"You have lung cancer." Each year, 420,000 people in North America, Japan,
and Europe receive that frank, four-word diagnosis. Eighty five to ninety
percent of them don't survive the next five years; most succumb in just a few
months. An estimated one million people worldwide--160,000 in the U.S.
alone--die from the disease annually. Though the second most common cancer, it
is by far the most efficient killer.
Despite great effort, the cure rate for lung cancer in the U.S. has risen from 5% to just 13% in 30 years. The disease's lethality, however, lies not in a special virulence, but in the difficulty of diagnosis. By the time the oncologist spots the suspect lung spot, the cancer is typically too far along to be stopped.
Continuing to place great emphasis on tackling the disease at its later stages is inappropriate, says Branko Palcic, a senior scientist at the British Columbia Cancer Agency and scientific advisor at Xillix, a Vancouver-based developer of medical imaging devices. "It's like firefighters making better engines, better hoses, and better hydrants," he asserts, "but then always arriving at the fire when the house is half gone."
Instead, Palcic and a team of researchers and engineers set out to develop an early-detection method for cancerous and pre-cancerous lung tissue. Their work culminated in the creation of the Lung Imaging Fluorescence Endoscope, referred to as XillixLIFE Lung. In the U.S., the system received two patents in the first four months of 1996 and is under review for approval by the FDA.
The new imager attaches to a standard white-light bronchoscope, a factor that simplified the engineering and, hopefully, will ease government approval of the device. Both LIFE and the white-light systems employ the same fiber-optic endoscope--the several-foot long, pencil-thin rod that is threaded into the patient's bronchi to illuminate and capture video images. But the two devices differ dramatically in the wavelength of light used for illumination and in the type of return signal captured for viewing.
LIFE exploits the phenomenon of tissue autofluorescence. When stimulated with UV or near-UV, lung cells emit their own spectrum of light that begins at a wavelength of about 480 nm and extends into the red region past 700 nm. In this case, the actual illuminating source is an industrial, 150-mW helium-cadmium laser--converted for medical duty--that produces a deep-blue light at 442 nm. Though actual UV would be superior for stimulating fluorescence, it also happens to be a wavelength known to damage tissue.
Unlike the reflectance of white light--a surface effect--fluorescence "excites the molecular structures, and they become like little light bulbs," explains Bruno Jaggi, Xillix's VP and chief engineer.
More importantly, healthy tissue and cancerous or pre-cancerous tissues radiate quite differently. Both emit similar quantities of red light, but healthy tissue pumps out roughly eight times more green light.
Using this information to identify suspect lung tissue is the challenge. The level of fluorescence measures just 1 picowatt/mm,2. With roughly 20 mW of blue laser light illumination resulting in 64- mW/mmirradiance at a typical distance of 1cm, the ratio of input to output is about 64 million to one.
The endoscope gathers this dim irradiance and transmits it along 10,000 to 20,000 optical fibers (depending on endoscope brand) to an image acquisition and display module. There, the light is split spectrally into red and green color bands by a dichroic mirror, further defined by red and green band-pass filters, and amplified by a factor of 5,000 to 10,000 with a pair of image-intensified cameras that trace their heritage to military night-vision equipment. Each camera captures the corresponding red or green monochrome scene on a 512 512- pixel CCD and encodes it as a video signal. These signals are then manipulated and compared by a proprietary image-processing card running in an industrial 486-class PC. The resulting output routes to a color video monitor that displays healthy lung tissue in green and diseased tissue in red.
|∑ Image intensifiers derived from night-vision equipment
∑Real-time imaging board and proprietary processing algorithm
medical application industrial HeCd laser
It would seem that detecting cancerous regions would be as simple as looking for low levels of green light. "That would be nice, but it doesn't work," says Jaggi. Light intensity varies as an inverse function of the square of the distance to the radiant source. Since both the blue illuminating light source and the fluorescence obey this physical law, the actual intensity of fluorescence is 1/r4, where r is the distance between the tissue and the tip of the endoscope. So without knowing the distance to every part of an illuminated scene, there is no way to tell apart tissues that emit abnormally low levels of green light from those that are simply further away.
