Diagnosing disease with spectral imaging

Gary Boas

Spectral imaging holds promise for a range of diagnostic applications. It could be used to differentiate diseased from healthy tissue based on intrinsic spectral signatures, obviating the need for contrast agents. For example, researchers at Cedars-Sinai Medical Center in Los Angeles recently explored the possibility of using it for cancer detection, demonstrating the ability to identify cancerous tissue in both animal and human samples.

These results underscore the potential of spectral imaging for intrasurgical guidance, which can contribute to decision making during surgery by assessing tissue status in real time. Investigator Daniel L. Farkas recently noted, however, that, to assess its efficacy for such applications, it would be best to start with “simpler problems” than cancer.

Using spectral imaging, researchers have demonstrated differentiation of normal and diseased tissue in a mouse model of Hirschsprung’s disease. Shown here are a black-and-white image of a portion of a bowel in a mouse (top) and a spectral image (bottom) with a mean-square-error classification of individual pixels in the spectral image stack. In the latter image, blue represents normal/blood; green, aganglionic (diseased) regions; and red, a transition region. Reprinted with permission of the Journal of Biophotonics.

For this reason, Farkas and colleagues have reported using this technique to determine the presence of Hirschsprung’s disease in mouse models. This disorder is a congenital absence of ganglion cells that affects primarily the lower portion of the colon. Farkas explained that assessment of the disease requires only a binary type of imaging — straightforward differentiation of normal from abnormal tissue. Thus, Hirschsprung’s disease offers a simpler model with which to test the effectiveness of spectral imaging.

The researchers performed surgery on and acquired in vivo spectral image sets of the lower colon of control mice as well as mice that were mutant models of Hirschsprung’s disease. The data sets were obtained with an Applied Spectral Imaging Sagnac interferometer-based spectral imaging system and with an acousto-optic tunable filter-based system, developed in-house. The investigators extracted a characteristic spectral signature from selected regions of interest, then applied an algorithm they developed to distinguish between normal and aganglionic colon with high spatial resolution and high reproducibility. They demonstrated that they could clearly separate the two, confirming this finding with pathological analysis. Sensitivity was 97 percent, and specificity 94 percent.

They currently are planning clinical trials in humans, to test the efficacy of the technique for intrasurgical decision making regarding Hirschsprung’s disease. Treatment for the disease typically involves a minimally invasive procedure in which the diseased part of the colon is removed and the remaining healthy colon is attached to the anus. The surgeon must be careful, however, to determine precisely where the healthy portion ends and the diseased part begins. If any of the diseased colon remains, the patient might suffer from worsening constipation. If too much of the colon is removed, however, chronic diarrhea and other major health problems might develop.

Currently, surgeons make the determination by collecting samples from the colon wall and sending them to the laboratory for pathological analysis. This process can be time-consuming, however, requiring the patient to remain under general anesthesia for as much as an hour. “There’s also a fair amount of subjectivity,” said Dr. Philip K. Frykman, part of the research team. In contrast, spectral imaging affords immediate, objective determination of what is healthy and what is diseased colon, thus eliminating the need for pathological analysis and reducing the amount of time in the operating room.

Journal of Biophotonics, published online Feb. 21.