Digital breast tomosynthesis (DBT) is a new and promising technology. The future roles of this technique, whether in a clinical or a screening setting, are considered in this article
Randi Gullien
Senior Radiographer
Robin Lee Hammond RT (R)(M)
Lead Mammographer
Per Skaane MD PhD
Breast Imaging Centre,
Department of Radiology,
Oslo University Hospital,
University of Oslo, Norway
The Norwegian Breast Cancer Screening Program (NBCSP), administrated by the Cancer Registry in Norway, invites all women 50–69 years of age biannually for routine screening mammography. The invitation includes the date, time and place for their screening mammograms. The full-field digital mammographic examinations (FFDM) include cranio-caudal (CC) and mediolateral-oblique (MLO) views of both breasts. Image interpretation is carried out on dedicated picture archiving and communication systems (PACS) workstations.
Independent double reading is the standard of practice. The radiologists performing the independent double reading use a 5-point rating scale for probability of cancer. A score of 2 or higher is defined as ‘positive’ in the NBCSP. If one or both radiologists gives a score of 2 or higher, the mammographic examination is discussed in an arbitration (or consensus) meeting. The interpretation scores are recorded directly into the NBCSP database. The same group of radiologists perform the diagnostic work-up including additional views, ultrasound, magnetic resonance imaging (MRI) and needle biopsy, if indicated. Short-term follow-up is not used.
Digital breast tomosynthesis
Digital breast tomosynthesis (DBT), a new promising technology for acquiring and displaying three-dimensional (3D) mammograms, uses the FFDM platform.(1–3) The 3D images are obtained in the same projections (CC and MLO) as for conventional FFDM and during the same compression as for 2D images. The tomographic cross-sectional images or ‘slices’ are typically of only 1mm thickness, but data may also be displayed as a volume consisting of several slices or ‘slabs’.
By reducing or eliminating overlapping tissue, the shape and margins of masses will be more delineated more clearly.(4,5) It is suggested that DBT might improve specificity compared with conventional 2D mammography alone. Furthermore, the elimination of overlapping breast tissue using DBT should theoretically improve the detection of hidden lesions in dense breast parenchyma and increase the sensitivity.
The performance of DBT regarding detection and characterisation of microcalcifications compared with FFDM has been a matter of concern owing to the so-called ‘thin-slice effect’. Only some calcifications might be seen on each slice and not all the calcifications of a cluster as on a conventional 2D image. On DBT, only a limited number of the calcifications are demonstrated on each of the 1-mm slices. Viewing the calcifications in a ‘slab’-mode can resolve some of these problems.(6,7) Another interesting finding from a clinical study has been that DBT is superior to 2D mammography especially for the measurement of breast cancer size.(8) The correlation between DBT and ultrasound is better than with 2D measurements.
All DBT examinations at the Breast Imagine Centre (BIC) and the screening centre were performed on Selenia Dimensions digital breast tomosynthesis units with dedicated workstations (Hologic Inc; Bedford, MA, USA).
Project in the clinical setting
A clinical project was undertaken at the BIC before the Oslo Tomosynthesis Screening Trial (OTST) commenced. The regional ethics committee approved the project and all women gave informed consent. Women who attended the BIC for imaging work-up were asked if they voluntarily wanted a work-up with tomosynthesis as well as a standard work-up. The main reason for this project was to compare FFDM and DBT in a side-by-side feature analysis for cancer conspicuity, and to determine whether there is an additional potential value of DBT compared with standard state-of-the-art conventional imaging work-up with regards to the detection of malignancies. There were different reasons why the women came to the BIC, including call-backs from the screening, a palpable lump, and routine mammography after prior breast cancer surgery.
The results from this study are based on a small number of cases and therefore cannot permit any final conclusion. The study indicates, however, that DBT seems to raise the malignancy conspicuity for cancers demonstrated at both FFDM (2D) and DBT and the ability to find cancers at DBT that are missed or overlooked at FFDM.9 DBT showed an increase in the cancer detection rate of 8% compared with combined conventional imaging (mammography plus ultrasound); this is an important premise for implementing DBT in mammography screening.
