating all voxels above a threshold of twice the mean of the reconstructed values at a volumetric region of interest around the reconstructed ...

Viral Gene Delivery. To further assess the ability to concurrently record multiple FPs, we performed multispectral tomography to visualize a GFP-expressing oncolytic herpes simplex virus mutant propagating in YFP-labeled tumoral cells (2224) similar to studies previously reported for microscopy (16, 17) or epi-illumination imaging (25). Fig. 5 a and b shows reconstructed coronal slices of the original spectrally mixed images corresponding to detection at the YFP and GFP channels, respectively. Fig. 5 c and d shows coronal slices from the reconstructed unmixed images of GFP and YFP, respectively, and Fig. 5e depicts an overlay of the two unmixed images.

Fig. 5.
Visualization of tumor- and virus-mediated gene delivery. (a and b) Reconstructed coronal slices corresponding to the unmixed signals at the GFP and YFP channels, respectively. (c and d) Spectrally deconvolved reconstructed slices representing the GFP-expressing ...

Discussion

In this work, we investigated the feasibility of performing noninvasive tomographic imaging of FPs in mice in vivo. Of particular significance was the use of (i) transillumination as raw data for tomography and (ii) appropriate mathematical forward models that go beyond the standard diffusion approximation developed for the NIR. Transillumination offers better symmetry in probing the animal volume compared with epi-illumination methods and was here the preferred mode for tomographic imaging. Combined with inversion methods, these data can yield 3D description of fluorescence biodistribution. Imaging or tomography in epi-illumination mode may be more sensitive in detecting superficial activity, but detection becomes more challenging for deep-seated activity, and image fidelity worsens over transillumination schemes (26).

A significant advantage of using reconstruction techniques even when imaging s.c. tumors is that these methods can correct or are insensitive to the background optical properties and tissue optical heterogeneity (27), offering more accurate quantification and fidelity than photographic methods at any depth (24). This multispectral tomographic approach could be expanded to more wavelengths for simultaneous monitoring of an increased number of biological processes, each marked by distinct fluorochromes or proteins. It was further observed that this strategy can minimize the effects of tissue autofluorescence especially when narrow-band filters are used. It is characteristic that the detection sensitivity in transillumination measurements, such as the ones shown in Fig. 2, were CCD noise-limited and not tissue autofluorescence-limited as common in epi-illumination methods. Spectral schemes developed for reducing the sensitivity to autofluorescence and improve detection in epi-illumination mode also could be propagated in transillumination and tomography (25).

The overall detection sensitivity of the method is ultimately determined by FP emission strength and tissue attenuation. In this study, we were able to detect ≈50 × 103 cells from the middle (≈6 mm deep from either side) of small mice postmortem. This sensitivity measure strongly depends on the depth and tissue optical properties and allows for general guidelines in terms of in vivo detection limits when imaging activity at the center of animals. Imaging cells closer to the surface could yield more sensitive detection (4). Similarly, the resolution of 400 μm previously observed for this set-up (9) at the center of a 1.2-cm slab is also depth- and optical-property-dependent, so that this value is a generic measure outlining a worst-case scenario. Generally, imaging more superficially improves the resolution. Conversely moving to lesser attenuation (as in the NIR) allows for more broad diffusive propagation and worsens the resolution.

With an increasing availability of potent red-shifted FPs, the sensitivity is expected to improve significantly to allow more sensitive detection. Currently, however, even the most red-shifted fluorochromes require excitation in the visible (<600 nm) (2832), where tissue is more absorbing than even in the green (33). Therefore, adept mathematical models for photon propagation in highly absorbing media and efficient photon delivery and detection methods are generally required for tomography. It is useful to observe, however, that in contrast to systemic administration of fluorescent probes, imaging of FPs may yield better contrast because of the absence of nonspecific background signals.

