Breast tomosynthesis – new perspectives on breast cancer screening

Sammanfattning: Abstract: Mammography is currently the established method in breast cancer screening, although the sensitivity is known be affected by overlapping tissue concealing tumours. Breast tomosynthesis takes advantage of multiple exposures at different angles reducing the negative effect of obscuring tissue. The aim of this thesis was to investigate the use of tomosynthesis in breast imaging with a special focus on the role of tomosynthesis in screening. Papers I and II was based on 7,500 women comprising the first half of the Malmö Breast Tomosynthesis Screening Trial (MBTST): a prospective population-based single arm study including randomly invited women 40–74 years old eligible for the screening programme in the city of Malmö. Women underwent one-view tomosynthesis with reduced compression force and mammography. The images were read and scored separately in a blinded double-reading procedure. The increase in cancer detection rate (6.3 to 8.9/1,000) and recall rate (2.6% to 3.8%) using tomosynthesis was significant (p<0.0001). The additional cancers detected were mainly invasive, with a tendency of downstaging (Paper I). Findings of stellate distortion simulating malignancy increased the false positive rate with tomosynthesis. The false positive rate was reduced over time, suggesting a learning curve (Paper II). Paper III was a human observer study designed to study the effect of different viewing procedures (free scroll volume browsing and initial cineloops at three different frame rates) and image orientations (vertical and horizontal) on reading performance. 40 three-dimensional simulations of breast malignancy were inserted in 55 normal tomosynthesis image volumes, creating four unique image sets, one for each viewing procedure. Detection accuracy was similar for the different reading conditions but reading time and visual attention were improved for horizontally oriented images. In Paper IV, 26 breast cancers subjected to discrepant detection in tomosynthesis and mammography in a previous reader study with 185 cases with normal and diseased breasts (95 cancer lesions), were analysed in a side-by-side review to assess factors related to lesion visibility and possible reasons for detection error. Breast cancers imaged with tomosynthesis were found to be more conspicuous with a higher proportion of discernable lesion periphery. Missing a cancer in tomosynthesis seemed to be related to interpretative error of visible lesions. In conclusion, one-view tomosynthesis, with reduced compression force, might be sufficient as a stand-alone screening modality.