Molecular imaging, the imaging of biological processes at cellular and molecular levels in vivo, is a challenging task and is typically not addressed with any single modality, but instead with a combination of imaging tools. Nuclear medicine techniques like SPECT and PET provide excellent sensitivity for detecting the presence of very small quantities of probe, but generally have poor spatial resolution making it difficult to localize where the probe resides.1 Optical imaging has excellent sensitivity and spatial resolution, but light photons interact strongly with tissues leading to poor penetration.2,3 Thus, optical imaging can only image events to limited depths. CT and MRI have exquisite spatial resolution, but much poorer sensitivity. To overcome the limitations of individual modalities, modern cameras have combined the best of two (or more) worlds, for example, merging nuclear medicine cameras with CT or MRI, and allowing the visualization of molecular processes with both high sensitivity and high spatial resolution.

Nuclear Medicine Techniques

Nuclear medicine probes contain a radioactive isotope that emits a particle of radiation. Nuclear medicine cameras are able to count those individual particles. The type of camera is dictated by the type of radiation emitted: SPECT cameras detect those isotopes that decay through the direct emission of single gamma rays, while PET cameras detect the pair of gamma rays emitted following a positron decay. Traditional SPECT cameras consist of a collimator, a scintillation crystal, an array of photomultiplier tubes (PMT), and a computer for recording and processing the signals (Figure 1A). The collimator acts as the lens of the camera by defining the direction that the gamma rays are traveling. It does this by blocking the rays that are not going in the desired direction and thus collimators are the Achilles heel of the gamma camera because they throw away >99% of the incoming rays. The overall geometrical sensitivity of a clinical gamma camera is ~2 × 10−4. Those rays that make it past the collimator are stopped in the scintillation crystal which converts the high-energy gamma ray into a shower of lower-energy light photons. The light signal is amplified by the PMT to create an electrical signal that can be measured and stored in the computer. New SPECT cameras that are dedicated to cardiac imaging have also recently been introduced for clinical use.4 These dedicated systems have large numbers of detectors and novel collimator designs which give them an increase in sensitivity of 4 to 8 times over traditional cameras. Approximate sensitivities and spatial resolutions of the nuclear medicine and optical imaging systems are given in Table 1.

Figure 1
figure 1

Nuclear medicine and optical cameras for molecular imaging: (A) gamma camera, (B) positron emission tomography scanner, (C) optical camera for bioluminescent imaging, (D) optical camera for fluorescent imaging, and (E) an optoacoustic imaging system

Full size image
Table 1 Comparison of imaging modalities
Full size table

PET imaging (Figure 1B) detects the pair of gamma rays that are produced when a positron (a particle of anti-matter that is like an electron but with opposite charge) interacts with an electron in the patient tissues. The electron and positron annihilate each other and produce two 511-keV gamma rays traveling in opposite directions. Like SPECT, PET uses scintillation crystals and PMTs to convert the high-energy gamma rays into a detectable electronic signal. The two gamma rays should arrive at the detectors at almost the same time, so a coincidence timing circuit is used to check if two detected events came from the same annihilation event. Because there are two gamma rays, when they are detected they define two ends of a line that corresponds to their direction of travel. This approach eliminates the need for the collimator and results in an increase in sensitivity of ~100× for PET over SPECT.

A strength of nuclear medicine imaging is that the exact same probes can be used in both small-animal and human imaging which can facilitate rapid translation from pre-clinical development and validation to clinical application. To capitalize on this, both small-animal SPECT and PET cameras have been developed. Although the basic imaging principles are same, there are some important differences between the small-animal and clinical scanners. With SPECT, the typical clinical spatial resolution is 10 to 15 mm which is insufficient for mouse imaging. To address this problem, the parallel-hole collimator is replaced with a pinhole collimator which acts as a magnifying lens. The sensitivity of pinhole cameras is very good for small sources located very close to the pinhole aperture, but falls off rapidly with distance. To increase sensitivity, most micro-SPECT cameras use multiple detectors each with multiple pinhole apertures (typically >36 pinholes) resulting in an overall sensitivity of up to 0.3% with resolutions of 1.5 mm in rats and 0.5 mm in mice. With micro-PET scanners, the diameter of the detector ring is greatly reduced and the bore’s axial length to diameter ratio is greatly increased. The changes in geometry lead to better solid-angle coverage and increase the sensitivity in the center of the field-of-view to ~10%. Using smaller crystals in the detectors, and techniques to better locate the depth in the crystal at which the gamma-ray is absorbed, provide improved spatial resolution of ~1 mm.

