In some ways, molecular imaging in small animals has been like having a map of the neighborhood without knowing where the neighborhood is within the city. Investigators have imaged targets deep within animals but have had trouble clearly resolving and identifying the associated internal organs and other anatomical features.Now Elizabeth M.C. Hillman of Columbia University in New York and Anna Moore of Massachusetts General Hospital in Charlestown have demonstrated an all-optical method for anatomical co-registration. The technique exploits the dynamics of dye distribution within the animal to create contrast and to delineate organs, thereby ensuring that the captured image has a context.“In vivo nondestructive imaging is great, but validation of what you are measuring is critical,” said Hillman, an assistant professor of biomedical engineering.In optical molecular imaging, fluorescent and bioluminescent probes label targeted cells and tissues that are often deep within the animal. The light that emerges is scattered and attenuated so much that locating where the signal originated can be very challenging. The problem is compounded because the better the probe targets a certain tissue or cell, the less the images reveal nearby organs that might help identify from where the photons are coming.One solution to the problem involves combining optical imaging with another technique such as x-ray, micro-CT, ultrasound or MRI. However, x-rays do not travel through tissue in the same way that visible light does, and micro-CT and ultrasound have little contrast in soft-tissue organs. Thus, registration of any of these with the optical images is difficult. Also, adding any of the secondary imaging modalities increases the cost and complexity of the setup.Biodynamics creates contrastWhile working on a method to differentiate between the uptake and the washout of a targeted dye, Hillman experimented with a mixture of the tracer dyes indocyanine green and dextran Texas red. The first has an emission peak in the near infrared and a low molecular weight, while the second has an emission peak in the visible and a high molecular weight. After injecting the dye mixture, Hillman imaged the whole animal and studied how the data changed over time.A tracer dye injected into an animal model follows a different time course in various organs (a), enabling researchers to determine where anatomical features are using only optical methods (b). Images reprinted with permission of Nature Photonics.“When I sat down and analyzed the image time-series, I was really amazed to see all the organs,” she recalled.The differentiation occurred because the injected dye followed a different time course in various organs. Thus, each organ produced a distinct optical signal that varied over time. Those signals can be separated, illuminating specific organs — the biodynamics within the animal create the contrast.In a demonstration of the imaging technique, the researchers positioned a mouse on a slightly raised stage between two mirrors mounted at right angles to each other and at 45¼ to the animal stage. The mirrors allowed the top and sides of the mouse to be observed simultaneously from above. In this dynamic contrast imaging setup, mirrors below the animal model provide views from three sides to the camera mounted above. Side illumination excites fluorescence of an injected tracer dye, with the two light sources allowing dye of two different excitation-emission wavelengths to be used. By synchronizing the light sources with the motorized filter changer and the camera, researchers can track two tracer dyes in the animal at the same time.They mounted a 12-bit cooled Photometrics CCD camera above the mouse and placed a computer-controlled emission Electro-Optical Products filter changer in front of the camera. The filter changer contained a 600-nm long-pass filter to detect the visible emission peak and an 830- to 870-nm bandpass filter for the near-IR emission peak.To excite the visible dye, they used white light filtered to 550 to 590 nm that traveled through a light guide to shine on the animal from both sides. For the near-IR dye’s light source, they used two 785-nm laser diodes. They were able to synchronize the illumination with the emission filters and the camera acquisition by using a shutter in the visible light path and by rapidly cycling the diodes on and off.As reported in the September issue of Nature Photonics, the researchers injected the dye mixture into the animal’s tail and then acquired 10 images of 50 ms each every 2 s for each emission-excitation pair. They did this for up to 40 min after the injection. From this data they extracted images for a time-series analysis. Qualitatively, noted Hillman, it is easy to see what happens by looking at the early data, particularly for the near-IR dye. “There are definite clear trends in the first few seconds — the dye goes to the heart, then the lungs, then the brain, the kidneys, washes quickly out of the brain and all ends up in the liver,” she said.For a more quantitative extraction of the time courses of the dye through the animal, they used principal component analysis. This approach pulled out the most common time courses and showed how they were represented by the images. However, principal component analysis is sensitive to noise in the data, and so, for a more robust algorithm, they used a non-negative least squares fit to time courses extracted from various regions of interest. They selected those regions from the raw data based on principal component analysis.According to the investigators, the results of the whole-animal imaging agreed well with a digital atlas. Attempts to quantify the error between the derived optical image and the actual anatomy using micro-CT were hampered by poor x-ray contrast in the soft tissue organs. To truly determine the error, it might be necessary to do microtome imaging of mice, with the animal sliced and photographed bit by bit. While Hillman indicated that such a study might be done in the future, she noted that it may not be that important. The fluorescent light generated by the tracer dye would follow the same scattering path that would be followed by light from a targeted dye, assuming the two have similar wavelengths and are in the same organs. Hence, it does not matter whether there is some distortion, as long as the distortion is the same for the tracer and the targeted dye.A patent has been filed and the technology licensed to CRI Inc., of Woburn, Mass., which offers multispectral imaging and analysis tools based on various emission wavelengths of dyes. Plans call for a product launch within the next few months, and Hillman is working with the company on this.On the research side, she now uses only the near-IR dye, as it offers some advantages, and the setup has no moving parts. She plans to extend the technique by, among other things, using different fluorescent tracer dyes. Doing this may take some work, but the biodynamics and sensitivities of various dyes could yield new results and offer new capabilities.“The dynamics of some dyes could even allow noninvasive evaluation of the function of the organs,” Hillman said.