Imaging Depths of Near-Infrared Quantum Dots in ... - SAGE Journals

12 downloads 0 Views 675KB Size Report
be possible to improve the signal to noise ratios by over. 100-fold by QDs that emit at .... immersed in biologic tissues at a depth z below the air/ tissue interface.
RESEARCH ARTICLE

Imaging Depths of Near-Infrared Quantum Dots in First and Second Optical Windows Nayoun Won, Sanghwa Jeong, Kangwook Kim, Jungheon Kwag, Joonhyuck Park, Sang Geol Kim, and Sungjee Kim

Abstract Potential advantages of quantum dot (QD) imaging in the second optical window (SOW) at 1,000 to 1,400 nm over the first optical window (FOW) at 700 to 900 nm have attracted much interest. QDs that emit at 800 nm (800QDs) and QDs that emit at 1,300 nm (1,300QDs) are used to investigate the imaging depths at the FOW and SOW. QD images in biologic tissues are processed binarized via global thresholding method, and the imaging depths are determined using the criteria of contrast to noise ratio and relative apparent size. Owing to the reduced scattering in the SOW, imaging depth in skin can be extended by approximately three times for 1,300QD/SOW over 800QD/FOW. In liver, excitation of 1,300QD/SOW can be shifted to longer wavelengths; thus, the imaging depth can be extended by 1.4 times. Effects of quantum yield (QY), concentration, incidence angle, polarization, and fluence rate F on imaging depth are comprehensively studied. Under F approved by the Food and Drug Administration, 1,300QDs with 50% QY can reach imaging depths of 29.7 mm in liver and 17.5 mm in skin. A time-gated excitation using 1,000 times higher F pulses can obtain the imaging depth of < 5 cm. To validate our estimates, in vivo whole-body imaging experiments are performed using small-animal models.

EAR-INFRARED (NIR) IN VIVO IMAGING with semiconductor quantum dots (QDs) can offer nonradioactive extraction of biologic information, promising highly multiplexed molecular imaging. NIR imaging using QDs exploits their large-absorption cross sections, high photostability, and relatively narrow and symmetric emission peaks, which can be tuned from visible to infrared wavelengths l.1–3 QDs are especially advantageous for the bright infrared emission over molecular fluorophores because they can be free from the molecular vibration–coupled nonradiative channels. As a result, QDs can be an optimal imaging contrast agent in the NIR optical window (700 # l # 1,400 nm), where the interference by the absorption and scattering from water and biologic tissues becomes minimal.4–6 The NIR optical window can be divided into the first optical window (FOW; 700 # l # 900 nm) and the second optical

N

From the Department of Chemistry and School of Interdisciplinary Bioscience and Bioengineering, Pohang University of Science and Technology, Gyeongbuk; and the Department of Surgery, College of Medicine, Kyungpook National University, Daegu, Republic of Korea. Address reprint requests to: Sungjee Kim, PhD, Department of Chemistry, Pohang University of Science and Technology; e-mail: [email protected].

DOI 10.2310/7290.2011.00057 #

2012 Decker Publishing

338

window (SOW; 1,000 # l # 1,400 nm).7 The SOW is more advantageous for imaging than the FOW because of the lower scattering level. In 2003, simple simulation studies of optical imaging in turbid media such as tissue or blood showed that it would be possible to improve the signal to noise ratios by over 100-fold by QDs that emit at the SOW when compared to those that emit at the FOW.6 QDs have been demonstrated for many in vivo imaging applications, including sentinel lymph node mapping,8–11 tumor neovascularization,12,13 visualization of lymphatic vessels,14 and intraoperative study of tumor delineation.15 Recently, QDs for the SOW have shown their potential for target-specific labeling of cells using lead selenide (PbSe) QDs.16 Carbon nanotubes have also been used for probes in the SOW; however, they are not as bright as QDs.17–19 Design considerations of QDs in terms of the toxicity issues for potential in vivo imaging are being actively developed. Conventional heavy metal elements for QDs have been replaced by ‘‘safer’’ elements.9,20–22 The effect of the hydrodynamic size of QDs has been studied to control the biodistribution and to facilitate the renal clearance.23–25 However, the imaging capability of QDs in the FOW and SOW has not been comprehensively studied because it is affected by a large number of parameters, including polarization and the incidence angle of propagating

Molecular Imaging, Vol 11, No 4 (July–August 2012): pp 338–352

Imaging Depths of Near-Infrared Quantum Dots in First and Second Optical Windows

photons, concentration of QDs, fluence rate F, and QD characteristics, including quantum yield (QY) and absorption cross section. In addition, different detectors are used to measure the emission in the FOW (typically siliconintensified [Si] charge-coupled device [CCD] cameras) and in the SOW (typically Indium gallium arsenide (InGaAs) or Mercury cadmium tellurideHg (CdTe) CCD cameras). Different characteristics of these detectors complicate quantitative comparison of the results. Herein we report quantitative and comparative studies of NIR QD imaging capability in the FOW and SOW based on a series of reflectance fluorescence imaging experiments using QDs that emit at around 800 nm (800QDs) and QDs that emit at around 1,300 nm (1,300QDs) in various phantoms and biologic tissues of different thicknesses.

Materials and Methods Materials Cadmium acetate dihydrate (99.999%) was purchased from Alfa Aesar (Ward Hill, MA). Selenium shot (99.99%, < 2 mm), tellurium shot (99.999%, 1–2 mm), lead(II) acetate trihydrate (99.999%), trioctylphosphine (TOP; 90%), trioctylphosphine oxide (TOPO; 90%), stearic acid (95%), oleic acid (90%), diphenyl ether (99%), hemoglobin from bovine blood, and intralipid (20% emulsion) were purchased from Sigma-Aldrich (St. Louis, MO).

339

synthesized by following the cadmium-limited condition method.27 The selenium precursor was previously prepared by dissolving selenium shots (0.3 mmol) in TOP (0.3 mL) in the glove box. The tellurium precursor was previously prepared by dissolving tellurium shots (0.2 mmol) in TOP (0.4 mL) in the glove box. The selenium precursor was mixed with tellurium precursor, and additional TOP (2 mL) was added. Cadmium acetate dehydrate (1 mmol), stearic acid (20 mmol), and TOPO (5 g) were loaded into a three-neck flask and heated to 270 uC under nitrogen gas flow. At this temperature, the mixture of selenium and tellurium precursors was quickly injected into the reaction flask, and the temperature was maintained at 250 uC. The reaction mixture was stirred until the desired size of CdTeSe nanocrystals was obtained. On completion, the mixture was cooled to room temperature and diluted with hexanes. PbSe QDs were prepared by the following previously reported methods.28 The selenium precursor was previously prepared by dissolving selenium shots (5 mmol) in TOP (5 mL) in the glove box. Lead(II) acetate trihydrate (2 mmol), oleic acid (4 mmol), and 10 mL of diphenyl ether were loaded into a three-neck flask and heated to 120uC under nitrogen gas flow. At this temperature, the selenium precursor was quickly injected into the reaction flask, and the temperature was maintained at 100uC. The reaction mixture was stirred until the desired size of PbSe nanocrystals was obtained. On completion, the mixture was cooled to room temperature and diluted with hexanes.

