In vivo diagnosis and UC: A good fit?
Patients with ulcerative colitis (UC) are at increased risk of colorectal cancer, and current guidelines recommend periodic surveillance colonoscopy with random and targeted biopsies. But because most lesions aren't dysplastic, standard surveillance results in a large number of tissue samples with very low yields.
Michael F. Picco, M.D., of Mayo Clinic in Florida, says probe-based confocal laser endomicroscopy (pCLE), offers an alternative to traditional biopsy and histopathology.
He explains, "pCLE provides high-definition images of the gastrointestinal mucosa in real time, allowing in vivo visualization of the histology of the mucosal epithelium."
Although in vivo diagnosis shows great promise for reducing unnecessary biopsies and improving early detection of dysplasia, its use in UC is complicated by chronic inflammation of the surrounding tissue.
To assess the clinical applicability of pCLE for in vivo diagnosis of dysplasia in patients with UC, Dr. Picco and colleagues first conducted a study to determine the accuracy of the technology and whether novice observers could agree on criteria for the detection of precancerous and cancerous growths.
To accomplish this, 25 patients undergoing surveillance colonoscopy had pCLE imaging of polyps, which were identified using white light or chromoendoscopy. This was followed by matching tissue sampling for histopathology. Three novice observers then reviewed the recorded pCLE videos offline, identifying a total of 61 lesions (nine dysplastic, 15 hyperplastic and 37 pseudopolyps). All observers identified eight of the nine abnormal growths, suggesting that they could diagnose and differentiate dysplasia with reasonable accuracy. But, Dr. Picco says, there was also a tendency to overcall.
"In vivo diagnosis is challenging because the images, which cover a very small area, are running by in real time and are hard to interpret. UC increases the difficulty because there is so much background information. We wondered whether this technique would work for general GI physicians who are not trained pathologists, and we tried to see if there might be a better way of allowing novice observers to discriminate between normal and abnormal tissue," he explains.
From video to stills
So rather than watching recorded videos, Dr. Picco and his colleagues asked observers to interpret computer-generated summary mosaic images, believing this would aid diagnosis. But in fact, the mosaic scheme produced worse results.
"Unlike run-of-the-mill polyps, the computer couldn't synthesize the images because of the background noise, so it was unclear whether this technique had any validity in looking at lesions in patients with UC," Dr. Picco says.
The next step was to add more mosaic images — 150 of them. These were given to experienced investigators for comparison. But even these observers did not fare well, identifying precancerous lesions just 50 percent of the time.
"Apparently learning isn't the issue," Dr Pico says. "Although investigators performed well with video images, results were poor with mosaic images because they did not provide the essential information needed to make a diagnosis." The background in people with UC stands in stark contrast to that in people without bowel disease; the inflammation makes everything so much more difficult."
He adds, "We have this fancy technology and it would be nice if any endoscopist could look at a simple image generated by the computer and use it. But we need to make the technology more user-friendly and more useful."
He notes that mosaics can be made by the observer or by the computer, but the observer mosaic is biased. "We want the computer to put the summary image together. The next step is to determine how mosaics are generated and whether they can be tweaked. If they can't, the technique isn't useful, but we don't know the answer to that yet."