The solution is to use the red emission band--which is similar in intensity for both healthy and unhealthy tissue--to normalize the reading of the green channel. Areas fluorescing proportionally less green light than red then appear suspect regardless of their distance from the endoscope.
At first, engineers designed an imaging system that normalized the signals by computing the ratio of green light to red light at each pixel location on the two CCDs. This method involved roughly 250,000 division operations per video frame. And given the limited computing power of a PC, it proved impossible to calculate 30 frames every second and maintain a real-time image. Ultimately, Xillix, in collaboration with researchers at the British Columbia Cancer Agency, developed a proprietary algorithm--patent pending--that functions like a ratio but can be computed more quickly, in real-time, at video rates.
Today, clinicians routinely use white-light bronchoscopy to examine lung tissue for cancer. Yet the medical journal Chest describes the method as, "a very poor technique for visualizing bronchial dysplasias and early carcinomas such as carcinoma in situ (CIS)." Fewer than 30% of CIS lesions can be identified positively with white light. The alternative, X-rays, is even worse, spotting actual cancers only after they've grown to 1 cm or more in diameter.
LIFE, on the other hand, can identify lesions as small as 1 mm, with detection rates for CIS as high as 91.4% and a specificity of 92%. "We are detecting about three times more early cancer lesions (than with white-light bronchoscopy alone)," says Jaggi.
Discovering autofluorescence in lung tissue involved a bit of serendipitous research and engineering. In the late 1980s, a team including Palcic, Jaggi, and clinician Dr. Stephen Lam began their investigation into early cancer-detection by imaging the traces of photosensitive drugs injected into patients. The drugs are known to collect in tumorous tissue, and the thought was that by shining UV down the bronchoscope, the researchers could spot the glow of drug-rich cancerous areas.
|∑ Low-light vision systems
∑ Video and computer image-processing devices
∑ Early cancer detection for other
The method presented many drawbacks. Following treatment, patients wouldn't be able to go out in the sun--or even expose themselves to bright light--for several weeks or more. It was prone to a fairly high degree of false-positive readings. And, "if you have to take so much as an aspirin for diagnosis, it's going to be an issue (with obtaining government approval)," says Palcic. Attempting to limit these disadvantages, researchers progressively reduced the amount of photosensitive drugs until, "to our surprise, once the dose was zero, we still got a reasonable signal," Jaggi says. It cut five to 10 years from the approval process.
Efforts to create the current LIFE-Lung imager sent engineers down several developmental cul-de-sacs, as well. Jaggi looked at using a single camera to capture and compare a sequence of alternating green and red images filtered with a rotating color wheel. Yet due to the elapsed time between the two images, even slight movements of the endoscope caused subtle differences in scene content that caused unacceptable artifacts.
He also developed a prototype system that used a beam splitter to send exact copies of the captured light, passed through either a red or green filter, to two cameras. However, splitting the beam instead of the spectrum, as with the dichroic mirror, cut the already minuscule light level in half.
Today, LIFE-Lung finds application in Canada and parts of Europe and Asia. There, patients who exhibit signs of early cancer--verified by biopsy--can sometimes have those lesions simply scraped off. In the US, several clinics and hospitals are performing investigational studies to provide the FDA with results to be used in the approval process.
Not content with applying LIFE solely in the lung, Xillix has entered a collaborative development and marketing agreement with Japan's Olympus--the largest manufacturer of endoscopes and accessories--to investigate applications in the gastrointestinal tract. Jaggi notes that other LIFE devices may someday aid with the diagnosis of cancer in the bladder and pharynx/larynx.
Competition? "I'm not aware of anybody (else) doing early cancer detection without drugs," claims Jaggi.