Optical imaging sheds light on cancer's signature

June 14, 2004

Optical imaging may address one of the great paradoxes in radiology. X-ray mammography can depict breast microcalcifications, the smallest object relevant to a radiological diagnosis that can be seen with the human eye. Yet two of every 10 breast cancers still escape detection.

Although beset with technical challenges, near-infrared (NIR) optical imaging can potentially detect and monitor the treatment of breast cancers before the first suggestion of a problem appears with mammography, PET, or MR imaging. On a broader front, optical imaging techniques have already been embraced as an essential instrument for molecular-based drug research and development. Optical imaging has for years fueled discoveries in molecular biology laboratories. Today, researchers are applying its strengths-and wrestling with its challenges-in an effort to shed light on molecular secrets deep within living tissues.

Driving progress in optical molecular imaging are the development of biocompatible fluorescent markers, advances in imaging technology, and refinements in the mathematical models that make sense of the chaotic paths photons travel as they careen through the body. These advances are helping researchers accomplish important goals in securing the future of optical molecular imaging: improvement of light penetration, image resolution, and the quantification and localization of disease.


Several research teams are using diffuse optical tomography to detect breast tumors by searching out two intrinsic cancer signatures: increased blood flow, as shown by the total hemoglobin concentration, and hypermetabolism, as shown by a drop in oxygen concentration (Figure 1). Britton Chance, Ph.D, emeritus professor of biophysics, biochemistry, and radiology at the University of Pennsylvania, and colleagues have studied more than 100 women, including 30 with breast cancer, using diffuse optical tomography. They are finding the sensitivity and specificity for breast cancer detection to be more than 90% with this approach.

Chance, Brian W. Pogue, Ph.D., an associate professor of engineering at Dartmouth, and others are part of a network funded by the National Institutes of Health that is working toward a multicenter trial of NIR optical tomography of the breast. The imaging industry has taken notice as well: Montreal-based Advanced Research Technologies has developed a time-domain optical tomographic breast imaging device known as SoftScan and is testing it clinically in collaboration with McGill University, the Ottawa Regional Women's Breast Health Centre, and others (Figure 2).

Researchers are also investigating whether changes in blood flow and oxygenation can effectively evaluate the response to chemotherapy, detect and quantify fetal hypoxia, and diagnose ischemic stroke. Chance has used NIR spectroscopy to open a window into the brain, charting the effects of cognitive activity on blood flow in the prefrontal cortex (Figure 3).

Optical imaging offers several advantages over other technologies for molecular investigations in small animals and, to a lesser extent, in humans. It's easy to use, a number of highly sensitive benchtop fluorescent probes have the potential for biocompatibility, there is no need for ionizing radiation, and the equipment is relatively inexpensive and rapidly advancing on the coattails of optoelectronic technology in general.

"This is why I have espoused optical imaging. It is an eminently practical field," Chance said.

Although optical imaging encompasses diverse techniques (see accompanying story) and makes use of various wavelengths of light, a great deal of excitement in molecular research lies in the use of tomographic and fluorescence techniques to image living tissues with near-infrared light.

NIR light is essential for examining molecular activity at a depth of more than a few millimeters. That's because in biologic tissues, hemoglobin, water, and fat are least absorbent in the near-infrared spectrum-roughly 650 to 900 nanometers-so photons can penetrate tissue most deeply.

"The name of the game is penetration through the skin and skull into the brain cortex, or through the layers of adipose tissue that may overlie a breast cancer," Chance said. "It is a question of using a color of light that will penetrate the tissue, and that is near-infrared."

Tissue autofluorescence is also at a minimum at NIR wavelengths. Eliminating background signal improves detection.


At its simplest, NIR optical imaging involves the illumination of tissue with a broad beam of light and detection of the resulting spectral or fluorescence data. In addition to surface examinations, planar techniques are useful for simple applications such as detecting the mere presence of fluorescence. Determining on-off genetic expression through use of an activatable probe, such as green fluorescent protein, is one example.

Work at Dartmouth College led by Pogue offers another intriguing example. Pogue is using planar imaging to detect the production of protoporphyrin IX, an endogenous fluorophore that accumulates in metabolically active tissues, particularly mitochondria-rich tumor cells.

Planar imaging yields limited information, however. Penetration depth is only a few millimeters. Equally important, it is impossible with planar techniques to accurately localize or quantify contrast materials because photons do not travel in ordered, linear paths through the body. Instead, they scatter wildly, skewing interpretation of the actual location or distribution of their source.

"To exploit the true potential of optical imaging you need to get below the surface," Pogue said. "Research is focused on how to get light signals out of thick tissues. It's a very difficult problem."

