What factors affect liver metastasis detection performance?

Thursday, December 2 | 8:00 a.m.-9:00 a.m. | SSPH14-2 | Room S402
Image navigation patterns and subspecialization was associated with better performance of abdominal CT scans to detect liver metastasis, according to this scientific presentation.

In a study led by presenter Scott Hsieh, PhD, from the Mayo Clinic in Rochester, MN, 25 readers -- including nine subspecialized abdominal radiologists, five nonabdominal subspecialists, and 11 residents/fellows -- interpreted 40 portal phase abdominal CT exams with challenging lesions on a custom workstation.

Abdominal specialists produced a mean area under the jackknife-adjusted free receiver operating characteristic (JAFROC) curve of 77%, sensitivity of 85%, and false-positive count of 57 across all 40 exams, compared with mean JAFROC of 69%, sensitivity of 83%, and false-positive count of 56 for nonabdominal staff. Trainees had a mean JAFROC of 71%, sensitivity of 83%, and false-positive count of 60. The differences between all the readers groups were statistically significant (p < 0.01).

Although they didn't appear to be better at detecting possible liver metastases, abdominal subspecialists also exhibited better judgment at rating risk of malignancy once the metastases were detected, according to the researchers.

In other findings, time in liver windows, time spent looking at coronal images, and total number of marks correlated with improved sensitivity (p < 0.05). Subspecialized training or the use of zoom was not significantly correlated with improved sensitivity, however.

Subspecialized training and practice correlated with improved JAFROC scores. After performing clustering analysis, the researchers were also able to identify groups of metastases that were better or not as well detected by abdominal subspecialists.

"Certain patterns of image navigation are associated with higher sensitivity in the task of liver metastasis detection," the authors wrote. "Subspecialists are able to attain higher performance through better judgment of malignancy probability."

Check out this Thursday morning talk to learn more.

Page 1 of 603
Next Page