SS 602a - Breast density and background parenchymal enhancement
SS 602a - Breast density and background parenchymal enhancementThursday, March 2, 10:30 - 12:00 Room: E1 Session Type: Scientific Session Topics: Evidence-Based Imaging, Breast Digital Evaluation: Open Digital Evaluation for this Session Moderators: W. Alomaim (Dublin/IE), R. A. Kubik-Huch (Baden/CH) Add session to my schedule In your schedule (remove)
Purpose: Certain single nucleotide polymorphisms (SNPs) are associated with mammographic density and breast cancer. Our purpose was to investigate whether certain SNPs known from literature (i.e., candidate SNPs) were associated with mammographic density in women with breast cancer in a large Swedish cohort - the Malmö Diet and Cancer Study (MDCS).
Methods and Materials: From 1991-2007 a total of 826 women with breast cancer have been identified in the MDCS and radiological parameters for those women have been registered. Genotyping was performed on 762 nonrelated women with breast cancer. 15 SNPs in relation to mammographic density (fat involuted and moderately dense vs dense) were analysed with logistic regression yielding odds ratios (OR) with 95% confidence intervals (CI) adjusted for age at diagnosis, body mass index, and hormone replacement therapy (at baseline).
Results: For SNP rs6557161, the minor homozygote allele showed a significant relation to high mammographic density as compared to the major homozygote (ORadj 1.91 (1.12-3.26)). For SNP rs7289126, the minor homozygote (ORadj 0.60 (0.37-0.97)) and the heterozygote allele (ORadj 0.65 (0.45-0.94)) showed a significant relation to low mammographic density. For 13 of the selected SNPs no relation to mammographic density was observed.
Conclusion: We have found that two SNPs were associated with mammographic density in a large Swedish cohort, the MDCS, and further analyses on breast cancer survival will be performed. Increased knowledge regarding a potential shared genetic basis for mammographic parameters such as density could potentially help in the development of breast cancer risk scores including SNPs.
Determining factors of radiation dose in digital breast tomosynthesis (DBT) and full-field digital mammography (FFDM) utilising automated breast density and volumetric measurements
Purpose: To determine the factors affecting the radiation dose of stand-alone DBT versus FFDM.
Methods and Materials: IRB-approved study on women attending their surveillance mammography from February 2015 to March 2016 were performed. All women had both FFDM and DBT performed at same time under same degree of compression.The breast density and volumetric measurements were generated Quantra software (version 2.2, Hologic Inc, USA). The correlation between average glandular dose (AGD) per view with age, compressed thickness, and volumetric data of both FFDM and stand-alone DBT were performed with regression analysis.
Results: Retrospective analysis performed on 4283 study sets with mean AGD per view of FFDM and DBT equals 1.58mGy and 1.53mGy respectively.On regression analysis, there is positive correlation of AGD with compressed thickness, volume of fat, fibroglandular tissue and total breast volume for both FFDM and DBT, and negative correlation with age. All these parameters show stronger correlation in FFDM compared to DBT with better goodness of fit (R2) values, as well as larger regression coefficients. Among all, compressed breast thickness and volume of fibroglandular volume demonstrated largest regression coefficient measuring 0.039 and 0.0057 respectively, R2 = 57% and 32% respectively in FFDM, while for DBT the regression coefficient measured 0.021 and 0.0029 respectively.
Conclusion: The increase in compressed thickness and volume of fibroglandular tissue resulted in more significant increment in radiation dose in FFDM than in DBT, indicating that there was relatively less radiation dose in patients having stand-alone DBT compared to FFDM in population with predominantly dense breasts.
Purpose: To compare automated volumetric breast density measured by two different applications.
Methods and Materials: Mammograms of 110 women using a spectral photon-counting scanner (MicroDose mammography SI; Philips), from Jan 2015 to Feb 2016, were used. Breast volume (BV) (cm3), glandular volume (GV) (cm3), and volumetric breast density (VBD) (%) of two views of each breast were obtained from spectral energy data at the time of acquisition, using Breast Density Measurement (Philips). Three parameters were also obtained by Volpara (version 1.5.1; Volpara Solutions) using the raw data. Correlation coefficients (R) between MLO and CC views of the same breast for each system, and between two systems for each view were calculated. Weighted κ analyses were performed for density scores between the two systems, and between visual assessment and each system.