Image Data Process by Global Thresholding The global thresholding process was conducted using the MATLAB (The MatWorks, Inc.) program based on the global thresholding algorithm.26 The raw intensity data of NIR fluorescence images were imported to MATLAB. The initial threshold (T) value was estimated using the mean of maximum intensity and minimum intensity of the entire pixel intensities. The histogram of the image was divided by two regions using the T value, and the mean intensities, m1 and m2, of two regions were calculated. A new threshold was created by the mean of m1 and m2. These processes were iterated until the T value converged. The pixels with the same or higher intensity than the T were regarded to signal. And the pixels with the lower intensity than the T were regarded to background. Synthesis of QDs Preparation of Cadmium telluride selenide (CdTeSe) alloyed QDs with homogeneous internal structures was

Imaging Setup and Configurations For reflectance fluorescence imaging, the light source is typically uniform and diffuse and has an incidence angle h with respect to the detector that is perpendicular to the air/ tissue interface. We used a glass container of 1 cm lateral dimension filled with a QD solution. The container was immersed in biologic tissues at a depth z below the air/ tissue interface. An Si CCD camera (Hamamatsu, ORCAAG) was used for the FOW imaging. The Si CCD camera was attached with 780 nm long-pass filter (Edmund, 32757) and with a zoom lens (Optem, NIR Zoom 70XL). An InGaAs CCD camera (FLIR, SC2500-NIR) was used for the SOW imaging. The InGaAs CCD camera was attached with a 1,250 nm long-pass filter (Thorlabs, FEL1250) and with a zoom lens (Navitar, Zoom 6000 NIR). The working distance between samples to the zoom lens was about 40 cm, and the field of view was 4.5 3 5 cm. For the experiment using biologic tissues, fresh bovine liver tissues

340

Won et al

and porcine skin tissues were obtained from a local butcher’s shop and were immediately frozen. Before the experiments, they were sectioned and the thickness was measured by caliper. In Vivo Imaging Experiment A female nude mouse (BALB/c-nu/nu, 8 weeks old, 25 g, from Orient Co. Ltd, Korea) was used in accordance with our institution’s guidelines on animal care and use. The mouse was anesthetized with 0.015 mL g–1 of intraperitoneal tribromoethanol (Avertin). A cylindrical glass container of 5 mm width was filled with 0.3 mL of a solution of QD in hexanes. The solution contained both 800QDs and 1,300QDs at the concentrations adjusted to have identical absorption at excitation wavelength lex. The ends of the QD cylinder were capped by rubber septa. The cylinder was inserted into the abdominal cavity of a mouse below the intestine, and the skin incision was closed using a suture. The Si CCD camera was used for the imaging of 800QDs in the FOW, and the InGaAs CCD camera was used for the imaging of 1,300QDs in the SOW. Identical excitation conditions (F 5 20 mW cm–2) with an excitation wavelength of 660 nm were used for both cases. The images were obtained within 30 minutes. Part of the entire image that contains the mouse body, not the background, is used to calculate contrast to noise ratios (CNRs).

Results and Discussion NIR photons are known to penetrate < 10 cm in biologic tissues.4 However, photon propagation length is little meaningful for practical imaging applications. The ability to distinguish an object from the background can be evaluated by the CNR, which is the difference between the signal intensity and the background intensity divided by the standard deviation of the background intensity.29 The Rose criterion states that to be detectable, an object’s CNR must exceed 430–32; however, the CNR criterion alone is not sufficient to evaluate the quality of image. Another criterion is necessary to judge the spatial resolution of image. Spatial resolution of an optical image can be majorly affected by the scattering characteristics of the medium and the thermal noise level of the CCD camera. Spatial resolution in medical imaging is conventionally defined as the ability to resolve details in the object and can be described by the line spread function or the modulation transfer function.33,34 To determine the resolution, test

phantoms with stripes of alternating bright and dark contrast with different frequencies are typically used.35 When the lateral blurring of the stripe line approaches the width of the line (in other words, when each line blurs to appear as double of the actual width), the imaging modality fails to resolve the stripes. From this simplified concept, we imposed another criterion for image quality. The relative apparent size (RAS) should be less than 200%, where the RAS here is defined by the apparent lateral width of the QD sample divided by the actual width. In our condition, a QDimaged object must meet both the CNR and the RAS criteria to be considered properly imaged. The imaging depths of NIR QD reported here should be regarded as minimum depths that can be achieved without the aid of sophisticated data processing methods. To determine the CNR and the RAS, we adopted a simple image data process of segmentation by global thresholding. The raw reflectance fluorescence image was binarized into signal area or background area by the global threshold (Figure 1A). The global threshold that minimizes the variance between the intensity values of the signal area and between those of the background area was determined iteratively (see Materials and Methods, above).26 The binarized image was used to obtain the CNR and RAS values and to determine the imaging depth. Other than Figure 1A, all QD images presented are raw and as taken by the CCD camera. Any bright spots owing to reflections or detector thermal noise may be segmented into signal, especially when the CNR is low. Images with such falsepositive spots may pass the CNR criterion, but the spots significantly increase RAS and may fail the RAS criterion. Thus, the CNR and RAS criteria can complement each other. Based on these criteria, we compared the NIR QD imaging capability in the FOW and SOW as focused on the imaging depths in different environments. We report the imaging depths by the CNR and the RAS criteria, respectively. The depth limit that meets both criteria should provide the minimum guaranteed imaging depth; however, the actual imaging depth limit may lie somewhere between the limits of the CNR and RAS depending on the nature of the object and the medium. We performed reflectance fluorescence imaging (rather than back-illuminating configuration) to obtain information for potential intraoperative applications (Figure 1B). Because QDs will likely be used in the future for tumor targeting–based diagnosis, and specifically for detection of small collections of malignant cells, we simulated them using a glass container of 1 cm lateral dimension filled with

Imaging Depths of Near-Infrared Quantum Dots in First and Second Optical Windows

341

Figure 1. A, Near-infrared (NIR) fluorescence image of a quantum dot (QD) sample surrounded in porcine skin tissue and its binarized image. B, Schematic diagram of the experimental setup for the NIR fluorescence imaging of a QD sample in biologic tissue.

