DOI: 10.1148/radiol.2211001507
Blinded Review of Retrospectively Visible Unreported Breast Cancers: An Eye-Position Analysis1
Calvin F. Nodine, PhD,
Claudia Mello-Thoms, PhD,
Susan P. Weinstein, MD,
Harold L. Kundel, MD,
Emily F. Conant, MD,
Rose E. Heller-Savoy, MD,
Susan E. Rowlings, MD and
Julia A. Birnbaum, MD
1 From the Pendergrass Laboratory (C.F.N., C.M.T., H.L.K.) and Department of Breast Imaging (S.P.W., E.F.C., R.E.H.S., S.E.R., J.A.B.), University of Pennsylvania, Philadelphia. From the 2000 RSNA scientific assembly. Received September 8, 2000; revision requested October 26; final revision received January 31, 2001; accepted February 15. C.F.N. supported in part by DAMD17-97-1-7130 grant. Address correspondence to C.F.N., Department of Radiology, 3600 Market St, Suite 370, Rm 103, Philadelphia, PA 19104-2647 (e-mail: nodine@oasis.rad.upenn.edu).
View larger version (133K):
[in a new window]
|
Figure 1. A mammographer scanning the display workstation containing CC (left) and MLO (right) images of one breast. Eye position was monitored during the initial interpretation phase with an eye-head tracker. The eye-head tracker was worn on a headband by the viewer. It contained an infrared light source that imaged the viewer’s left eye and a television camera that picked up limbus-reflection changes from the cornea as the eye scanned the mammographic display. The eye-position data were sent to a control unit for analysis to determine the axis of gaze. The headband also contained a magnetic sensor that picked up changes in the head movement within a magnetic field produced by a transmitter located behind the viewer. Eye-head position data were integrated to determine the location of the axis of gaze on the mammogram.
|
|
View larger version (91K):
[in a new window]
|
Figure 2. An eye-position record of an experienced mammographer scanning a cancer-containing mammogram (prospective case). The first 16 seconds of scanning are shown. Fixations are indicated by small circles (not shown to the viewer) connected by lines that create the eye-fixation scan path. The malignant calcifications are identified on the CC (left) and MLO (right) views by large circles with thin lines. On the CC view, the initial fixation started within the circled lesion area. Fixations tended to cluster within the circled areas, indicating that both lesions attracted the viewer’s attention early in the scanning.
|
|
View larger version (101K):
[in a new window]
|
Figure 3. The same cancer-containing mammograms (CC view on left and MLO view on right) scanned by the same viewer as in Figure 2 show the entire eye-fixation scan path, which took 29 seconds. Long-dwell fixation clusters (>1,000 msec) are indicated by large circles with thick lines. In addition to the two long-dwell fixation clusters generated earlier, as indicated in Figure 2, two additional long-dwell fixation clusters were generated in new areas before the viewer made her final decision.
|
|
View larger version (99K):
[in a new window]
|
Figure 4. Mammograms obtained in the same case as in Figures 2 and 3, but in addition to the eye-fixation scan path and the cancer indicated by large circles with thin lines, the zoom locations are depicted by squares on both views. Note that the zoom locations selected by the viewer fall directly over the true cancer locations and that these zooms coincide with the large groupings of fixations on the CC (left) and MLO (right) views. Note also that the first (and only) zooms were on the lesion on both views.
|
|
View larger version (106K):
[in a new window]
|
Figure 5. Mammograms give an eye-position record of the same experienced mammographer as in Figures 2-4 scanning CC (left) and MLO (right) views of a cancer-free case (left breast). Fixations are small circles connected by lines. The starting fixation is indicated by an asterisk on the CC view. The eye-fixation scan path is extensive over both views and shows 46 fixation clusters in 36 seconds of scanning. The long-dwell fixation clusters (>1,000 msec) are indicated by large circles with thick lines. Attention areas containing long-dwell fixation clusters indicate where visual attention was presumably focused on normal-image perturbations.
|
|
View larger version (107K):
[in a new window]
|
Figure 6. Mammograms obtained in the same case as in Figure 5 show the entire eye-fixation scan path and all zooms, which are indicated by squares. There were five zooms on the CC view (left) and seven zooms on the MLO view (right). In this case, the overlap between fixations and zooms is almost 100%. Note how much greater the coverage is for this cancer-free case than it was for the malignant case in Figures 2-4.
|
|
Copyright © 2001 by the Radiological Society of North America.