The Oslo Tomosynthesis Screening Trial
In November 2010, the screening centre in Oslo initiated a new prospective study with DBT. The OTST is an ongoing study and will continue throughout 2012. The regional ethical committee approved the study. Women residing in Oslo are offered the opportunity to participate voluntarily in the study at the time of their appointment. All attendees must sign an informed consent before performing the mammograms.
Women who participate in the study will have conventional 2D FFDM in both CC and MLO views. In the same compression as for 2D, the tomosynthesis (3D) mammograms are performed. The combination of both 2D and 3D is called a ‘combo’ exam. The compression time for a ‘combo’ is only an additional 4.5 seconds per image than for a single 2D mammogram.
A single DBT mammographic view gives approximately the same radiation dose as a single 2D mammographic view. Thus, the participating women received about double dose with a ‘combo’ as for a 2D examination only.
The positioning and compression of the breast is performed as in standard mammography. Performing the DBT, the grid is moved away from the receptor and an exposure is made while the X-ray tube moves along an arc of 15 degrees, allowing 15 low-dose projections to be acquired. The radiographers can view the projected images during the exposure on the unit console, because these images appear before the compression time and the exposure is completed. The data from these projections are used to produce reconstructed 1mm tomo slices immediately following the exposure. The number of slices in the 3D data set depends on the thickness of the compressed breast. For example, a 40mm breast is equal to 40 tomo slices.
At the workstation, the radiologists viewing the images have the option of layering multiple overlapping slices and to create ‘slabs’.
All the radiographers at the screening centre are trained in how to use and perform DBT on the mammography units. The women are positioned the same way for the DBT as for the 2D mammography exam alone. The only difference is the design of the tomosynthesis equipment. The positioning and the compression of the breast can be performed by any of the screening radiographers, no matter if it is 2D or a ‘combo’.
Eight radiologists, with between two and 20 years of experience in mammography screening, participated in this study. They were all trained in the interpretation on the workstations before the study began. All exams are interpreted independently in a batch mode for each of the four arms (Figure 1).
The hanging protocol included the conventional 2D FFDM in both views, CC and MLO as well as tomosynthesis in both views, CC and MLO, for arms C and D. If available, prior screening mammograms are reviewed. A scheduler was used to determine independently the specific batch to be interpreted and which of the eight radiologists would be assigned to interpret.
Each radiologist who made the interpretation independently rated each examination by breast using a NBCSP 5-point rating scale. A score of 2 or higher is defined as ‘positive’ in the NBCSP. If one or more of the radiologists recorded a score of 2 or higher, mammographic findings had to be specified. These cases are discussed in consensus meetings. Interpretation time and the interpretation scores were automatically recorded directly into the NBCSP database.
The OTST has four independent arms: A-D. Arm A includes 2D only, arm B includes 2D + CAD, arm C includes 2D + 3D (‘combo’), and arm D is a synthesised 2D + 3D. Figure 1 shows a synthesised 2D image, whch is a 2D image that has been reconstructed from the 3D dataset.
The NBCSP is offering women mammography with independent double reading, which is ensured in arm A and arm B. Including CAD in one arm, arm B, gives the opportunity/possibility to evaluate CAD with 2D mammograms in a screening setting (and to compare with arm A, 2D alone). CAD is more commonly used in the US than in Europe.
Arm C and arm D include tomosynthesis and 2D images displayed in different ways, arm C as standard 2D and arm D as synthetic 2D.
There are four different workstations, one for each arm. Only the hanging protocol for the specific arm is available, and there is a different log-in for each workstation. When a reader is logged in to one arm, he or she has no access to the other arms. Standardised hanging protocol arms are used.
Interpretation time is crucial in organised high-volume mammography screening using batch reading. There has been much concern regarding the increased reading time necessary for tomosynthesis in a screening setting.