a QD solution.36,37 The container was immersed in biologic tissues at a depth z below the air/tissue interface. CdTeSe 800QDs and a Si CCD camera (Hamamatsu, ORCA-AG) were used for FOW imaging; PbSe 1,300QDs and an InGaAs CCD camera (FLIR, SC2500-NIR) were used for the SOW imaging. The QDs were prepared using previously reported methods (see Materials and Methods and Figure S2).27,28 Given that two different CCD cameras are used for the FOW and the SOW, the different thermal noise levels cannot be easily normalized for comparison. Instead of the thermal noise level normalization, we adopted the criteria of CNR and RAS for the image quality comparison. The thermal noise levels from different CCD cameras can be reflected by the CNR and RAS criteria values. QD imaging capabilities were investigated at different h using different polarizing filter configurations. For the FOW, 800QDs were used in bovine liver or porcine skin tissues of different thicknesses. Liver and skin tissues represent the extremes of tissue characteristics from highly absorbing to highly scattering. A 150 W halogen lamp with a 650 nm short-pass filter was set to have h 5 10u, 30u, or 50u. A linear polarizing filter was attached at the end of the excitation lamp with the polarization direction parallel to the air/tissue interface. F was maintained at 5 mW cm–2 regardless of h. A 780 nm long-pass filter with a linear polarizing filter was attached to the detector with the polarization direction either parallel or perpendicular to that of excitation polarization. Because the polarization of QD emission is isotropic, the polarizing filters were employed mostly to minimize interference from direct reflections at the air/tissue interface. Fluorescence images, CNRs, and RASs (Figure 2, A–C) were obtained from 1 mm thick bovine liver tissue while at h 5 10u, 30u, or 50u. Fluorescence images (see Figure 2A)

show many bright spots, which originated from the reflection at the rough air/liver interfaces. The reflection was greatly reduced by the crossed polarizing filter configuration, whereas the parallel polarizing filter configuration showed interference from the reflection spots which is similar with the case of no polarizing filter configuration. Consequently, CNRs were larger for the crossed polarizing filter configuration than for the parallel or no-filter configurations (see Figure 2B). The RASs were smaller for crossed polarizing filter configuration because the filters effectively removed the reflection spots (see Figure 2C); h has a greater effect on the CNR and RAS than does the polarizing filter configuration. The number of reflection spots decreased significantly as h increased (see Figure 2A). Because the bright spots are caused by rough surface sites that can directly reflect the excitation light to the detector, the surface reflectivity should decrease as h increases. However, this effect began to saturate at h . 30u. The CNRs and RASs at h 5 50u were marginally better than at h 5 30u. Experiments were repeated using thicker (3 mm) bovine liver tissue (Figure 2, D–F). Polarizing filter configuration showed effects similar to those obtained for the 1 mm tissue, but the largest CNRs and the smallest RASs were achieved at h 5 30u than at 50u. Reflections were minimal at h 5 50u; however, the background was imaged with uniformly higher intensities when h 5 50u than those for h 5 10u and 30u. This may be due to multiply scattered photons: because forward-directed scattering is dominant in biologic tissues, an incidence angle geometry that allows photons to reach the detector after multiple scatterings with higher probability should result in higher background noise and consequently smaller CNRs and larger RASs.38 Multiply scattered photons become more isotropic in

342

Won et al

Figure 2. Effects of imaging geometry (h 5 10u, 30u, 50u) and polarizing filter (PF) configuration. Near-infrared fluorescence images (left in A, D, G, J), contrast to noise ratios (CNRs) (middle in B, E, H, K), and relative apparent sizes (RASs) (right in C, F, I, L) for the cases in 1 mm thick bovine liver tissue (first row, A–C), in 3 mm thick bovine liver tissue (second row, D-F), in 2 mm thick porcine skin tissue (third row, G-I), and in 3 mm thick porcine skin tissue (last row, J-L). Polarizing filters were tested with no filter and with filter polarization oriented parallel or crossed to quantum dot emission polarization.

polarization and can account for the background noise that cannot be easily reduced by the polarizing filters.39 This multiple scattering effect becomes predominant when the tissue becomes sufficiently thick and thus switches the

optimal h from 50u to 30u as the liver tissue thickness increases from 1 to 3 mm. Experiments were repeated using 2 mm or 3 mm thick porcine skin tissue (Figure 2, G–L). Bright reflection spots

Imaging Depths of Near-Infrared Quantum Dots in First and Second Optical Windows

were not observed for the air/skin surface (Figure 2, G and J). Liver tissue is glossy, and highly specular reflections are predominant, whereas skin tissue reflects more diffusely, so the images portrayed pores and wrinkles on the skin surface. The skin reflections can be effectively removed by crossed polarizing filter configuration. The effect of h in skin tissue was similar to that observed in the liver tissue, and h 5 30u incidence angle provided better images with the largest CNRs and the smallest RASs than h 5 10u and h 5 50u. h 5 30u became more optimal as skin thickness increased (Figure 2, K and L). The optimal imaging configuration of h 5 30u was used for the rest of the reflectance fluorescence imaging experiments reported herein. Liver and skin tissues are representative of highly absorbing and highly scattering environments and more relevant to actual imaging applications. However, biologic tissues, including liver, are inherently scattering, and the effect of scattering cannot be rigorously decoupled from the absorption. Therefore, liquid phantom experiments were performed for comparative studies on NIR QD imaging capabilities as simulating purely absorbing or purely scattering conditions. The liquid phantoms were prepared by dispersing hemoglobin or intralipid in water. The imaging properties of 800QDs in the FOW (800QD/ FOW) and 1,300QDs in the SOW (1,300QD/SOW) were compared while varying the phantom concentrations. An LED excitation lamp (excitation wavelength lex 5 660 nm) was used at F 5 10 mW cm–2. The concentrations of 800QD and 1,300QD solutions were adjusted to match the absorbance at lex (A 5 3.5). We believe that normalizing absorbance A at lex is more meaningful for comparison than matching the molar concentrations and is more universal because molar concentration is dependent on the QD size and the material choice. Both 800QDs and 1,300QDs had a QY of < 5%. A 2 mL cylindrical vial that contained a solution of QDs in hexanes was immersed in the liquid phantom to a depth of 5 mm. Fluorescence images (Figure 3A) using 800QD/FOW and 1,300QD/SOW were captured while increasing the hemoglobin concentration from 0 to 2%, which approaches the hemoglobin content in blood. As the hemoglobin concentration increased, the CNR decreased for both 800QD/FOW and 1,300QD/SOW (Figure 3B). CNRs were higher for 800QD/FOW than in 1,300QD/ SOW because the water absorption of NIR light is higher in the SOW than in the FOW. However, the effect became less significant as the hemoglobin concentration–increased. RASs did not show noticeable difference between 800QD/ FOW and 1,300QD/SOW (Figure 3C). This shows that