Results: Both systems showed very strong correlations between CC and MLO, for BV, GV and VBD (R=0.970, 0.977, 0.990 by Philips, R=0.963, 0.898, 0.864 by Volpara).BV and GV of two systems showed very strong correlations (R=0.993-0.997, 0.865-0.880), whereas VBD showed strong correlation (R=0.775-0.794). Density scores between Philips and Volpara, visual assessment and Philips, and visual assessment and Volpara showed moderate agreement (κ=0.510, 0.523, and 0.419).
Conclusion: Volumetric measurements of BV, GV, and VBD by both systems showed good precision. Philips system showed slightly better performances in correlation of VBD between CC and MLO, and agreement of density scores with visual assessment, which may be because direct measurements of spectral energy data are undertaken instead of processing of raw digital data.
Reproducibility of automated mammographic density measures between two digital mammography device vendors
Purpose: This study examined the reproducibility of automated mammographic density measures for mammograms acquired from two digital mammography device vendors during a one year period. Continuous measures of mammographic density included percent breast density, dense breast area, and total breast area; categorical measures of mammographic density included a 3-category scale and the 4th edition BI-RADS density scale.
Methods and Materials: Pairs of “for presentation” digital mammography images were obtained from two mammography units for 128 women who had a screening mammogram on one vendor unit followed by a diagnostic mammogram on a different vendor unit. All exposures occurred within a single department by the same group of mammography-certified technologists and under the direction of a single medical director, technical manager, and quality assurance officer. A fully-automated density software was used to generate mammographic density measures from all available standard screening view images from the two vendors. Intra-class correlation coefficients (ICC), Pearson’s correlation coefficient (PCC), weighted kappas, and scatter plots were used to evaluate mammographic density measures between vendors.
Results: ICCs of 0.94, 0.90, and 0.98 and PCCs of 0.94, 0.92, and 0.99 were observed for percent density, dense breast area, and total breast area measurements respectively. Weighted kappas for a 3-category scale and the 4th edition BI-RADS scales were 0.85, 0.83 respectively.
Conclusion: Overall agreement of mammographic density measures automatically generated from pairs of “for presentation” digital mammography images was almost perfect between the two vendors.
Purpose: To compare the reliability of automated breast density (ABD) measurements and visual assessment in terms of reproducibility.
Methods and Materials: 159 women (excluding those with previous breast operation) attending surveillance mammography at least once with digital breast tomosynthesis (DBT) from October 2014-June 2016 recruited with IRB approval. Breast density assigned using BIRADS 5th edition (Fatty: a, scattered fibroglandular tissue: b, heterogeneously dense: c, markedly dense: d). Simply dense (category c+d) and non-dense breast (category a+b) grouping done and analysed. Quantra software [version2.2; Hologic Inc, USA] utilised for ABD. For visual assessment, two blinded radiologists (Readers 1&2) assigned breast density to same group of patients for visits 1 and 2. Kappa statistics used to assess agreement between different comparison groups.
Results: [Same visit] Agreement of Quantra ABD of both breasts (assuming symmetrical) compared: Kappa coefficient 0.8 (CI 0.6-0.9). As only one measurement was given to each patient in visual assessment, agreement of Readers 1 and 2 assigned breast density assignment compared for same visit, but kappa coefficient was only 0.27 (CI 0.13-0.41). [Same patient, different visits] Agreement of Quantra assigned breast density category at two visits compared: Kappa co-efficient 0.82 (CI 0.73-0.91). For visual assessment reproducibility, the agreement of the Reader 1 in the first visit was compared with Reader 2 in the second visit, but kappa coefficient was only 0.16 for visual assessment (CI 0-0.3).
Conclusion: Automated breast density is more reliable in terms of reproducibility of breast density assessment than visual assessment, which has high inter-observer variability.
Accuracy of fully automated volumetric FGT measurement with MRI of the breast: correlation with anthropomorphic breast phantoms
Purpose: To investigate the accuracy of fully automated measurements of the amount of fibroglandular tissue (FGT) with magnetic resonance imaging (MRI) using different MRI sequences based on anthropomorphic breast phantoms as the ground truth.
Methods and Materials: In this study, ten anthropomorphic breast phantoms that consisted of different known fractions of adipose and fibroglandular tissue, which closely resembled normal breast parenchyma, were developed. Anthropomorphic breast phantoms were imaged with a 1.5 Tesla unit (Siemens, AvantoFit) using an 18-channel breast coil. The sequence protocol consisted of an isotropic Dixon sequence (Di), an anisotropic Dixon sequence (Da), and T1 3D FLASH sequences with and without fat saturation (T1). Fully automated, quantitative, volumetric measurement of FGT for all anthropomorphic phantoms and sequences was performed and correlated with the amount of fatty and glandular components in the phantoms as the ground truth.
