A powerful color-based imaging technique is making the jump from
remote sensing to the operating room—and a team of scientists* at the National Institute
of Standards and Technology (NIST) have taken steps to ensure it performs as well
when discerning oxygen-depleted tissues and cancer cells in the body as it does
with oil spills in the ocean.
The technique, called hyperspectral imaging (HSI), has frequently
been used in satellites because of its superior ability to identify objects by color.
While many other visual surveying methods can scan only for a single color, HSI
is able to distinguish the full color spectrum in each pixel, which allows it to
perceive the unique color “signatures” of individual objects. Well-calibrated
HSI sensors have been able to discern problems from diseases in coral reefs to pollution
in the atmosphere as determined by the distinct spectral signature at a location.
“Because diseased tissues and cells also have distinct spectra,
scientists have been trying to use HSI for medical applications as well,” says
NIST physicist Jeeseong Hwang. “But any time you tell a machine to scan for
something, you need to be sure it is actually looking for what you want, and you
have to make sure that the image analysis algorithm extracts the correct color information
out of a complex multicolor data set. We decided to create a way to calibrate an
HSI device and to test its algorithm as well.”
Matthew Clarke, a former National Research Council-supported
postdoctoral fellow in Hwang’s group who is currently working in the National Gallery
of Art in Washington, D.C., wrote new software for a device called
a microarrayer, so named because it is capable of laying down hundreds of tiny sample
droplets in specific places on a microscope slide’s surface. Normally a microarrayer
creates DNA arrays for genetic research, but the team remade it into an artistic
tool, programming it to select chemicals of different hues and lay them down on
the slide’s surface.
The results, which look a bit like dot-matrix printing, can be
used to calibrate medical HSI devices and image analysis algorithms. When combined
with HSI in a medical imaging application, this effort could allow a surgeon to
look for cells with a specific chemical makeup, as determined by the cells’ color.
“Scientists and engineers can create a custom slide with
the exact colors representing the chemical makeup they want the HSI devices to detect,”
Hwang says. “It could be a good way to make sure the HSI devices for medical
imaging perform correctly so that surgeons are able to see all of a tumor or diseased
tissue when operating on a patient.”
This project is part of a larger effort to evaluate and validate
optical medical imaging devices, led by the NIST team members, David Allen, Maritoni
Litorja, Antonio Possolo, Eric Shirley and Jeeseong Hwang. Hwang adds that the special
issue** of Biomedical Optics Express in which the team’s findings appear is
the output of a recent NIST-supported international workshop on the topic.
*M.L. Clarke, J.Y. Lee, D.V. Samarov, D.W. Allen, M. Litorja,
R. Nossal and J. Hwang. Designing microarray phantoms for hyperspectral imaging
validation. Biomedical Optics Express, Vol. 3(6), pp. 1291-1299 (June 2012), doi:
** See www.opticsinfobase.org/boe/virtual_issue.cfm?vid=168.