Detecting Cancer With Light

Shining some light on skin lesions and measuring the reflection may help dermatologists identify the most dangerous form of skin cancer more easily, according to a new study. The technique could help pick up the cancer in its early stages and save patients from needless biopsies.

Melanoma, the most aggressive form of skin cancer, first shows up as a darkly pigmented patch on the skin, But it's difficult to tell such a malignant lesion from a harmless mole or freckle. Dermatologists can have a computer compare subsequent images and look for suspicious patterns, but they usually rely on telltale signs like ragged edges or multiple colors. Even specialists can have trouble detecting melanoma, however, and most would like to spare patients the stress of a biopsy when possible.

In 1991, an Italian team developed a new technique. When they shone a beam of light on skin lesions, they found that the spectra of reflected light were different for melanomas and benign skin lesions. Now a research team from the Institute of Cancer Research and the Royal Marsden Hospital, United Kingdom, has taken that technique a step further.

Physicist Vincent Wallace and his colleagues first measured the spectra from 121 "worrisome lesions," which were subsequently removed and diagnosed in the lab. They found seven key differences between the spectra of benign lesions and melanomas, which they used to develop software that analyzes the signal and judges whether the lesion is likely to be a melanoma or not. They incorporated this into an experimental scanner, which they tested on 47 lesions. As the team reports in the March Physics in Medicine and Biology, they could tweak the system so that it could pick out every malignancy, yet produce few false alarms. The system did better than the current computerized image analysis, and as good as visual inspection by most dermatologists, says physicist and team member Jeffrey Bamber. The method is also quicker and could examine smaller lesions. And because it's completely automated, it could be used by nondermatologists, such as general practitioners, says Bamber.

But Ronald Barr, a dermatologist at the University of California, Irvine, says he would be wary to use the technique to routinely screen patients--at least in the U.S.--because he is afraid it might fail to detect some malignancies. "We have such a litigious society," says Barr. Instead, he says the system may be handy to screen many lesions on a single patient known to have melanoma, or to follow up patients after treatment.