Radiology: Colon CAD improves reader sensivity even for small polyps
Magnified colon polyp via CT colonography is shown in 3D.
Image Source: David J. Vining, MD, University of Texas MD Anderson Cancer Center in Houston
Colon computer-aided detection (CAD) improved reader sensitivity per segment, per patient and per polyp for small polyps and adenomas, while slightly reducing specificity, according to a study published in the September edition of Radiology.

Although stand-alone studies have indicated CAD’s sensitivity, readers studies are required “to demonstrate the practical value of CAD,” explained Abraham H. Dachman, MD, of the department of radiology at the University of Chicago, and colleagues.

Dachman and colleagues sought to expand on previous multi-reader CAD studies with small patient cohorts or small numbers of readers and assess the effect of CAD on CT colonography (CTC) accuracy when CAD is used as a second reader.

The researchers collected 100 colonoscopy-proved cases with an average patient age of 57.9 years from the Naval Medical Center San Diego and Walter Reed Army Medical Center in Washington, D.C.

Five radiologists with more than 500 cases of CTC experience finalized the classification of each patient finding as positive, negative or excluded from the study. Among positive cases, only those with polyps visible in retrospect were considered in order to differentiate CAD performance from CTC.

Nineteen readers interpreted the 100 cases, which included a stratified random sample of patients with positive findings with at least one large polyp (?10 mm), patients with positive findings with only small polyps between 6 to 9 mm, and patients with negative findings. The final group of cases included 17 patients with polyps of 10 mm or larger, 35 patients with polyps between 6 to 9 mm and 48 patients with negative findings.

Each reader interpreted 50 cases with CAD and 50 without CAD in two sessions separated by 27 to 58 days and used their typical clinical reading mode of primary 2D or primary 3D. They rated each polyp according to a 100-point confidence scale.

“The primary objective was to measure the mean change in the readers’ area under the ROC curve (AUC) from the AUC for reading without CAD to the AUC for reading with CAD,” wrote Dachman. The average segment-level AUC for the CAD read was 0.758, which was significantly greater than the average AUC of 0.737 for the unassisted read, reported the authors. When the patient was used as measure of analysis, AUC increased from 0.711 without CAD to 0.727 with CAD, which was not significant.

The researchers found improved sensitivity with CAD use at the segment, patient, or polyp level among subcategories of  ?10 mm, 6 mm to 9 mm and adenomatous polyps. Seventy-four percent of readers showed a larger sensitivity improvement for smaller adenomas.

Other study results include a decrease in specificity from 0.984 without CAD to 0.975 with CAD and an average increase of 4.5 minutes in interpretation time with CAD. Patient treatment changed in about 22 percent of patients; reader assessment changed from false-negative to true-positive in 14.1 percent of patients and from true-positive to false-negative in 7.9 percent of patients, according to Dachman and colleagues.

“The readers’ decrease in patient-level specificity from 0.929 to 0.904 is acceptable,” wrote Dachman, who added that “the improvement in sensitivity with CAD, while small, is clinically important.” The researchers summed, “We conclude that the use of CAD results in a significant improvement in overall reader performance.”

Around the web

Eleven medical societies have signed on to a consensus statement aimed at standardizing imaging for suspected cardiovascular infections.

Kate Hanneman, MD, explains why many vendors and hospitals want to lower radiology's impact on the environment. "Taking steps to reduce the carbon footprint in healthcare isn’t just an opportunity," she said. "It’s also a responsibility."

Philips introduced a new CT system at ECR aimed at the rapidly growing cardiac CT market, incorporating numerous AI features to optimize workflow and image quality.

Trimed Popup
Trimed Popup