Results: Fully automated quantitative measurements of FGT with MRI were performed successfully in all phantoms 6.71-59.72% (mean 32.82%) and for all sequences; Di, Da, and T1 ranged from 6.98-58.05% (mean 32.89%), 7.43-61.05% (mean 34.76%), and 6.44-54.69% (mean 34.67%), respectively. All sequences yielded good correlation with actual FGT content. However the best correlation of FGT measurement results and anthropomorphic breast phantoms was identified for Dixon sequences; Di (R2=0.998), Da (R2=0.997), and T1 (R2=0.910).
Conclusion: Fully automated quantitative measurements of FGT with MRI provide accurate information on the actual amount of FGT. Dixon type sequences showed the highest correlation and reproducibility of automated, quantitative FGT measurements on anthropomorphic breast phantoms, compared to conventional sequences.
Purpose: To evaluate the distribution of MRI BPE among different breast cancer subtypes searching for any correlation with immunohistochemical and receptorial panel (Estrogen Receptor -ER, Progesterone Receptor - PR, Human Epidermal Growth Factor Receptor 2 - HER2).
Methods and Materials: 41 consecutive patients affected by breast cancer underwent breast DCE-MRI. Two radiologists evaluated all subtracted MR enhanced images for classifying normal BPE. ER, PR and HER2 expression was assessed by immunohistochemical analysis. ER and PR status was evaluated using Allred score (positive values: score ≥3). The intensity of the c-erbB-2 staining was scored as 0, 1+, 2+, or 3+ (positive values: ≥ 3+; negative: 0 and 1+; 2+ value assessed with silver in-situ hybridisation). Patients were subdivided into four categories based on cancer subtypes: ER/PR+HER2+, ER/PR+HER2-, ER/PR-HER2+, ER/PR-HER2- (triple negative). Distribution of BPE into the four categories was assessed and any significant difference was calculated with the comparison of proportions, with p≥0.05 considered as significant.
Results: 6/41 (14.5) patients were ER/PR+Her2+, 24/41 (58.5%) were ER/PR+HER2-, 2/41 (5%) were ER/PR-HER2+, 8/41 (19.5%) were ER/PR-HER2-. No significant correlation between BPE pattern and receptor expression among breast tumours was found (p>0.05). Mild/moderate BPE were the most prevalent patterns in ER/PR+HER2+, ER/PR+HER2-, ER/PR-HER2+ groups; in the ER/PR-HER2- (triple negative) group, moderate/marked BPE prevailed.
Conclusion: BPE shows no significant correlation with breast tumour receptorial panel, except for triple negative breast tumours, in which an atypical trend for moderate/marked BPE prevalence could be supposed. Further studies need to be performed to confirm this preliminary data.
Association of breast density and region of interest size with apparent diffusion coefficient value of normal fibroglandular tissue at MRI
Purpose: To determine association of breast density and region of interest (ROI) size with apparent diffusion coefficient (ADC) value of normal fibroglandular tissue on diffusion-weighted imaging (DWI) at MR.
Methods and Materials: The retrospective study included 28 women who underwent clinical MR exams. Breast density was evaluated at MRI and divided into low and high density category, the former including breasts interpreted as being fatty or having scattered fibroglandular tissue, and the latter including heterogeneously dense and dense breasts. DWI was performed at 1.5T using b = 0, b = 400 and b = 800. ADC of normal fibroglandular tissue was calculated by manually drawing two sets of multiple circular ROI´s of sizes equalling 15 mm2 and 30 mm2. Areas of fibroglandular tissue not large enough to encompass the larger ROI were excluded from measurements to avoid partial volume averaging. Intrasubject mean ADC values for the two ROI sizes were compared by Pearson’s correlation coefficient. Association between mean ADC and breast density level was assessed using Student's t-test.
Results: Mean ADC of fibroglandular tissue was 1.79 ±0.18 x10-3 mm2/s and 1.8±0.2 x10-3 mm2/s on smaller and larger ROI respectively. Intrasubject ADC measurements for the two ROI sizes were highly correlated (r=0.954, p<0.001). ADC was not associated with breast density level (p>0.5).
Conclusion: ADC reflects microstructural characteristics of fibroglandular tissue unrelated to breast density or ROI size. Contrast between lesions and fibroglandular tissue on DWI will not be affected by breast density.
