Medicsight study shows high colon cancer detection with ColonCAD
Medicsight, a provider of computer-aided detection (CAD) and image analysis software, presented data showing that its ColonCAD helped radiographers detect 100 percent of cancers in routine primary reporting of virtual colonoscopy (VC) at the European Society of Gastrointestinal and Abdominal Radiology (ESGAR) in Istanbul, Turkey, in mid-June.

“These findings are exciting because whilst demand for colorectal cancer investigation is on the rise, there is a shortage of radiologists trained to report VC in many countries," said David Burling, MD, consultant radiologist from St. Mark's Hospital in London. “If radiographers can be trained to interpret scans more effectively using CAD technology, both radiology teams and patients will benefit, with the ultimate goal of decreasing the incidence and morbidity of colorectal cancer, the second most common cause of death from cancer in Europe.”

In the study, 304 patients (61 percent female) underwent VC as part of routine clinical care. Two trained radiographers read the exams, supplemented by CAD as a ‘second reader.’ Radiographers recorded colonic abnormalities and interpretation times and proposed one of six different codes for each exam (inadequate/normal/6-9 mm polyp/10 mm+ polyp/cancer/diverticular stricture).

The findings were compared to that of an experienced radiologist and supplemented by colonoscopic findings if available, Medicsight said.

The company said the radiographers, assisted by CAD, detected 100 percent (17/17) of cancers; 72 percent (21/29) of 10 mm+ polyps and 67 percent (42/63) of 6-9 mm polyps. Overall, the agreement between radiographer and radiologist examinations was good, demonstrating that use of CAD by trained radiographers provides an effective triage strategy in initial patient management following VC, Medicsight said.

Notably on retrospective analysis, CAD had actually detected 60 percent of polyps missed by radiographers. Therefore, had radiographers accepted these correct CAD annotations, their detection rates would have been 83 percent and 89 percent for medium and large polyps respectively, according to the researchers.

The study “confirms the CAD's accuracy in a user group (radiographers) that will play a vital role in any colorectal screening program being implemented in the world today,” said David Sumner, CEO of Medicsight.