Therefore, an automatic interpretation time recording was implemented in the OTST. The reading time for all radiologists in each arm is recorded. The reading times for the first three months of the trial were analysed, showing an overall mean reading time for 2D only (arm A and B) of 45 seconds compared with an overall mean reading time of 90-100 seconds for 2D plus tomosynthesis (arm C and D). Thus, the reading time for 2D plus 3D is about double that for conventional 2D alone. We think that the additional reading time of about 50 seconds can be justified if the cancer detection rate is significantly increased when using tomosynthesis.
The design of the OTST was a great challenge because we wanted to address several aspects of new advanced applications for mammographic screening, in this first large-scale trial of tomosynthesis in a screening setting. The women invited to the NBCSP should be offered the scheduled double reading of conventional 2D examinations. Computer-aided detection has been used widely in the US in mammography screening, but there is little experience with this new technology in organised screening programs in Europe.
Furthermore, there is no agreement whether tomosynthesis should be applied for one or both projections (CC and MLO) in addition to 2D only. We decided to use tomosynthesis in both projections, which means a double dose of radiation is administered to the participants.
This could be justified for this trial only because the double dose is relatively small compared with other X-ray examinations such as computed tomography (CT), and because this is a once-in-a-lifetime event for the women. A double dose of radiation as standard in future mammographic screening programs would, however, be unacceptable.
Therefore, the potential role of so-called ‘synthetic 2D’ images based on 3D tomosynthesis data of tomosynthesis is of special interest. If it turns out that these synthetic 2D views in combination with tomosynthesis increase the cancer detection rate significantly compared with 2D only using the same radiation dose, this could be an interesting technique in the future. These considerations were the background for the rather complicated design of the OTST (Figure 1).
The interim analysis based on the first three months of the study including 3488 women showed a statistically significant higher cancer detection rate for the combo mode (2D plus tomosynthesis) compared with 2D only (Figure 2). The OTST is scheduled for one complete screening round (two years) and the study will be closed at the end of December 2012.
Approximately 70% of the women attending the screening programme want to participate in the tomosynthesis trial, and consequently we expect that approximately 25,000 women will be included when the trial is closed at the end of this year. A two-year follow-up of the study population is needed in order to make corrections for interval cancers having true positive and false negative scores. It is suggested that tomosynthesis may increase the cancer detection rate significantly (Figure 3A, B). The results of this first large-scale prospective trial, implementing tomosynthesis in mammography screening, might give important information for the planning of future breast cancer screening.
Conclusions
It is not clear whether the future role of breast tomosynthesis and its implementation will be in breast cancer screening or in diagnostic mammography, or in both settings. Our results from the OTST to date indicate that DBT will represent an important adjunct to conventional imaging in mammography screening. It is still unclear whether tomosynthesis should be used in only one view or both views, and whether synthetic 2D images can replace the conventional 2D mammograms. Furthermore, large prospective studies similar to the OTST studies are needed to confirm the promising experience from our DBT screening trial.
References
- Baker JA, Lo JY. Breast tomosynthesis: State-of-the-art and review of the literature. Acta Radiol 2011;18:1298–310.
- Gennaro G et al. Digital breast tomosynthesis versus digital mammography: a clinical performance study. Eur Radiol 2010;20:1545–53.
- Andersson I et al. Breast tomosynthesis and digital mammography: a comparison of breast cancer visibility and BIRADS classification in a population of cancers with subtle mammographic findings. Eur Radiol 2008;18:2817–25.
- Gur D et al. Digital breast tomosynthesis: Observer performance study. Am J Roentgenol 2009;193:586–91.
- Hakim CM et al. Digital breast tomosynthesis in the diagnostic environment: A subjective side-by-side review. Am J Roentgenol 2010;195:172–6.
- Kopans D et al. Calcifications in the breast and digital breast tomosynthesis. Breast J 2011;17:638–44.
- Spangler ML et al. Detection and classification of calcifications on digital breast tomosynthesis and 2D digital mammography: A comparison. Am J Roentgenol 2011;196:320–4.
- Fornvik D et al. Breast tomosynthesis: Accuracy of tumor measurement compared with digital mammography and ultrasonography. Acta Radiol 2011;51:240–7.
- Skaane P et al. Digital breast tomosynthesis (DBT): initial experience in a clinical setting. Acta Radiol 2012;53:524–9.