The technology presented here may offer a more accurate alternative to photographic imaging methods for resolving deep-seated and superficial activity and could find several applications in imaging genetically engineered animals. It can be further used to interrogate tumor metastasis and drug action, stem cell migration, and immunological responses. Imaging in the visible will restrict applications to small animal imaging because of the strong light attenuation in this spectral region; however, development of red-shifted FPs (2832) could further enable imaging of larger animals as well and increase the detection sensitivity.

Supplementary Material

Supporting Information:

Acknowledgments

We thank Kazue Kasai (Massachusetts General Hospital, Harvard Medical School, Charlestown, MA) for preparation of herpes simplex virus amplicon vectors and Kinya Terada and Ken Ishii for their input on the initial phase of the project. G.Z. thanks Antoine Soubret, Gordon Turner, and George Themelis for useful discussions. This work was supported in part by National Institutes of Health Grants R01 EB 000750-1, R33 CA 91807, P50 CA 86355, and R24 CA 92782. J.R. was supported by European Union Integrated Project “Molecular Imaging” LSHG-CT-2003-503259 and European UnionSTREP “TRANS-REG” LSHG-CT-2004–502950.

Notes

Author contributions: Y.S. and V.N. designed research; G.Z., H.K., H.S., J.R., J.G., and Y.S. performed research; G.Z. and J.R. analyzed data; and G.Z., R.W., and V.N. wrote the paper.

Conflict of interest statement: R.W. is a founding member of VisEn Medical, a molecular imaging company. V.N. has financial interests in VisEn Medical.

This paper was submitted directly (Track II) to the PNAS office.

Abbreviations: CCD, charge-coupled device; CT, computed tomography; FP, fluorescent protein; FPT, FP tomography; NIR, near-IR; YFP, yellow FP.

