SS 1502 - Breast density: an update
BI-RADS-based subjective estimation of fibroglandular breast tissue with magnetic resonance imaging: comparison to automated quantitative assessment
Purpose: To evaluate the inter- and
intra-observer agreement of BI-RADS-based subjective visual
estimation of FGT with MRI, and to investigate whether FGT
assessment benefits from an automated quantitative MRI measurement
by comparing both approaches.
Methods and Materials: Eighty women (mean age, 46; range, 21 to 86 years) with no imaging abnormalities on mammography or ultrasound (BI-RADS 1 and 2) were included in this IRB-approved prospective study. All women underwent un-enhanced breast MRI using the Dixon sequence (TR/TE 6ms/1.45ms/2.67ms, 256 slices, matrix 352x352, 1 mm isotropic, TA 3:38min). Four radiologists independently assessed FGT with MRI by subjective visual estimation according to the revised ACR BI-RADS atlas. Automated quantitative measurement of FGT with MRI was performed using a previously described measurement system. Inter- and intra-observer agreement of qualitative and quantitative FGT measurements were assessed using Cohen’s kappa (k).
Results: There was only a moderate inter- and intra-observer agreement for subjective visual estimation of FGT with MRI for inexperienced readers. Experienced readers achieved better results, with a substantial inter-observer agreement and perfect intra-observer agreement. Practice and experience reduced observer-dependency. Automated quantitative measurement of FGT with MRI was successfully performed for every examination and revealed only fair to moderate agreement (k= 0.209-0.497) with subjective visual estimations of FGT.
Conclusion: Subjective visual estimation of FGT with MRI shows moderate intra- and inter-observer agreement, which can be improved by practice and experience. Automated quantitative measurements of FGT with MRI are necessary to allow a standardized risk evaluation, appropriate management, and the assessment of preventive breast cancer measurements.
Purpose: High breast density is associated
with increased breast cancer risk. 2D mammographic percent density
(MPD) quantification is limited, reflecting X-ray attenuation
characteristics and is not used in younger women. MRI offers an
ionising-radiation free 3D alternative with measurements of
water-based tissue in the breast correlating strongly with MPD.
Ultrasound tomography (UST) is an emerging whole breast 3D-imaging
technique that obtains quantitative tomograms of speed of sound (as
well as other properties) of the entire breast. The volume averaged
speed of sound (VASS)holds promise for improving on traditional
assessment by measuring density at each voxel and offering a cheap,
patient-acceptable, ionizing-radiation free alternative. This study
evaluated UST by comparing VASS with a measurement of water-based
tissue from non-contrast MRI.
Methods and Materials: 50 healthy volunteers from the Generations study (median age 40 years, range 30-64 years) underwent bilateral breast UST, 46 of whom underwent MRI using a 2-point Dixon technique. VASS and percentage water density were measured in both breasts and compared using Pearson’s correlation coefficient.
Results: The mean VASS for the cohort was 1446+/-148ms-1 (range 1434-1541ms). There was high similarity between measurements from the right and left breasts (1463+/-29ms-1, 1459+/-29ms-1 respectively (p=0.516))(ICC=0.98). Mean percentage water density for the cohort was 34.6+/-14.5% (range 13.5-74.4%) with good right-to-left consistency (35.7+/-15.3%, 34.4+/-14.6% respectively (p=0.55)). There was excellent correlation between VASS and percentage water density (r2= 0.97, p<0.0001).
Conclusion: UST holds promise as a novel accurate, reproducible and ionising-radiation free technique to evaluate breast density.
Mammographic breast density and HER2 overexpression assessment improves the Nottingham Prognostic Index prognostic ability in patients with invasive breast cancer
Purpose: To examine the possible additional
value of very low mammographic breast density (VLD), HER2, ER and
PR statuses in a patient group within matched Nottingham Prognostic
Index (NPI) categories. We also aimed to investigate whether those
variables could be incorporated into the NPI.