343

1,300QD/SOW does not have much advantage over 800QD/FOW in purely absorbing environments. Liquid phantom experiments were repeated using intralipid solutions (Figure 3, D–F). The intralipid concentration was varied up to 2%; this concentration makes the phantom scatter NIR light more than does the skin. Images obtained using 800QD/FOW became increasingly blurred as the intralipid concentration increased, but those obtained using 1,300QD/SOW were significantly less sensitive to the intralipid concentration than 800QD/FOW (see Figure 3D). For both 800QD/FOW and 1,300QD/ SOW, CNRs dropped significantly as intralipid concentration increased to 0.25% and then stayed constant at higher concentrations (see Figure 3E). Contrary to the hemoglobin phantom cases, 1,300QD/SOW showed larger CNRs than did 800QD/FOW despite the water absorption. This occurs because 800QD/FOW emission scatters more than that from 1,300QD/SOW and thus generates higher background intensity. In the case of 800QD/FOW, RAS increased dramatically even at low intralipid contents: at 0.25% intralipid concentration, RAS approached 200%. In contrast, 1,300QD/SOW maintained RAS close to 100% even at intralipid concentration of 2% (see Figure 3F). These results demonstrate that 1,300QD/SOW imaging has a strong advantage over 800QD/FOW imaging even in slightly scattering media. The liquid phantom experiments also corroborate the validity of our criteria for imaging depth capability using both CNR and RAS. The CNR criterion works properly for absorbing media, and RAS does so for scattering media. Because most biologic tissues are significantly scattering, 1,300QD/SOW is expected to have better imaging capability than 800QD/FOW. Convinced by the possibility of 1,300QD/SOW for deep tissue–penetrating imaging in biologic tissues, we compared the imaging capabilities of 800QD/FOW and 1,300QD/SOW while varying the thickness of bovine liver tissue or porcine skin tissue. QD samples and imaging configurations were the same as used for the liquid phantom experiments except that the samples were biologic tissues instead of liquid phantoms. As the thickness of bovine liver tissue increased, both 800QD/ FOW and 1,300QD/SOW images began to show interference from the background (Figure 4A). The 800QD/ FOW images began to show reflection spots as the tissue thickness exceeded 6 mm, indicating that the Si CCD camera loses the ability to discriminate the QD signal from the reflection spots as the QD signal intensity is diminished owing to the increased tissue thickness. In 1,300QD/SOW images, the background interference was mostly due to thermal noise in the InGaAs CCD camera

344

Won et al

Figure 3. Imaging capabilities of 800 nm emitting quantum dots (solid lines with &) and 1,300 nm emitting quantum dots (QD) (dashed lines with #) in liquid phantoms. Near-infrared fluorescence images (A, D), contrast to noise ratios (CNR) (B, E), and relative apparent sizes (RAS) (C, F) for hemoglobin phantom (A–C) and for intralipid phantom (D–F). Dotted lines represent imaging depth criteria. A 660 nm LED lamp was used for excitation at the fluence rate of 10 mW cm–2. The depth of the liquid phantom was 5 mm.

and so was more uniformly distributed than in 800QD/ FOW images. CNRs decreased exponentially as the tissue thickness increased in both 800QD/FOW and 1,300QD/ SOW images (Figure 4B). In nontransparent media, the measured light intensity I decreases exponentially as tissue thickness increases as I 5 I0 exp(2z/d), where I0 is the intensity of incident light, z is the thickness of the medium, and d is the mean penetration depth for the medium.40–42 Measured values of d were 8.7 mm for 800QD/FOW and 6.0 mm for 1,300QD/SOW. SOW imaging showed slightly smaller mean penetration depth than that of FOW imaging. By extrapolating the measured CNRs using the

equation, imaging depths by the CNR criterion were obtained as 13.0 mm for 800QD/FOW and 10.0 mm for 1,300QD/SOW. RASs of 800QD/FOW images increased more rapidly than those of 1,300QD/SOW images (Figure 4C). The imaging depths by the RAS criterion were 11.2 mm for 800QD/FOW and 12.3 mm for 1,300QD/SOW. By applying both CNR and RAS criteria, the imaging depths were 11.2 mm for 800QD/FOW and 10.0 mm for 1,300QD/SOW. For liver tissue, we obtained similar imaging depths for 800QD/FOW imaging and 1,300QD/SOW imaging. Similar to the results of hemoglobin liquid phantom experiments, the imaging depths of

Imaging Depths of Near-Infrared Quantum Dots in First and Second Optical Windows

345

Figure 4. Imaging capabilities of 800 nm emitting quantum dots (solid lines with &) and 1,300 nm emitting quantum dots (QD) (dashed lines with #) in biologic tissues. Near-infrared fluorescence images (A, D), contrast to noise ratios (CNR) (B, E), and relative apparent sizes (RAS) (C, F) in the cases of bovine liver tissues (A–C) and in the cases of porcine skin tissues (D–F). Dotted lines represent imaging depth criteria. A 660 nm LED lamp was used for excitation at the fluence rate of 10 mW cm–2. The thickness of the tissues was varied from 0 to 10 mm.

800QD/FOW and 1,300QD/SOW are not noticeably different in liver tissue, which is highly absorbing. The experiments were repeated using porcine skin tissues of different thicknesses. As the thickness of the skin tissue increased, 800QD/FOW images began to show severe blurring when compared to 1,300QD/SOW images. Bright spots often occurred in 800QD/FOW images of thick skin tissue (Figure 4D). Identical skin epidermal surface was used for all experiments, including the 800QD/ FOW and 1,300QD/SOW imaging. The thickness was varied by stacking layers of skin tissues beneath the skin surface. The bright spots in 800QD/FOW images seemed

to originate from autofluorescence from the skin.43 They were not seen when tissue samples were thin. Although the autofluorescence intensity was negligible when compared to the QD signal in measurements of thin tissue samples, it approached the QD signal levels as the skin tissue thickened. In contrast, autofluorescence was not observed in 1,300QD/SOW images. This demonstrates that SOW can be freer from the autofluorescence problems as the detection wavelengths lie at the longer wavelengths. However, the existence of autofluorescence spots in 800QD/FOW images in the FOW did not show a significant effect on CNRs or RASs. We also performed

346

Won et al

the experiments using different skins that did not have autofluorescence spots and obtained similar results (see Figure S2 online). CNRs decreased exponentially as the skin tissue thickness increased in both 800QD/FOW images (d 5 8.5 mm) and 1,300QD/SOW images (d 5 11.6 mm) (Figure 4E). By extrapolating the measured CNRs, imaging depths by CNR criterion were estimated to be 8.0 mm for 800QD/FOW and 11.6 mm for 1,300QD/ SOW. RASs of 800QD/FOW images increased more rapidly than those of 1300QD/SOW images (Figure 4F). By the RAS criterion, imaging depths were 4.2 mm for 800QD/FOW and 11.9 mm for 1,300QD/SOW. Contrary to the liver cases, 1300QD/SOW in skin tissue showed deeper imaging depths by both CNR and RAS criteria. By applying both CNR and RAS criteria, the imaging depths were 4.2 mm for 800QD/FOW and 11.6 mm for 1,300QD/ SOW. Overall, the skin imaging depth is almost three times deeper in the SOW than in the FOW (Table 1). QDs have the potential for unparalleled imaging capabilities, especially in the SOW. Molecular fluorophores are not available for SOW wavelengths. Lanthanide-doped nanoparticles can be used for the SOW but show smallerabsorption cross sections than QDs, limited emission wavelength tunability, and typically low QYs.44 We used 5% QY PbSe QDs to determine the imaging depths of QDs in SOW. However, QDs in SOW can be bright, with a QY well above 50%.28,45,46 Imaging capabilities using 1,300QD/ SOW were studied while varying the QYs of QDs. We prepared 1,300QD solutions with three different QYs. The QDs were identical in the size and composition, and the concentration of the solutions was also same. They look alike under a visible camera (Hitachi, HV-D37A), but the difference in their fluorescence intensities is easily detected using an InGaAs CCD camera (Figure 5A). The brightness (signal intensity) of the QD sample was linearly proportional to QY. Using the same methods as in the tissue-depth assessment experiments (see Figure 4), thicknesses of bovine liver tissues and porcine skin tissues were varied and the corresponding SOW imaging data