Non-nuclear Techniques

Optical techniques have been available for more than a century for the evaluation of molecular processes in tissue samples, but over the last 20 years there has been a tremendous growth in the application of these techniques for in vivo imaging.2,3 Optical imaging has sensitivity similar to nuclear medicine approaches and superior spatial resolution for much lower costs, but the depth of signal penetration is limited. This has restricting it to small-animal applications and to imaging processes near to the surface. Imaging is done with optical lenses and CCD cameras, using filters to select the frequency band of the light to image. Small-animal subjects are often placed in a dark box to limit background light sources. Two common approaches to optical imaging are bioluminescence (BLI) and fluorescence imaging (FLI) (Figure 1C, D). In BLI, probes are labeled by a substrate like luciferin which is oxidized in the presence of an enzyme (luciferase) to produce light. Choosing a compound that is not present naturally yields a very specific signal with low background. Light strongly interacts with tissues such that the signal that escapes from the tissues is heavily attenuated and has scattered multiple times. Mutant forms of luciferase have been developed that emit light closer to near-infrared energies and so interact much less with tissues, but imaging remains limited to within a few millimeters of the surface. With fluorescence imaging, fluorophores (fluorescent compounds) in the subject are excited using an external light source. The fluorophores can be either endogenous or inserted as part of a fluorescently labeled probe. A disadvantage of FLI is that the auto-fluorescence of other tissues in the subject leads to a substantial background signal. Also, scatter and attenuation affect both the exciting and emitted light and so FLI, like BLI, is limited to imaging at depths of only a few millimeters.

Both BLI and FLI are planar imaging techniques, but there has also been a growth in 3D optical methods. Diffuse optical tomography (DOT) uses arrays of light emitters and detectors placed around a subject. The diffusion and attenuation of light as it passes through the tissue between the emitter and the receiver is modeled and a 3D image of the tissue’s optical properties can be created. With fluorescence molecular tomography (FMT), an exciting light source illuminates the subject from multiple directions and fluorescent images are acquired each time. With FMT, the problem of reconstructing an image of the fluorophores from the acquired images is more complicated because the scatter and attenuation of both the exciting and emitted light must be modeled. However, it provides both a 3D map of the distribution of the fluorophores in addition to a map of the tissue optical properties. The FMT reconstruction problem is very ill posed resulting in many potential solutions and degrading image quality, but the use of anatomical imaging (CT or MRI) can provide prior information about the distribution of tissues that greatly aids the reconstruction. DOT and FMT have better depth sensitivity than FLI/BLI and provide 3D information, but photon scatter and absorption still limit the depth at which they can be used. Optoacoustic imaging (also called photoacoustic imaging) improves on this. Energy from an exciting light source is absorbed by the fluorophore and it undergoes thermal expansion generating ultrasound waves (Figure 1E). The signal is then detected using ultrasound transducers. Because ultrasound waves are attenuated by 3 orders of magnitude less than light photons, the depth of imaging can be extended to several centimeters. Furthermore, with multispectral optoacoustic tomography, multiple frequencies of exciting light can be used to distinguish different fluorophores based on their spectral signature.

Hybrid Imaging

Both nuclear and optical techniques benefit from co-registration with high-resolution anatomical imaging modalities like CT and MRI. The relatively poor spatial resolution of nuclear techniques can make it difficult to accurately localize the detected signal. In addition, because they are functional imaging modalities, the nuclear images may conform only loosely to the anatomical structures and identification of anatomical landmarks usually depends on non-specific uptake in other organs. As molecular imaging tracers become more specific, the background signals disappear leaving us with a bright light from our functional imaging probe that is sitting in a dark room. Co-registered anatomical images provide context in which to evaluate the information provided by functional imaging. Optical imaging draws the same benefits of localization and, in addition, the structural information provided by CT or MRI can be used as prior information to help constrain the reconstruction of 3D optical tomographic images, directly improving the image quality. Finally, the benefits of combined anatomical and functional imaging have led to the proliferation of multi-modality cameras (both clinical and pre-clinical) which in turn opens the door to combining the functional imaging capabilities from more traditional techniques of nuclear medicine and optical with those from the new molecular imaging capabilities being developed for MRI.

Many different imaging modalities are available to probe molecular and cellular processes. Each modality has its strengths and weaknesses and the best camera to use will depend on the particular pathological process being examined. Although huge advances have been made over the last two decades, the technologies behind each modality and techniques for combining them continue to improve, enhancing the accuracy and the sensitivity with which molecular processes can be imaged.

New Knowledge Gained

This article has briefly reviewed some of the key concepts behind the tools used for nuclear medicine and optical imaging of molecular processes. Nuclear medicine imaging has the advantage of high sensitivity and direct translation between pre-clinical development and clinical implementation. Optical imaging has both high sensitivity and excellent spatial resolution at a low cost, but it’s clinical use is limited to targets near to the surface. All of these techniques benefit from combination with anatomical imaging such as CT or MRI.