In 6 years, the database has collected 9,000 colonoscopy images. So-
called convolutional neural networks accurately localized polyps 98%
of the time, says Dr. Karnes. He's now working to get the technology
running in real time while colonoscopies are performed. "AI could
make lower-performing colonoscopists who have pattern recognition
difficulties suddenly find more polyps," says Dr. Karnes. "That could
help close the gap between ADR and adenoma prevalence, and hope-
fully put an end to colon cancer."
Eliminating infection risk?
You've seen the headlines and read the reports of antibiotic-resistant
bacteria outbreaks linked to difficult-to-clean duodenoscopes. The
FDA recently issued written warnings to the 3 makers of duodeno-
scopes for failing to complete studies to determine if reprocessing
staffs were able to properly clean and disinfect their devices. The dan-
gers of cross-contamination aren't limited to duodenoscopes. In
January 2016, Baystate Noble Hospital in Westfield, Mass., notified
293 patients that they might have been exposed to bloodborne
pathogens after it was discovered that colonoscopes used at the facili-
ty during a 10-month period weren't properly disinfected.
There's talk of sterilizing endoscopes between uses, but single-use
devices could limit infection risks and eliminate reprocessing's costs
and concerns. First, though, Dr. Traverso calls for a cost-benefit analy-
sis that incorporates the different factors involved in adding dispos-
able scopes. As with all advances designed to make it easier to per-
form more precise screenings, "the key," says Dr. Traverso, "is to find
the right fit for maximizing the impact on patient care at a reasonable
cost."
OSM
A P R I L 2 0 1 8 • O U T PA T I E N T S U R G E R Y. N E T • 6 5