In Pogue's lab and in a handful of others across the country, researchers are using a combination of advanced-photon and tomographic techniques to overcome the limitations of conventional optical imaging. Light is delivered not as a continuous wave but, instead, either in short pulses (time-resolved imaging) or with rapidly oscillating intensity (frequency-domain imaging) at multiple points around the target tissue. It is then collected by detectors that also surround the tissue.

The process is repeated at various wavelengths that correspond to the absorption peak of intrinsic chromophores, such as oxy- and deoxyhemoglobin. In the case of extrinsic fluorophores, absorbed light is almost immediately re-emitted at a longer wavelength; therefore, both excitation and emission light are collected. Data are then reconstructed according to complex mathematical models that account for the diffuse propagation of photons in living tissue.

Together these advances enable better penetration of light (to as deep as 15 cm in breast tissue) and independent characterization of the absorption and scatter of photons. The result is improved resolution (typically a few millimeters, but less than 1 mm in one recent study) and better discrimination of normal and abnormal tissue. Perhaps most important, advanced imaging techniques produce an accurate, quantitative, 3D assessment of the target tissue.

Improving Quality

To improve reconstruction quality, Vasilis Ntziachristos, Ph.D., and his colleagues at Massachusetts General Hospital have taken a further step. They have developed a system that replaces the standard fiber-optic array that delivers and detects light through direct contact with the skin, substituting a noncontact charge-coupled device. A similar system is being used at the University of Pennsylvania.

"With fiber-based systems, it is difficult to get the number of measurements you need to do high-quality reconstructions," said Ntziachristos, who directs the hospital's bio-optics and molecular imaging laboratory. "How many fibers can you bring in contact with tissue?"

Promising as optical tomography is when used with intrinsic contrast, cancer researchers are looking toward a brighter future as targeted fluorescent probes become available for human use, Ntziachristos said.

Of the many commercially available fluorescent agents, only indocyanine green (ICG) is approved for use in humans. A nontargeted blood pool agent, it is useful for tracking blood flow and gauging both the extent of angiogenesis in tumors and the characteristic "leakiness" of their chaotic blood vessels. In small animals, however, many more fluorescent options are available.

Chance's team, for example, is testing the utility of the fluorescent glucose compound pyropheophorbide 2-dexoyglucosamide (Pyro-2DG) to detect hypermetabolism in rat glioma tumors. There are targeted fluorescent reporter probes that home in to the genetic signatures of tumor cells, such as the somatostatin or folate receptor. Ntziachristos and his fellow researchers are studying fluorescent probes that not only are highly specific for certain proteins or proteases, but are activated by interaction with their targets. One example is a probe targeted to cathepsin B, an enzyme that is commonly upregulated in cancer cells. The probe is dark until a portion of its carrier molecule is cleaved by a cathepsin, an action that sends out a bright fluorescent signal. Using another probe targeted to phosphatidylserine, cyanine-labeled annexin V, the Boston researchers are testing whether the effectiveness of chemotherapy can be gauged by imaging apoptosis.


Although light penetration and spatial resolution are improving, they remain challenges for NIR optical imaging. Experts doubt that the technology will top the 15-cm penetration achieved in breast tissue, and they believe it will not even come close to that in blood-rich organ and muscular tissue, where light absorption by hemoglobin is a more critical factor. This means that NIR optical imaging is expected to have tailored, selected uses in humans. In small-animal research, however, the future is wide open.

Fluorescence molecular imaging is a boon to drug research and development, for example. When used with small animals, it enables visualization of the molecular distribution and action of drugs long before anatomic or functional effects can be observed. Multispectral techniques make it possible to examine more than one molecular process at a time. Equally important, fluorescence imaging will make animal research more efficient and cost-effective because it will reduce the need to sacrifice animals in order to assess drug efficacy.

As for limitations in spatial resolution, physiological information-not anatomic detail-is optical imaging's strength, according to Chance. With this is mind, the Penn group pioneered the coregistration of MR and optical images, combining the spatial resolution of MR and the spectroscopic information of optical imaging.

The approach has widespread appeal. Ntziachristos, for instance, has combined MR with planar and fluorescence-mediated optical tomography to improve visualization of tumors in mice (Figure 4). Pogue and his colleagues have integrated an NIR optical breast imaging system into a clinical MR scanner. Data are acquired by both systems during a single scan, and spectroscopic data on hemoglobin concentration are then superimposed on the MR T1-weighted images. This approach not only compensates for shortcomings in the spatial resolution of optical tomography, it improves MR breast imaging as well.

"The hemoglobin signature in the tumors could add a level of specificity that would improve the ability of MR to characterize tumors," Pogue said.