Spectral Imaging Method Could Improve Pathologist’s Ability to ‘See’ Cancer Cells

PITTSBURGH—A special microscope capable of “seeing” some 40 different wavelengths of the visible color spectrum is being used to add dramatic new colors to stained tissues, which may enhance the pathologist’s ability to detect malignancies.

“As a pathologist, I am very excited by this new technique because we can distinguish patterns that couldn’t be seen by the eye,” said Richard Levenson, MD, a specialist in spectral imaging at Carnegie Mellon University. He noted that pathology has a 150-year history of correlating patterns seen on slides with disease. “Now we have a whole new set of patterns.”

In an interview with Oncology News International, Dr. Levenson explained that our eyes have only three color receptors. Within the visible range of light—wavelengths from about 400 to 700 nm—we perceive everything at about 400 to 500 nm as blue, 500 to 600 nm as green, and 600 to 700 nm as red. All the various shades of color that we recognize are basically combinations of those three. “We dump a lot of complex spectral behavior into just three bins,” he said.

Spectral classification breaks these three color bins down into 10 or 15 subdivisions each, providing a more detailed image in multiple colors. “We then can look at 40 primary colors (40 different wavelengths), instead of just three,” he said. “It increases the dynamic range of our color vision.”

Using a spectral imaging microscope and specially developed software, Dr. Levenson has obtain detailed spectral images of stained tissue samples that highlight cancer cells (see Figure). Panel B of the figure shows a spectral image in which each pixel is classified and assigned a pseudocolor based on its similarity to the classification spectra. Cancer cells are spectrally identifiable (in red in panel B) even in a complex and visually confusing background. “Thus,” he said, “these spectrally classified images reveal very small differences that are reproducible and distinguishable.”

Dr. Levenson and his colleague Daniel Farkas, PhD, are in the early stages of developing the technology necessary to make spectral imaging applicable to pathology. He noted that introduction of the technique into clinical practice will require a major investment in studies to correlate spectral findings with biologic findings.

To some degree, he said, “spectral instruments have been built because they can be, not necessarily because they have a real world application yet.” But if spectral imaging is a technology “looking for a home,” that home may be pathology. “Pathologists with vision who’ve seen it are very excited by its possibilities,” Dr. Levenson commented.

One particular avenue of research that could yield a practical application of the technique, he said, is automated vision analysis, which would employ the same kind of computer-automated technology currently being used to review Pap smears.

“Most of these automated instruments, surprisingly, don’t use any color information, just things like texture, density, and various morphologic measures,” he said. “They throw out the color information entirely, so this represents a whole new dimension for computer-automated technology.”