Methods and Materials: Altogether 270 patients with newly diagnosed invasive breast cancer were included in the analysis. Patients with mammographic breast density of <10% were considered as patients with VLD breasts. We compared the performance of NPI with and without VLD, ER, PR and HER2 statuses. Cox multivariate analysis, time-dependent receiver operating characteristic curve (tdROC), concordance index (c-index) and prediction error (0.632+ bootstrap estimator) were used to derive an updated version of NPI.
Results: Both mammographic breast density (VLD) (p<0.001) and HER2 status (p=0.049) had a clinically significant effect on the disease free survival of patients in the intermediate and high risk groups of the original NPI classification. The incorporation of both factors (VLD and HER2 status) into the NPI increased the prognostic power (C-index 0.872 vs 0.779, p<0.001) and provided improved patient outcome stratification by decreasing the percentage of patients in the intermediate prognostic groups.
Conclusion: Very low breast density and HER2 positivity are prognostic factors for breast cancer independent of the NPI. Moreover, their addition to the NPI helps increase its accuracy and thus offers an improved, but still readily available method for the evaluation of patient prognosis.
Error in recorded compressed breast thickness measurement impacts on volumetric density classification
Purpose: Mammographic density has been
demonstrated to predict breast cancer risk. It has been proposed
that it could be used for stratifying screening pathways and
recommending additional imaging. Volumetric density tools use the
recorded compressed breast thickness (CBT) of the breast measured
at the x-ray unit in their calculation, however the accuracy of the
recorded thickness can vary. The aim of this study was to
investigate whether inaccuracies in recorded CBT impact upon
volumetric density classification and to examine whether the
current QC standard is sufficient for assessing mammographic
Methods and Materials: Raw data from 52 digital screening mammograms were included in the study. For each image, the clinically recorded CBT was artificially increased and decreased to simulate measurement error. Increments of 1 mm were used up to ±15% error of recorded CBT was achieved. New images were created for each 1 mm step in thickness resulting in a total of 572 images which then had Volpara Density Grade (VDG) and volumetric density percentage assigned.
Results: A change in VDG was recorded in 38.5% (n= 20) of mammograms when applying ±15% error to the recorded CBT and 11.5 % (n= 6) were within the quality control (QC) standard prescribed error of ±5mm.
Conclusion: The current QC standard of ±5mm error in recorded CBT creates the potential for error in mammographic density measurement. This may lead to inaccurate classification of mammographic density. The current QC standard for assessing mammographic density should be reconsidered.
Purpose: The American College of Radiology
BI-RADS4 quantitative breast density scale differs from the
BI-RADS5 qualitative scale, which emphasises both the amount and
the masking effect of dense tissue. This study evaluates the
agreement between these two classification scales.
Methods and Materials: Six radiologists assessed breast density composition on a set of 375 cases from four standard screening views using the 4th and 5th editions of the BI-RADS density composition scale (labeled 1/2/3/4 and A/B/C/D, respectively). Overall between-rater agreement was evaluated using the multi-rater weighted kappa statistic. A consensus (majority) assessment was used to measure between-scale agreement using the two-rater weighted kappa statistic, and to calculate observed proportions in each density category for both scales.
Results: The observed proportions for the 1/2 and 3/4 density categories were 63 % and 37 % respectively, virtually identical to the A/B and C/D density categories which were 62 % and 38 % respectively Correspondence within the 1&A, 2&B, 3&C and 4&D pairs was not as close: of studies classified as 1, only 78 % were classified as A, for 2 only 85 % were classified B, for 3 only 89 % were classified C, and for 4 only 51 % were classified D. Between-rater agreement was substantial for BI-RADS4 (Kappa=0.76) and BI-RADS5 (Kappa=0.79) and agreement between the two scales was excellent (Kappa=0.89).
Conclusion: BI-RADS5 has high agreement with BI-RADS4, but a small proportion of patients who had previously been classified as having “dense" breasts using BI-RADS4 may no longer be classified as “dense” using BI-RADS5 and vice versa.