were analyzed (raw data: see Figure S3 online). In liver tissue, imaging depths based on the CNR criterion increased logarithmically with the QY (Figure 5C). Because the QD solutions had identical absorbing power, the number of photons emitted from the QD samples was directly proportional to the QY. The number of photons should decrease exponentially while propagating through the highly absorbing liver tissue. Because the CNR criterion requires that a certain number of photons over the background level must reach the detector, imaging depth should be logarithmically proportional to the QY. Simple extrapolations show that the imaging depth in liver tissues can be extended to 21.2 mm for our type of 1,300QDs with 50% QY; this is 2.1 times larger than the 10.0 mm depth obtained using QDs with 5% QY (see Table 1). Imaging depths based on the RAS criterion also increased logarithmically with the QY (see Figure 5C). RAS is predominantly dependent on the scattering properties of the medium, and photon propagation decays exponentially under scattering conditions. Imaging depths by the RAS criterion for 1,300QDs with 50% QY were 128.1 mm, which is 10.4 times larger than the 12.3 mm for 5% QY QD sample (see Table 1). In liver tissue, imaging capability is predominantly limited by the CNR criterion (see Figure 4, A–C). Experiments were repeated using skin tissues of various thicknesses. Estimated imaging depths using the CNR and RAS criteria for 1,300QD in skin tissue with 50% QY were 19.3 mm and 17.5 mm, respectively. These are 1.7 times and 1.5 times greater, respectively, than those obtained using QDs with 5% QY (see Table 1). In skin tissue, imaging capability was more limited by the RAS criterion than by the CNR criterion (see Figure 4, D–F), and 1,300QD/SOW imaging depth in skin showed 1.5 times enhancement as the QY increased from 5 to 50%. Imaging depths were also investigated using different concentrations of the 1,300QD. For future tumor targeting-based diagnosis, the QD concentration is directly related to the local QD concentration in a tumor mass or a

Table 1. Imaging Depths Using Quantum Dots in Tissues Evaluated by Contrast to Noise Ratio and Relative Apparent Size Criteria Imaging Depth (mm) Type of Tissue Bovine liver Porcine skin

Emission Wavelength of QD (nm)

By CNR Criterion

By RAS Criterion

800 1300 800 1300

13.0 10.0 8.0 11.6

11.2 12.3 4.2 11.9

CNR 5 contrast to noise ratio; QD 5 quantum dot; RAS 5 relative apparent size.

Imaging Depths of Near-Infrared Quantum Dots in First and Second Optical Windows

347

Figure 5. Effects of quantum yield (QY), concentration, and fluence rate on the imaging depth of 1,300 nm emitting quantum dots (QD) in biologic tissues. A and B, Color CCD image (left), InGaAs CCD image (center), and signal intensity profile at the horizontal line in the InGaAs CCD image (right) for the three QD samples of different QYs (A) and for the four QD samples of different concentrations (B). Imaging depths by contrast to noise ratio (CNR) and relative apparent size (RAS) criteria versus QYs (C), versus concentration (D), and versus fluence rate (E). It is noted that the plots have logarithmically scaled horizontal axes.

small collection of malignant cells. For in vivo imaging experiments such as tumor diagnosis or lymph node mapping, administration of QD solutions to tissues results in QD concentrations being diluted by a factor of up to 1,000.8 Although QDs can accumulate in the targeted area, the QD signal can be diluted by excretion and cell division.47 We studied the dependence of imaging capability on QD concentration while diluting a 10 mM 1,300QD solution up to 30 times. Figure 5B shows vials of 1,300QD solutions of four different concentrations under a color CCD camera and under an InGaAs CCD camera. Contrary to the linear dependence of the brightness over the QY, the brightness was not linearly proportional to the concentration because of the significant reabsorption in the QD solutions of concentrations . 1 mM. Imaging depths of the four 1,300QD samples were determined using liver tissues (Figure 5D and Figure S3 online). The imaging depths were less dependent on concentration than on the QY. It is noted that the horizontal axes of Figure 5, C and D, are both logarithmically scaled but with different ranges. When the QY of the 1,300QD sample increased from 3.6 to 8.7%, the imaging depth in the liver increased

by 4.1 mm but increased by only 1.9 mm when the QD concentration increased from 1 to 3 mM. The experiments were repeated while varying the skin tissue thickness. The imaging depth in skin increased by only 0.2 mm when the QD concentration increased from 1 to 3 mM. The concentration dependence experiment suggests that the imaging capability loss owing to dilution of QDs is not severe if a reasonable concentration (presumably . 1 mM) of QD can reach the target area. This also suggests that the imaging depth can be effectively extended by using brighter QDs (ie, higher QY) than by using higher concentrations of QDs. The fluence rate of incident light is also a critical parameter that affects imaging depth. In principle, penetration depth, which is defined as the depth at which the light intensity is attenuated to 1/e times the original intensity, should be independent of the power of the incident light.41 However, in reality, and especially when using an InGaAs CCD camera, both CNR and RAS are dependent on the fluence rate F because of thermal noise: as tissue thickness increases, the exposure time required for optimal imaging increases; thus, the thermal noise

