Per Skaane MD, PhD, Presenter: Nothing to Disclose

Ashwini Kshirsagar, Abstract Co-Author: Employee, Hologic, Inc

Solveig Sand-Hanssen Hofvind, Abstract Co-Author: Nothing to Disclose

Gunnar Jahr, Abstract Co-Author: Nothing to Disclose

Ikuyo Endo, Abstract Co-Author: Nothing to Disclose

Ronald Augustus Castellino MD, Abstract Co-Author: Employee, Hologic, Inc

To determine the standalone performance of CAD on the baseline screening mammograms for (a) cancers missed by one of two readers on the baseline screen and (b) cancers subsequently diagnosed but missed by one or both readers on the baseline screen, in a population based, independent double reading mammography screening program randomized to SFM and FFDM (Oslo-II study).

23,929 women aged 45-69 underwent standard 4 view screening mammograms. The prospective independent double readings using a 5-point rating scale for probability of cancer were compared with the results of retrospectively applied CAD (R2 ImageChecker V8.3, Hologic) on 207 baseline screening examinations. There were 103 cases of screen detected cancers (64 SFM and 39 FFDM) and 104 (80 SFM and 24 FFDM) cases of subsequently detected cancers (interval cancer or cancer diagnosed on the next screening round). The baseline examinations of women with subsequent cancers were classified as normal, minimal sign non-actionable, minimal sign actionable, and false negative (“missed”). Minimal sign actionable and false negative exams should have been recalled for assessment if detected by the reader.

One of the two independent readers missed 16 of the 64 SFM screen-detected cancers; CAD correctly marked 15/16 (93.8%). Of the 39 screen-detected FFDM cancers, 8 were missed by one reader; CAD correctly marked all 8 (100%). For the 104 subsequent cancers, CAD correctly marked 18/20 (90%) of the baseline SFM and 5/5 (100%) of the baseline FFDM classified as minimal sign, actionable or false negative, which had been prospectively read as normal by one or both readers.

Visible and actionable cancers were prospectively overlooked by one of the two readers at baseline screening in 23.3% (24/103) and by one or both readers for subsequently diagnosed cancers in 27.9% (29/104). Although comparable, CAD correctly marked more of the cancers missed on FFDM (100%) than on SFM (90-94%).

CAD correctly marked between 90-100% of cancers missed in an independent double reading breast cancer screening program. The potential benefit of CAD in this setting is comparable for SFM and FFDM.

Skaane, P, Kshirsagar, A, Hofvind, S, Jahr, G, Endo, I, Castellino, R, Breast Cancer Screening with Independent Double Reading using Screen-Film Mammography (SFM) and Full-Field Digital Mammography (FFDM): Standalone Performance of Computer-Aided-Detection (CAD) on Missed Cancers.  Radiological Society of North America 2008 Scientific Assembly and Annual Meeting, February 18 - February 20, 2008 ,Chicago IL. http://archive.rsna.org/2008/6011775.html