Purpose: BRCA1/2 mutations account for
30-50% of hereditary breast cancers and bilateral oophorectomy is
associated with a reduced risk of BC in these patients. Breast
density is a well-established breast cancer risk factor and is also
associated with increased risk in BRCA carriers. The relationship
between oopherectomy and breast density and which method of breast
density interpretation is best for temporal change has not been
Methods and Materials: Retrospective study of 50 BRCA1/2 patients who underwent oophorectomy and had at least a baseline and post surgery mammogram. Mammographic breast density was determined by Volpara and visual assessment by a single radiologist.
Results: At baseline, there was a trend to decreased density with increasing age which was not significant. Patients with a family history of breast cancer also had increased breast density but this difference was not statistically significant. Breast density significantly decreased after oopherectomy with an average 2.1% absolute decrease (p value <0.001). There was a higher absolute decrease in breast density in patients aged between 40-50 due to the higher baseline density. Using Volpara Density Grades (analogous to BI-RADS density categories), 84 % of women displayed a decrease in density category over the study period compared to only 76 % using the radiologist visual classification.
Conclusion: Oopherectomy is associated with a decrease in breast density and younger patients exhibit a larger absolute decrease. Volpara is more sensitive to change over time compared to visual assessment.
Purpose: Literature has established a
strong association between mammographic density and breast cancer,
as well as the difficulty in evaluating the breast with increased
mammographic density. Perception studies are an important component
of radiology, these provide an understanding of the scan patterns
and the interpretation behaviour of the reporting
Methods and Materials: Breast radiologists from the UK were voluntarily recruited to review sets of anonymized mammographic images (n=170) and to subjectively rate the breast density according to the BI-RADS categorization. The images were used along with the use of a TOBII eye-tracker (TOBII Technologies, Sweden) to track the eye movement of the Radiologists while assessing the images. Images were reviewed using standardized viewing conditions and Ziltron software. The results were analysed using a paired sample T-test and ANOVA.
Results: Complete data was obtained for 17 radiologists (1-10+years experience) and results showed radiologists spent significantly more time observing MLO (MLO: 1.67sec vs CC: 1.32sec; mean values) and right images (right: 1.63sec vs left: 1.36sec; mean values) on the mammographic images (p<0.05). Radiologists spend significantly more time observing images with BI-RAD 3 (mean=1.71sec) density than the other 3 BI-RADs (p<0.001).
Conclusion: The impact of breast density on radiologist fixation time in mammography varies according to the density. The variation in how radiologists viewed projections and different sides of the breast warrants further research as does potential impact upon abnormality identification. An understanding of radiology review patterns will inform training and support the amelioration of potential interpretation differences.
Purpose: Mammographic breast density and
breast cancer personal/family history are independent risk factors
in the general population. Our aim was to evaluate the association
between these factors.
Methods and Materials: From August to September 2015, we prospectively enrolled women presenting for screening mammography, excluding those with history of cancer or short life expectancy. Each patient was proposed the questionnaire for the breast or ovarian cancer family history using the IBIS RISK EVALUATOR based on Tyrer-Duffy-Cuzick model (Stat Med 2004), providing the individual lifetime risk (LTR) and 10-year risk (10YR). An expert radiologist visually evaluated mammographic breast density using the American College of Radiology classification: a (lowest quartile), b, c, d (highest quartile). Data were given as median and percentiles. Association between breast density and risks or patient age was estimated using the Kruskall-Wallis test; multivariate linear regression analysis was also performed.
Results: In the study period, 207 women were enrolled (median age 59 years; percentiles 49-66 years). Median 10YR was 3%(2-4%); median LTR was 7 % (4-11%). From density class a to d, median patient age was 62, 63, 52 and 50 years (p<0.001); the median 10YR was 3 % for all (p=0.938); the LTR was 5 %, 6 %, 9 % and 8 % (p=0.005). Multivariate linear regression analysis using LTR as dependent variable showed a highly significant contribution for age (p<0.001) but only a borderline significant contribution for mammographic density (p=0.066).
Conclusion: Our data suggest that the association between breast density and cancer risk based on personal/family history is mediated by the patient age.