348

Won et al

effect increases. As a result, a higher incident light intensity can increase the imaging depth. We studied the effect of F on the depth of 1,300QD imaging in the SOW using liver and skin tissues. The concentration of the QD sample was 10 mM, and the QY was 8.7%. According to the US Food and Drug Administration (FDA), human access during operation of a continuous laser at 400 # l # 1,400 nm is permitted up to 0.5 W.48 In our 5 3 5 cm field, the FDA-allowed F corresponds to 20 mW cm–2. Imaging depths were obtained while varying the thickness of liver or skin tissues under 3 # F # 20 mW cm–2 (Figure 5E and Figure S3 online). In liver tissue, the imaging depths of 1,300QD/SOW at F 5 20 mW cm–2 were 13.3 mm by the CNR criterion and 27.1 mm by the RAS criterion. In skin tissue, the imaging depths were 17.0 mm by the CNR criterion and 13.3 mm by the RAS criterion. By increasing the F from 3 to 20 mW cm–2, the imaging depth (CNR and RAS combined) was extended by 1.4 times in liver and by 1.8 times in skin. Increased F can continuously increase QDs’ NIR absorption because they are surrounded by biologic media and the effective fluence rate (photon flux that reaches QDs) remains much lower than the QD absorption saturation level. The imaging depth of 1,300QD/SOW can be further increased by using longer lex than 660 nm. We used 660 nm excitation for the comparison between QD imaging in the FOW and the SOW. The 1,300QD/SOW imaging (see Figure 4) was repeated to compare the effects of lex at 660 nm and 850 nm under a constant F 5 10 mW cm–2 (Figure 6). In liver tissue, samples excited at 660 nm showed higher background noise levels than those excited at 850 nm (see Figure 6A). As tissue thickened, CNRs of samples excited at 660 nm decreased more rapidly than did those excited at 850 nm (see Figure 6B). This difference occurs primarily because liver tissue absorbs 660 nm photons more effectively than 850 nm photons. The imaging depths by the CNR criterion were 10.0 mm for lex 5 660 nm and 14.4 mm for lex 5 850 nm. The imaging depths by the RAS criterion were 12.3 mm for lex 5 660 nm and 23.9 mm for lex 5 850 nm (see Figure 6C). As in Figure 4, the imaging depth in liver tissue was more critically limited by the CNR criterion than by the RAS criterion. Based on the CNR criterion, the 1,300QD imaging depth in liver was increased 1.4 times by switching lex from 660 to 850 nm. The experiments were repeated using porcine skin tissues of different thicknesses (see Figure 6, D–F). Images obtained using 1,300QD/SOW at lex 5 660 nm were similar to those at lex 5 850 nm. The imaging depths by the CNR criterion were 12.0 mm for lex 5 660 nm and

12.2 mm for lex 5 850 nm. The imaging depths by the RAS criterion showed similar values at both lex 5 660 nm and lex 5 850 nm. Skin tissue does not absorb many photons, so the imaging properties are heavily governed by photon scattering. Because the scattering difference between 660 and 850 nm photons is small, these two wavelengths showed marginal difference in skin imaging depth. The results of these studies of 1,300QD/SOW imaging capability, while varying the QY, QD concentration, F, and lex, can be used to estimate 1,300QD/SOW imaging depths for hypothetical conditions of QY, F, and lex. In liver tissue, 1,300QDs with 8.7% QY under F 5 10 mW cm–2 at lex 5 660 nm showed imaging depths of 12.4 mm (CNR limited) and 11.1 mm in skin tissue (RAS limited). As F increased to 20 mW cm–2, the imaging depths were extended to 13.3 mm for both liver and skin tissues. If the QY of the QD sample can be increased to 50%, the imaging depths can be extended to 21.2 mm for liver tissue and 17.5 mm for skin tissue. If the lex were switched to 850 nm, the liver imaging depth can be further extended to 29.7 mm, whereas the skin imaging depth remains the same at 17.5 mm. This case represents optimal 1,300QD/SOW imaging under continuous excitation at the FDA-approved F. QD imaging in the SOW is expected to have guaranteed imaging depths of 2 to 3 cm under optimal continuous excitation. In addition, the QY of QDs can exceed 50% and the imaging capability can be improved using softwarebased postprocesses. An imaging depth . 2 cm can provide numerous potential applications of QD imaging. It can be applied to diagnostic tools for tumors and lesions near the skin. The imaging depth may approach that required for breast tumor diagnosis. Given that the fluorescence imaging herein can be easily extended to endoscopic equipments, the imaging can find more applications in internal organs as well. However, the imaging depth does not support whole-body imaging of humans, as magnetic resonance imaging and computed tomography do. Given that time-gated pulse F can be higher than the continuous illumination limit, the imaging depth for F that exceeds the FDA limit was estimated. If F were increased 10 times (200 mW cm–2), imaging depths could be expected to be 36.3 mm for liver tissue and 25.1 mm for skin tissue. An increase of 100 times results in the imaging depths of 42.9 mm for liver tissue and 32.8 mm for skin tissue. Achieving < 5 cm imaging depth requires a 1,000 times increase in F and QDs with 100% QY (Table 2). This approaches the condition to achieve whole-body imaging of humans.

349

Imaging Depths of Near-Infrared Quantum Dots in First and Second Optical Windows

Figure 6. Imaging capabilities of 1,300QDs when excited by 660 nm (solid lines with &) or by 850 nm (dashed lines with #) in biologic tissues. Near-infrared fluorescence images (A, D), contrast to noise ratios (CNR) (B, E), and relative apparent sizes (RAS) (C, F) are shown in bovine liver tissues (A–C) and in porcine skin tissues (D–F). Dotted lines represent imaging depth criteria. The thickness of the tissues was varied from 0 to 12 mm. It is noted that the CNR axes are logarithmically scaled.

Table 2. Measured and Predicted Imaging Depths of 1,300QDs by Contrast to Noise Ratio and Relative Apparent Size Criteria under Different Conditions

Quantum Yield (%) 8.7 8.7 50 50 50 50 100

Imaging Depth (mm)

Excitation Wavelength (nm)

Fluence Rate (mW cm–2)

Bovine Liver (CNR limited)

Porcine Skin (RAS limited)

660 660 660 850 850 850 850

10 20 20 20 200 2000 20000

12.4 13.3 21.2* 29.7* 36.3* 42.9* 54.1*

11.1 13.3 17.5* 17.5* 25.1* 32.8* 42.9*

CNR 5 contrast to noise ratio; QD 5 quantum dot; RAS 5 relative apparent size. *Predicted value.

350

Won et al

It is stressed that the imaging depths estimated here are closer to the lower bound. The actual imaging depth can be deeper and can also be enhanced by postimage processes. It is also noted that the imaging depths are already deep enough for whole-body imaging of small animals. These predicted imaging depths under different conditions are listed in Table 2. Comparative studies between 800QD/FOW and 1,300QD/SOW imaging have been performed showing significant advantages of the SOW over the FOW. We performed small-animal (mouse) whole-body imaging experiments to compare 800QD/FOW and 1,300QD/SOW imaging capabilities and to validate our estimates. Identical excitation conditions (F 5 20 mW cm–2) with the lex of 660 nm were used. A cylindrical glass container of 5 mm width was filled with 0.3 mL QD hexane solution. The QD solution had both 800QDs and 1,300QDs. The concentrations were adjusted so that the A of the 800QDs and the 1,300QDs at lex were identical. The QYs of both QDs were both < 5%. The QD cylinder was imaged using an Si CCD camera and an InGaAs CCD camera (Figure 7A). The cylinder looked slightly brighter under the Si CCD camera (CNR 5 20.1) than under the InGaAs CCD camera (CNR 5 18.3). The cylinder was inserted into the abdominal cavity of a mouse below the intestine, and the skin incision was closed using a suture (see Materials and Methods, above). In vivo fluorescence images of the mouse were obtained using Si and InGaAs CCD cameras. Considering the imaging setup, the imaging depths for this configuration were 14.5 mm in liver tissue and 6.5 mm in skin tissue for 800QDs and 11.6 mm in liver tissue and 14.2 mm in skin tissue for 1,300QDs. The QD cylinder in the mouse was