References

1. Weissleder, R. & Ntziachristos, V. (2003) Nat. Med. 9, 123–128. [PubMed]
2. Wouters, F. R., Verveer, P. J. & Bastiaens, P. I. H. (2001) Trends Cell Biol. 11, 203–211. [PubMed]
3. Brown, E. B., Campbell, R. B., Tsuzuki, Y., Xu, L., Carmeliet, P., Fukumura, D. & Jain, R. K. (2001) Nat. Med. 7, 864–868. [PubMed]
4. Hoffman, R. M. (2002) Lancet Oncol. 3, 546–556. [PubMed]
5. Ntziachristos, V., Tung, C., Bremer, C. & Weissleder, R. (2002) Nat. Med. 8, 757–760. [PubMed]
6. Ntziachristos, V. & Weissleder, R. (2002) Med. Phys. 29, 803–809. [PubMed]
7. Tsien, R. Y. (2005) FEBS Lett. 579, 927–932. [PubMed]
8. Ripoll, J., Yessayan, D., Zacharakis, G. & Ntziachristos, V. (2005) J. Opt. Soc. Am. A 22, 546–551. [PubMed]
9. Zacharakis, G., Ripoll, J., Weissleder, R. & Ntziachristos, V. (2005) IEEE Trans. Med. Imaging 24, 878–885. [PubMed]
10. Graves, E. E., Ripoll, J. R. W. & Ntziachristos, V. (2003) Med. Phys. 30, 901–911. [PubMed]
11. Aronson, R. & Corngold, N. (1999) J. Opt. Soc. Am. A 16, 1066–1071. [PubMed]
12. Schulz, R. B., Ripoll, J. & Ntziachristos, V. (2004) IEEE Trans. Med. Imaging 23, 492–500. [PubMed]
13. Ntziachristos, V. & Weissleder, R. (2001) Opt. Lett. 26, 893–895. [PubMed]
14. Kak, C. & Slaney, M. (1988) Principles of Computerized Tomographic Imaging (IEEE, New York).
15. Graves, E. E., Culver, J. P., Ripoll, J., Weissleder, R. & Ntziachristos, V. (2004) J. Opt. Soc. Am. A 21, 231–241. [PubMed]
16. Zimmermann, T., Rietdorf, J. & Pepperkok, R. (2003) FEBS Lett. 546, 87–92. [PubMed]
17. Papadakis, A., Stathopoulos, E., Delides, G., Berberides, K., Nikiforidis, G. & Balas, C. (2003) IEEE Trans. Biomed. Eng. 50, 207–217. [PubMed]
18. Ichikawa, T., Hoegemann, D., Saeki, Y., Tyminski, E., Terada, K., Weissleder, R., Chiocca, E. A. & Basilion, J. P. (2002) Neoplasia 4, 523–530. [PMC free article] [PubMed]
19. Saeki, Y., Fraefel, C., Ichikawa, T., Breakefield, X. O. & Chiocca, E. A. (2001) Mol. Ther. 3, 591–601. [PubMed]
20. Moore, A., Marecos, E., Simonova, M., Weissleder, R. & Bogdanov, A. J. (1998) Microvasc. Res. 56, 145–153. [PubMed]
21. Turner, G. M., Zacharakis, G., Soubret, A., Ripoll, J. & Ntziachristos, V. (2005) Opt. Lett. 30, 409–411. [PubMed]
22. Yang, M., Baranov, E., Jiang, P., Sun, F., Li, X., Li, L., Hasegawa, S., Bouvet, M., Al-Tuwaijri, M., Chishima, T., et al. (2000) Proc. Natl. Acad. Sci. USA 97, 1206–1211. [PMC free article] [PubMed]
23. Yang, M., Li, L., Jiang, P., Moossa, A. R., Penman, S. & Hoffman, R. M. (2003) Proc. Natl. Acad. Sci. USA 100, 14259–14262. [PMC free article] [PubMed]
24. Pfeifer, A., Kessler, T., Yang, M., Baranov, E., Kootstra, N., Cheresh, D. A., Hoffman, R. M. & Verma, I. M. (2001) Mol. Ther. 3, 319–322. [PubMed]
25. Gao, X., Cui, Y., Levenson, R. M., Chung, L. W. K. & Nie, S. (2004) Nat. Biotechnol. 22, 969–976. [PubMed]
26. Pogue, B. W., McBride, T. O., Osterberg, U. L. & Paulsen, K. D. (1999) Opt. Express. 4, 270–286. [PubMed]
27. Soubret, A., Ripoll, J. & Ntziachristos, V. (2005) IEEE Trans. Med. Imaging 24, 1377–1386. [PubMed]
28. Shaner, N. C., Campbell, R. E., Steinbach, P. A., Giepmans, B. N. G., Palmer, A. E. & Tsien, R. Y. (2004) Nat. Biotechnol. 22, 1567–1572. [PubMed]
29. Wang, L., Jackson, W. C., Steinbach, P. A. & Tsien, R. Y. (2004) Proc. Natl. Acad. Sci. USA 101, 16745–16749. [PMC free article] [PubMed]
30. Verkhusha, V. V. & Lukyanov, K. A. (2004) Nat. Biotechnol. 22, 289–296. [PubMed]
31. Fradkov, A. F., Chen, Y., Ding, L., Barsova, E. V., Matz, M. V. & Lukyanov, S. A. (2000) FEBS Lett. 479, 127–130. [PubMed]
32. Matz, M. V., Fradkov, A. F., Labas, Y. A., Savitsky, A. P., Zaraisky, A. G., Markelov, M. L. & Lukyanov, S. A. (1999) Nat. Biotechnol. 17, 969–973. [PubMed]
33. Ntziachristos, V., Ripoll, J., Wang, L. & Weissleder, R. (2005) Nat. Biotechnol. 23, 313–320. [PubMed]

Articles from Proceedings of the National Academy of Sciences of the United States of America are provided here courtesy of National Academy of Sciences
PubReader format: click here to try

Formats:

Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...

Links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...