Data selection
Anonymized patient data were collected over 3 years (2016-2019) from five systems: Siemens Inspiration, Siemens Inspiration PRIME, Siemens Revelation PRIME, Hologic Selenia Dimensions and GE Senographe Pristina using the dose monitoring application DOSE (QaelumNV, Belgium). Data were selected based on age (45-75 years), compression force (80-120N) and breast thickness (10-100 mm), resulting in 7333 DBT series and 49576 DM images (Table 1). Mean glandular dose (MGD) was calculated following Dance’ method [7-8] as well as retrieved from the image header (indicated dose). We should note that only a limited number of DBT series could be retrieved for the Hologic system of which the majority had small breast thicknesses (median=48 mm). This led to an overall lower median MGD for DBT compared to DM mode for the Hologic system.
Table 1. Demographic data and mean glandular doses (MGD) for DM and DBT mode for the five systems included in this work. Data are represented as median ± standard deviation.
|
Siemens Inspiration
|
Siemens Inspiration PRIME
|
Siemens Revelation PRIME
|
Hologic Selenia Dimensions
|
GE Senographe Pristina
|
|
DM
|
DBT
|
DM
|
DBT
|
DM
|
DBT
|
DM
|
DBT
|
DM
|
DBT
|
Number of Images
|
35246
|
4499
|
8665
|
1423
|
712
|
535
|
4543
|
281
|
410
|
595
|
Patient age(years)
|
58 ± 7
|
60 ± 8
|
58 ± 7
|
61 ± 8
|
58 ± 7
|
61 ± 8
|
60 ± 8
|
58 ± 8
|
62 ± 7
|
64 ± 7
|
Breast thickness (mm)
|
50 ± 14
|
56 ± 14
|
51 ± 14
|
53 ± 14
|
55 ± 14
|
59 ± 15
|
60 ± 15
|
48 ± 15
|
59 ± 12
|
59 ± 13
|
Compression force (N)
|
109 ± 10
|
108 ± 10
|
111 ± 10
|
109 ± 10
|
107 ± 11
|
105 ± 10
|
108 ± 10
|
105 ± 11
|
109 ± 6
|
105 ± 11
|
MGD dicomheader (mGy)
|
1.21 ± 0.51
|
2.05 ± 0.65
|
1.07 ± 0.49
|
1.86 ± 0.56
|
1.08 ± 0.55
|
1.74 ± 0.54
|
1.62 ± 0.81
|
1.55 ± 0.71
|
1.36 ± 0.35
|
1.27 ± 0.34
|
MGD calculated (mGy)
|
1.21 ± 0.56
|
2.13 ± 0.80
|
1.01 ± 0.57
|
1.85 ± 0.70
|
1.13 ± 0.68
|
1.71 ± 0.66
|
1.42 ± 0.74
|
1.37 ± 0.59
|
1.39 ± 0.42
|
1.33 ± 0.42
|
Data analysis
The running 5th percentile, median and 95th percentile of MGD were calculated. From these data three breast thickness ranges were further selected covering small, medium and larger breasts i.e. 25-35, 50-60 and 70-80 mm for DM and DBT overall, and for each device separately.
Finally differences between indicated and calculated MGD were compared using the Wilcoxon test and plotted as a function of breast thickness.
Results - DM versus DBT doses
The running median MGD at 25-35, 50-60 and 70-80 mm for all DBT systems combined was 1.14, 1.70 and 2.77 mGy, this was an increase of 46%, 39% and 40% compared to the median of all DM systems combined. The 5th-95th percentile dose distributions for medium breasts for the five systems combined was 0.90 to 1.86 mGy for DM and 1.32 to 2.39 mGy for DBT. The ratio of the median MGD for DBT/DM for a medium breast was 1.57, 1.63, 1.38, 1.39 and 0.99 for Siemens Inspiration, Siemens Inspiration PRIME, Siemens Revelation PRIME, Hologic Selenia Dimensions and GE Senographe Pristina respectively. These data show that MGD values were significantly higher for DBT mode compared to DM for the Siemens and Hologic systems. For the GE system DBT and DM doses were similar.
Results - Dose distribution versus breast thickness
Figure 1 shows scatterplots of the MGD as a function of breast thickness for the five systems under investigation. Additionally the running median, 5th and 95th percentile are added on the graphs to facilitate recognition of outliers. These graphs show an increasing MGD for increasing breast thickness for the different vendors with the smallest increment for the GE system. In addition a larger spread of MGD, especially at larger breast thicknesses, is seen for the Siemens systems compared to the Hologic and GE system. The large spread in MGD for both the Siemens Inspiration and Siemens Inspiration PRIME systems might be due to the fact that these graphs contain many more datapoints compared to the other three systems.
Results: Comparison of MGD between systems
Figure 2 shows boxplots with MGD distributions for mammography for the five systems under investigation for three breast thickness ranges covering small, medium and large breasts. For small breasts there is a large variation in dose between the systems. This difference is reduced for medium and larger breasts showing similar dose distributions for the five systems. Figure 3 shows similar boxplots for DBT mode. For DBT, larger differences are found between the systems which remain for the three breast thickness groups. One can also notice existing dose differences between different system types of the same vendor.
Results - Indicated versus calculated dose
Figure 4 illustrates the relative differences between indicated (dicomheader) and calculated (Dance' method) MGD for DBT mode for the five different systems. In addition, the Wilcoxon matched-pairs test shows a significant difference (p<0.0001) between indicated and calculated doses for both DM and DBT and for all systems. Different trends are found for the different systems: for the GE system a fairly constant deviation between indicated and calculated dose was found as a function of breast thickness whereas for Hologic this difference increased from small to large breasts. For the Siemens systems, on the other hand, we noticed a fairly good correspondence between indicated and calculated MGD for medium breast sizes with increasing differences for small and large breasts.