imaged from the ventral side using both Si and InGaAs CCD cameras (Figure 7B). The QD cylinder images were seriously distorted by the internal organs. The QD cylinder glowed less brightly under the Si CCD camera than under the InGaAs CCD camera. The CNRs were 6.0 for the FOW image and 10.0 for the SOW image. The CNR of the 1,300QD/SOW image was almost double that of the 800QD/FOW image CNR. The mouse was flipped and imaged from the dorsal side (Figure 7C). No detectable image was obtained for 800QD/FOW. Reflections at the dorsal skin were imaged only under an Si CCD camera. To the contrary, the 1,300QD/SOW image of the QD cylinder was obtained under an InGaAs CCD camera. The CNR was only 2.5, which does not meet our CNR criterion. This confirms that our CNR and RAS criteria guarantee the depths at which discernible images can be obtained. As is the case in Figure 7C, objects beyond the imaging depths by our criteria can be visualized.

Conclusion QD imaging depths in biologic tissues were investigated using the criteria of CNR and RAS. In skin tissue, 800QD/ FOW had an imaging depth of 4.2 mm; under the same conditions, 1,300QD/SOW had an imaging depth of 11.6 mm. Owing to the reduced scattering in the SOW, the imaging depth in skin could be extended by approximately three times for 1,300QD/SOW. In liver, 1,300QD/SOW imaging showed similar imaging depth when compared to 800QD/FOW imaging. However, excitation of 1,300QD/SOW imaging can be shifted to

Figure 7. Small-animal (mouse) whole-body imaging using 800QDs and 1,300QDs. A, Image of a cylinder containing a mixture of 800QDs and 1,300QDs in hexane. B, Ventral views of mouse with the quantum dot (QD) cylinder inserted in the abdominal cavity. C, Dorsal views of the same mouse. In all rows: left, color CCD images; middle, NIR Si CCD images; right, NIR InGaAs CCD images. A 660 nm LED lamp was used for excitation at the fluence rate of 20 mW cm–2. CNR 5 contrast to noise ratio.

Imaging Depths of Near-Infrared Quantum Dots in First and Second Optical Windows

longer wavelengths; thus, the imaging depth in liver tissue can be extended by reducing the absorption of the illuminating photons. For example, changing the excitation from 660 to 850 nm at the same F extended the liver imaging depth by 1.4 times. Imaging depths using 1,300QD/SOW were further expected for hypothetical conditions. Based on our criteria, using continuous excitation of FDA-approved F under optimal conditions, 1,300QDs with 50% QY can attain imaging depths of 29.7 mm in liver tissue and 17.5 mm in skin tissue. Using time-gated excitations with the pulse F 1,000 times higher than FDA-permitted, timeaveraged F, 1,300QD/SOW imaging can approach an imaging depth of < 5 cm. The imaging depth based on our criteria is close to the minimum guaranteed tissue depth for a discernible image, and in reality, deeper imaging may be achievable. Advances in the sensitivity of cameras in the SOW and in software postprocessing of QD images are expected to further increase the imaging capabilities of QDs in the SOW. We are hopeful that QDs in the SOW can promise deep tissue penetrating in vivo imaging, which can open a new avenue for many biomedical applications.

Acknowledgment Financial disclosure of authors: This work was supported by a Korea Science and Engineering Foundation grant funded by Ministry of Science and Technology 20110018601, the Korea Health 21 R&D Project Ministry of Health & Welfare (A101626), the Priority Research Center Program through National Research Foundation of Korea 2010-0029711, 20110027727, and the Basic Science Research Programs 20110027236. Financial disclosure of reviewers: None reported.

References 1. Bruchez M, Moronne M, Gin P, et al. Semiconductor nanocrystals as fluorescent biological labels. Science 1998;281:2013–6, doi:10. 1126/science.281.5385.2013. 2. Chan WCW, Nie S. Quantum dot bioconjugates for ultrasensitive nonisotopic detection. Science 1998;281:2016–8, doi:10.1126/science. 281.5385.2016. 3. Medintz IL, Uyeda HT, Goldman ER, et al. Quantum dot bioconjugates for imaging, labelling and sensing. Nat Mater 2005;4: 435–46, doi:10.1038/nmat1390. 4. Weissleder R. A clearer vision for in vivo imaging. Nat Biotechnol 2001;19:316–7, doi:10.1038/86684. 5. Zaheer A, Lenkinski RE, Mahmood A, et al. In vivo near-infrared fluorescence imaging of osteoblastic activity. Nat Biotechnol 2001; 19:1148–54, doi:10.1038/nbt1201-1148.

351

6. Lim YT, Kim S, Nakayama A, et al. Selection of quantum dot wavelengths for biomedical assays and imaging. Mol Imaging 2003; 2:50–64, doi:10.1162/153535003765276282. 7. Smith AM, Mancini MC, Nie S. Second window for in vivo imaging. Nat Nanotechnol 2009;4:710–1, doi:10.1038/nnano. 2009.326. 8. Kim S, Lim YT, Soltesz EG, et al. Near-infrared fluorescent type II quantum dots for sentinel lymph node mapping. Nat Biotechnol 2004;22:93–7, doi:10.1038/nbt920. 9. Pons T, Pic E, Lequeux N, et al. Cadmium-free CuInS2/ZnS quantum dots for sentinel lymph node imaging with reduced toxicity. ACS Nano 2010;4:2531–8, doi:10.1021/nn901421v. 10. Soltesz EG, Kim S, Laurence RG, et al. Intraoperative sentinel lymph node mapping of the lung using near-infrared fluorescent quantum dots. Ann Thorac Surg 2005;79:269–77, doi:10.1016/ j.athoracsur.2004.06.055. 11. Ballou B, Ernst LA, Andreko S, et al. Sentinel lymph node imaging using quantum dots in mouse tumor models. Bioconjug Chem 2007;18:389–96, doi:10.1021/bc060261j. 12. Stroh M, Zimmer JP, Duda DG, et al. Quantum dots spectrally distinguish multiple species within the tumor milieu in vivo. Nat Med 2005;11:678–82, doi:10.1038/nm1247. 13. Smith BR, Cheng Z, De A, et al. Real-time intravital imaging of RGD2quantum dot binding to luminal endothelium in mouse tumor neovasculature. Nano Lett 2008;8:2599–606, doi:10.1021/ nl080141f. 14. Bhang SH, Won N, Lee T-J, et al. Hyaluronic acid2quantum dot conjugates for in vivo lymphatic vessel imaging. ACS Nano 2009;3: 1389–98, doi:10.1021/nn900138d. 15. Yoo JS, Won N, Kim HB, et al. In vivo imaging of cancer cells with electroporation of quantum dots and multispectral imaging. J Appl Phys 2010;107:124702-1–9, doi:10.1063/1.3447858. 16. Hyun B-R, Chen, , Rey DA, et al. Near-infrared fluorescence imaging with water-soluble lead salt quantum dots. J Phys Chem B 2007;111:5726–30, doi:10.1021/jp068455j. 17. Cherukuri P, Bachilo SM, Litovsky SH, et al. Near-infrared fluorescence microscopy of single-walled carbon nanotubes in phagocytic cells. J Am Chem Soc 2004;126:15638–9, doi:10.1021/ ja0466311. 18. Welsher K, Liu Z, Sherlock SP, et al. A route to brightly fluorescent carbon nanotubes for near-infrared imaging in mice. Nat Nanotechnol 2009;4:773–80, doi:10.1038/nnano.2009.294. 19. Welsher K, Sherlock SP, Dai H. Deep-tissue anatomical imaging of mice using carbon nanotube fluorophores in the second nearinfrared window. Proc Natl Acad Sci U S A 2011;108:8943–8, doi:10.1073/pnas.1014501108. 20. Li L, Daou TJ, Texier I, et al. Highly luminescent CuInS2/ZnS core/shell nanocrystals: cadmium-free quantum dots for in vivo imaging. Chem Mater 2009;21:2422–9, doi:10.1021/cm 900103b. 21. Yong K-T, Ding H, Roy I, et al. Imaging pancreatic cancer using bioconjugated InP quantum dots. ACS Nano 2009;3:502–10, doi:10.1021/nn8008933. 22. Bang J, Park J, Lee JH, et al. ZnTe/ZnSe (Core/Shell) type-II quantum dots: their optical and photovoltaic properties. Chem Mater 2009;22:233–40, doi:10.1021/cm9027995. 23. Choi HS, Liu W, Misra P, et al. Renal clearance of quantum dots. Nat Biotechnol 2007;25:1165–70, doi:10.1038/nbt1340.

