Breast Tomosynthesis: Aspects on detection and perception of simulated lesions

Abstract: The aim of this thesis was to investigate aspects on detectability of simulated lesions (microcalcifications and masses) in digital mammography (DM) and breast tomosynthesis (BT). Perception in BT image volumes were also investigated by evaluating certain reading conditions. The first study concerned the effect of system noise on the detection of masses and microcalcification clusters in DM images using a free-response task. System noise has an impact on image quality and is related to the dose level. It was found to have a substantial impact on the detection of microcalcification clusters, whereas masses were relatively unaffected. The effect of superimposed tissue in DM is the major limitation hampering the detection of masses. BT is a three-dimensional technique that reduces the effect of superimposed tissue. In the following two studies visibility was quantified for both imaging modalities in terms of the required contrast at a fixed detection performance (92% correct decisions). Contrast detail plots for lesions with sizes 0.2, 1, 3, 8 and 25 mm were generated. The first study involved only an in-plane BT slice, where the lesion centre appeared. The second study repeated the same procedure in BT image volumes for 3D distributed microcalcification clusters and 8 mm masses at two dose levels. Both studies showed that BT needs substantially less contrast than DM for lesions above 1 mm. Furthermore, the contrast threshold increased as the lesion size increased for both modalities. This is in accordance with the reduced effect of superimposed tissue in BT. For 0.2 mm lesions, substantially more contrast was needed. At equal dose, DM was better than BT for 0.2 mm lesions and microcalcification clusters. Doubling the dose substantially improved the detection in BT. Thus, system noise has a substantial impact on detection. The final study evaluated reading conditions for BT image volumes. Four viewing procedures were assessed: free scroll browsing only or combined with initial cine loops at frame rates of 9, 14 and 25 fps. They were viewed on a wide screen monitor placed in vertical or horizontal positions. A free-response task and eye tracking were utilized to record the detection performance, analysis time, visual attention and search strategies. Improved reading conditions were found for horizontally aligned BT image volumes when using free scroll browsing only or combined with a cine loop at the fastest frame rate.