Effect of parenchymal pattern in women with dense breasts, variation with age and impact on screening outcomes: observations from a UK screening programme
Purpose: To analyse mammographic parenchymal pattern (PP) in women with densest breasts; to determine how tissue pattern varies with age and impacts recall and cancer detection.
Methods and Materials: Breast density data (VolparaTM , 5th ed.) was obtained in a subset of women screened in a regional programme (April 2013-March 2015). Cases with densest breasts (Volpara Grade d [VGd]) were selected for PP categorization. All assessed VGd cases and 100 randomly selected VGd cases from non-assessed women were reviewed. Ten readers classified PP on a 1-5 scale (1-very smooth to 5-very nodular). Mean PP classification was compared between age groups and for assessed vs non-assesed women. Reader agreement was analysed using intraclass correlation. Likelihood of biopsy, cancer diagnosis and cancer characteristics were analysed by age and PP.
Results: 4,331/40,760(10.6%) screened women were VGd. Proportions of PP categories were similar at all ages for controls (p = 0.145) and for 280 assessed women (p=0.657). Inter-rater reliability for scoring PP in controls was good (ICC = 0.6302). 155 women underwent biopsy of which 34 had cancer. PP (1-5) ratios did not vary significantly with age in these groups (p = 0.580). There was significant correlation between cancer and nodular PP (p = 0.043). Cancer characteristics did not differ by age or PP.
Conclusion: Ratio of smooth to nodular pattern in women with the densest breasts did not vary with age. PP did not affect likelihood of recall to assessment or biopsy. There was a significant association between nodular PP and cancer diagnosis.
Purpose: To investigate the trends of PPV-1 (the percentage of screening detected breast cancers among the recall examinations due to abnormal mammographic findings) and PPV-2 (the percentage of screening detected breast cancers among the recall examinations that include an invasive diagnostic procedure) by categories of subjectively assessed mammographic density (MD).
Methods and Materials: We used information from 6363 recall examinations obtained from 5666 women screened with full field digital mammography in the Norwegian Breast Cancer Screening Programme, 2003-2010. The radiologists who performed the recall examination classified the mammograms into three categories based on MD: MD-1 (<30% of fibroglandular tissue), MD-2 (30-70%), and MD-3 (>70%). We used chi-square trend statistics to estimate trends of PPVs across MD categories. Logistic regression was used to estimate the odds ratio (OR) of screening detected breast cancer associated with MD among recalled women, adjusting for age.
Results: PPV-1 and PPV-2 decreased by increasing MD category (p<0.05 for the trend). MD-2 and MD-3 were associated with lower odds of breast cancer among recalled women compared with MD-1. Adjusted OR was 0.79 (95% CI: 0.68-0.92) and 0.75 (95% CI: 0.58-0.97) for MD-2 and MD-3, respectively.
Conclusion: High MD was associated with a higher amount of false positive results, and thus decreased radiologists’ performance, compared with low MD in the national screening programme in Norway.
Purpose: This study examined the variability in inter-rater reliability between radiologists assessments of mammographic density using the 4th and 5th edition of the BI-RADS density classification scales.
Methods and Materials: Six radiologists assessed mammographic density for 375 cases using the 4th and 5th editions of the BI-RADS density scale. A consensus assessment was calculated based on the majority assessment. Chance-corrected inter-rater agreement was evaluated using the weighted kappa statistic for all pairs of radiologists and between each radiologist and the consensus measure.
Results: Inter-rater agreement between all pairs of radiologists was moderate to almost perfect for the 4th edition scale (mean Kappa=0.76, range=0.53-0.90), and substantial to almost perfect for the 5th edition scale (mean Kappa=0.79, range=0.71-0.83). Inter-rater agreement between each radiologist and the consensus measure was substantial to almost perfect for the 4th edition scale (mean Kappa=0.85, range=0.69-0.92) and almost perfect for the 5th edition scale (mean Kappa=0.86, range=0.81-0.91).
Conclusion: In the absence of strict percent density cut-points in the BI-RADS 5th edition density scale, inter-rater agreement is significantly better than for the BI-RADS 4th edition density scale. The range of inter-rater variability between pairs of raters and between raters versus the consensus measure was considerably less for the 5th edition BI-RADS density scale compared to the 4th edition scale. Visual assessment of density using the BI-RADS 5th edition density scale may be more reliable than using the BI-RADS 4th edition density scale.