352

Won et al

24. Choi HS, Ipe BI, Misra P, et al. Tissue- and organ-selective biodistribution of NIR fluorescent quantum dots. Nano Lett 2009; 9:2354–9, doi:10.1021/nl900872r. 25. Gao J, Chen K, Xie R, et al. Ultrasmall near-infrared non-cadmium quantum dots for in vivo tumor imaging. Small 2010;6:256–61, doi:10.1002/smll.200901672. 26. Gonza´lez RC, Woods RE, Eddins SL. Digital image processing using MATLAB. Upper Saddle River, NJ: Pearson Prentice Hall; 2004. 27. Bailey RE, Nie S. Alloyed semiconductor quantum dots: tuning the optical properties without changing the particle size. J Am Chem Soc 2003;125:7100–6, doi:10.1021/ja035000o. 28. Yu WW, Falkner JC, Shih BS, et al. Preparation and characterization of monodisperse PbSe semiconductor nanocrystals in a noncoordinating solvent. Chem Mater 2004;16:3318–22, doi:10. 1021/cm049476y. 29. Wang LV, Wu H-i. Biomedical optics: principles and imaging. Hoboken, NJ: John Wiley & Sons, Inc.; 2007. 30. Rose A. Vision: human and electronic (optical physics and engineering). New York: Springer; 1973. 31. Song X, Pogue BW, Jiang S, et al. Automated region detection based on the contrast-to-noise ratio in near-infrared tomography. Appl Opt 2004;43:1053–62, doi:10.1364/AO.43.001053. 32. Hendrick RE. Breast MRI: fundamentals and technical aspects. New York: Springer; 2008. 33. Fujita H, Tsai DY, Itoh T, et al. A simple method for determining the modulation transfer function in digital radiography. IEEE Trans Med Imaging 1992;11:34–9, doi:10.1109/42.126908. 34. Sitter JDN, Goddard JS, Ferrell RK. Method for the measurement of the modulation transfer function of sampled imaging systems from bar-target patterns. Appl Opt 1995;34:746–51, doi:10.1364/ AO.34.000746. 35. Bushberg JT, Seibert JA, Leidholdt EM, Boone JM. The essential physics of medical imaging. 2nd ed. Philadelphia: Lippincott Williams & Wilkins; 2002. 36. Wray CJ, Lowy AM, Mathews JB, et al. The significance and clinical factors associated with a subcentimeter resection of

37.

38.

39.

40.

41.

42.

43.

44.

45.

46. 47.

48.

colorectal liver metastases. Ann Surg Oncol 2005;12:374–80, doi:10.1245/ASO.2005.06.038. Marom EM, Sarvis S, Herndon JE, et al. T1 lung cancers: sensitivity of diagnosis with fluorodeoxyglucose PET. Radiology 2002;223: 453–9, doi:10.1148/radiol.2232011131. Cheong WF, Prahl SA, Welch AJ. A review of the optical properties of biological tissues. IEEE J Quantum Electron 1990;26:2166–85, doi:10.1109/3.64354. Demos SG, Alfano RR. Temporal gating in highly scattering media by the degree of optical polarization. Opt Lett 1996;21:161–3, doi:10.1364/OL.21.000161. Tromberg BJ, Shah N, Lanning R, et al. Non-invasive in vivo characterization of breast tumors using photon migration spectroscopy. Neoplasia 2000;2:26–40, doi:10.1038/sj.neo.7900082. Dunn AK, Wallace VP, Coleno M, et al. Influence of optical properties on two-photon fluorescence imaging in turbid samples. Appl Opt 2000;39:1194–201, doi:10.1364/AO.39.001194. Wang X-Q, Chen J-Y, Mi L, et al. A comparison of tissue penetrations between single and two-photon excitations. Appl Phys Lett 2009;95:143705–3, doi:10.1063/1.3246793. De Grand AM, Lomnes SJ, Lee DS, et al. Tissue-like phantoms for near-infrared fluorescence imaging system assessment and the training of surgeons. J Biomed Opt 2006;11:0014007, doi:10.1117/ 1.2170579. Stouwdam JW, van Veggel FCJM. Near-infrared emission of redispersible Er3+, Nd3+, and Ho3+ doped LaF3 nanoparticles. Nano Lett 2002;2:733–7, doi:10.1021/nl025562q. Wehrenberg BL, Wang C, Guyot-Sionnest P. Interband and intraband optical studies of PbSe colloidal quantum dots. J Phys Chem B 2002;106:10634–40, doi:10.1021/jp021187e. Du H, Chen C, Krishnan R, et al. Optical properties of colloidal PbSe nanocrystals. Nano Lett 2002;2:1321–4, doi:10.1021/nl025785g. Frangioni JV, Hajjar RJ. In vivo tracking of stem cells for clinical trials in cardiovascular disease. Circulation 2004;110:3378–83, doi:10.1161/01.CIR.0000149840.46523.FC. Food and Drug Administration. Performance standards for lightemitting products. http://www.fda.gov (accessed March 29, 2011).