Developing standards for hyperspectral imaging
Currently, when doctors want to see if or how well a wound is healing, they often have to carry out a biopsy (or a similar invasive technique). These techniques can, however, really only provide information on a small area, as well as injuring an already injured patient.
But there is a technology that can offer doctors a noninvasive and painless way to discriminate between healthy and diseased tissue, and can also reveal how well damaged tissue is healing over a wide area. This technology is known as hyperspectral imaging. However, there is a catch involved in hyperspectral imaging; namely, a lack of calibration standards that is impeding its use.
The human eye, as well as consumer digital cameras, only sees red, green and blue light (a relatively narrow portion of the electromagnetic spectrum). However, each pixel of a hyperspectral image captures information for hundreds of narrow spectral bands, from the ultraviolet to the infrared.
Researchers at the National Institute of Standards and Technology (NIST) have begun gathering data on how human skin looks under various wavelengths of light, with the aim of using this data to develop the standards that are needed for hyperspectral imaging.
NIST researcher David Allen describes how being sensitive to so many wavelengths means hyperspectral imagers can see many different things that humans can’t see, including the amount of oxygen in human tissues, an indicator of healing.
“The potential of the technology has been proven, but the problem is that researchers are simply lacking a way to assure consistent results between labs,” explained Allen. “Standards development has itself been hindered by a lack of human skin reflectance data, especially in the ultraviolet and short-wave infrared.”
Catherine Cooksey, the project leader for the spectrophotometry program that establishes and maintains the national scale of reflectance, explains how we need to determine what so-called ‘normal’ tissue looks like, before we can explore what diseased tissue looks like hyperspectrally.
Furthermore, she says that they are looking to quantify the variability both within an individual and between individuals due to inherent biological differences. The initial NIST studies used 28 volunteer test subjects. The data collected included a photograph of the test area on the subject’s forearm and three reflectance measurements of the test area.
“Skin reflectance varies due to skin pigmentation, tissue density, lipid content and blood volume changes,” describes Cooksey. “And few, if any, studies of skin reflectance have been done with an estimated measurement uncertainty that is traceable to NIST or any other national metrology institute. We need good data from a wide variety of sources, and for that we need the help of our colleagues in the community.”
Once they collect enough data, the NIST researchers can feed it into NIST’s Hyperspectral Image Projector, a device that creates hyperspectral scenes that have all the spectral signatures of the real thing – in this case, tissue in various stages of repair. Medical imaging technicians can then use these ‘digital tissue phantoms’ to test their imagers’ ability to discern among and detect different tissue types and conditions.
Source: Seeing your true colors: standards for hyperspectral imaging.