Classification of mammographic densities and breast cancer risk: results from the Egyptian national breast screening study
Purpose: Our objective was to determine the
relation between breast ca and the recent ACR classification of
mammographic densities in our Egyptian screening population
Methods and Materials: From the medical records of the National screening program, 48 407 women assigned to digital mammography screening were collected, 561 cases of breast cancer were detected. Mammograms were classified into four, ACR-A, B, C & D categories of density, by 2 senior radiologists.
Results: From 2007 to 2015, 48407 women were outreached for screening by digital mammography, age between 40 and 65.They were all included in the study.The calculated total frequency of cancer was 0.046.Densities were ranked in descending pattern in the following order according to their frequency of positive cases:D (2.31%); C (1.62%); B (1.09%); A (0.71%).The frequency of breast cancer in ACR-A density group was 0.007 and in ACR-B group was 0.010, in the ACR-C group was 0.016 and in the ACR-D group was 0.023. Statistically, highly significant increased frequency of positive cases among in ACR Mammographic density (MD) B compared to A. The same between each 2 groups except class C versus D where there was no statistically significant change in the frequency of positive cases in one versus the other.There is positive significant risk among each higher MD compared to the lesser MD, except in ACR- classes C and D where the significant risk is equal.
Conclusion: In our population,increase in the mammographic breast density is associated with increase in risk for breast cancer.
Agreement between radiologists’ visual assessments and automated software: BI-RADS 5th edition density classifications
Purpose: Many automated breast density
tools have been validated in the context of breast density
quantification, however their ability to qualify the likelihood of
dense tissue masking a lesion has been less studied. This study
aims to quantify the agreement between radiologists’ assessments of
breast density and those from an automated algorithm that considers
the amount and appearance of breast density consistent with the 5th
edition of the BIRADS density lexicon.
Methods and Materials: Six radiologists assessed breast density composition from four de-identified standard screening views using the 5th edition of the BI-RADS density composition scale on a set of 375 cases. Agreement between each rater and the automated breast density measurement software (DM-Research, Densitas Inc). was measured using the weighted kappa statistic. Agreement between the radiologists’ consensus assessment and the algorithm was measured using the weighted kappa statistic.
Results: Overall agreement between the algorithm and each of the radiologists was excellent, Kappa = 0.81 to 0.90. Agreement between the algorithm and the consensus assessment was also excellent, Kappa = 0.89. Of the 375 cases reviewed, the algorithm scored A, B, C, and D in 21 %, 38 %, 36 %, and 5 % of cases, while the consensus radiologist assessment scored A, B, C, and D respectively in 28 %, 34 %, 34 %, and 4 % of cases.
Conclusion: Automated breast density software that accounts for the amount and dispersion of dense breast tissue can reduce the subjectivity of radiologist-assigned density scores and can have substantial agreement with radiologists’ visual assessments of density.
Background parenchymal enhancement is not associated with breast cancer in a non-high-risk population
Purpose: Previously, an association between
BPE and breast cancer was reported in high-risk populations. We
sought to determine, whether this was also true for non-high-risk
Methods and Materials: 668 consecutive patients unterwent breast MRI for assessment of breast findings (BI-RADS 0-5, no high-risk screening) and subsequent histological work-up. For this IRB-approved study, BPE and FGT were retrospectively assessed by two experienced radiologists according to the BI-RADS lexicon. Pearson correlation coefficients were calculated to explore associations between BPE, FGT, age and final diagnosis of breast cancer. Subsequently, mulitvariate logistic regression analysis considering covariate colinearities was performed, using final diagnosis as the target variable and BPE, FGT and age as covariates.
Results: Age showed a weak negative correlation (r<=-0.25) with FGT and BPE (p<0.001). BPE correlated moderately with FGT (r= 0.35, p<0.001). Final diagnosis displayed weak negative correlations with FGT and BPE (<=-0.18, p<0.004) and weak positive correlation with age (r=0.18, p<0.001). On logistic regression analysis, the only independent covariate for prediction of final diagnosis was age (OR 1.011, 95%-CI:1.004-1.018, p=0.001).
Conclusion: Based on our data, neither BPE nor FGT do independently correlate with breast cancer risk in non-high-risk patients. Our model retained only age as an independent risk factor for breast cancer in this setting.