Nicky B. Barr

Visual assessment of density in digital mammograms

This study compares visual assessment of density on full field digital mammograms using visual analogue scales (VAS) and written percentages Fifty normal digital screening mammograms were selected at random Nine readers viewed the images on two occasions, firstly indicating density on a VAS and then estimating the percentage of dense tissue in the breast Although the two methods were correlated, the degree of agreement between the density estimates varied considerably from reader to reader More experienced readers used a wider range of values, and inter-observer variability for both methods was higher for these readers The greatest difference between the methods was in mammograms with a mixed fatty-glandular appearance (density between 55% and 75%) Both methods are quick and convenient, although these results demonstrate a need for training to ensure they are used consistently by readers of different degrees of experience.

Same task, same observers, different values: the problem with visual assessment of breast density

The proportion of radio-opaque fibroglandular tissue in a mammographic image of the breast is a strong and modifiable risk factor for breast cancer. Subjective, area-based estimates made by expert observers provide a simple and efficient way of measuring breast density within a screening programme, but the degree of variability may render the reliable identification of women at increased risk impossible. This study examines the repeatability of visual assessment of percent breast density by expert observers. Five consultant radiologists and two breast physicians, all with at least two years’ experience in mammographic density assessment, were presented with 100 digital mammogram cases for which they had estimated density at least 12 months previously. Estimates of percent density were made for each mammographic view and recorded on a printed visual analogue scale. The level of agreement between the two sets of estimates was assessed graphically using Bland-Altman plots. All but one observer had a mean difference of less than 6 percentage points, while the largest mean difference was 14.66 percentage points. The narrowest 95% limits of agreement for the differences were -11.15 to 17.35 and the widest were -13.95 to 40.43. Coefficients of repeatability ranged from 14.40 to 38.60. Although visual assessment of breast density has been shown to be strongly associated with cancer risk, the lack of agreement shown here between repeat assessments of the same images by the same observers questions the reliability of using visual assessment to identify women at high risk or to detect moderate changes in breast density over time.

Volumetric and area-based measures of mammographic density in women with and without cancer

We compare mammographic density in 44 women with screen-detected breast cancer and a control group of 923 women with normal screening mammograms. Multiple regression was used to compare the effects of case-control status on breast density of the contra-lateral breast. Two breast density measures were investigated: the average visual assessment, recorded on a visual analogue scale (VAS), for the two views and two independent readers; and volumetric percentage density measured by QuantraTM. We adjusted for confounding factors of BMI, HRT use, age and menopausal status. Initially there was no significant difference in mean percentage density between cases and controls using either measure of density: VAS (cases 27.5%, controls 26.9%) and QuantraTM (cases 17.2%, controls 18.2%). However, when confounding factors were controlled for, case-control status had a statistically significant effect on breast density as measured by QuantraTM (adjusted means: cases 19.2%, controls 14.8%; p = 0.002) but not by VAS.

Factors Affecting Agreement between Breast Density Assessment Using Volumetric Methods and Visual Analogue Scales

Mammographic density in digital mammograms can be assessed visually or using automated volumetric methods; the aim in both cases is to identify women at greater risk of developing breast cancer, and those for whom mammography is less sensitive. Ideally all methods should identify the same women as having high density, but this is not the case in practice. 6422 women were ranked from the highest to lowest density by three methods: QuantraTM, VolparaTM and visual assessment recorded on Visual Analogue Scales. For each pair of methods the 20 cases with the greatest agreement in rank were compared with the 20 with the least agreement. The presence of microcalcifications, skin folds, suboptimally positioned inframammary folds, and whether or not the nipple was in profile were found to affect agreement between methods (p<0.05). Careful positioning during mammographic imaging should reduce discrepancy, but a greater understanding of the relationship between methods is also required.

Ethnic variation in volumetric breast density

Volumetric breast density was determined using QuantraTM (Hologic) in 1356 women undergoing routine breast screening. Self-reported ethnicity, age, HRT use, weight and height were also available. 1038 women declared themselves to be White (British or Irish), 71 Black, 77 Asian, 91 Jewish, 31 Mixed Race and 48 Other European. Most of the Jewish group were Ashkenazi, a group in which there is a high probability of genetic susceptibility to breast cancer. Women with screen-detected or previous cancers were excluded. The only significant difference in breast density found between ethnic groups was between the Jewish women and women of White (British or Irish) ethnicity, where mean volumetric densities were 19.61% and 16.89% respectively (p=0.012), however this difference is only of borderline significance (p=0.053) once adjustments are made for age, Body Mass Index (BMI) and use of Hormone Replacement Therapy (HRT). The Jewish women had on average a lower BMI and were more likely to have used HRT.

Longitudinal change in mammographic density and association with breast cancer risk: a case-control study

High mammographic breast density is associated with increased risk of breast cancer, but how risk varies with longitudinal change in density is less clear. To investigate, a case-control study of 30 women with screen-detected cancer and 30 women with a normal mammogram, all with two previous normal mammograms, was conducted. Percentage density for all mammograms was estimated with the thresholding software Cumulus. Mean density at first screen was not significantly different in cases and controls in contralateral (36.5 vs. 32.6, p = 0.23) or ipsilateral (36.0 vs. 32.9 p = 0.37) breasts, but mean reduction in density from first to third screen was significantly different in both contralateral (10.7 vs. 5.1, p = 0.02) and ipsilateral (11.7 vs. 6.2, p = 0.04) breasts. Using logistic regression, and controlling for age and HRT use, breast cancer risk was found to be associated with change in density from first to third screen.

Comparison of microcalcification detection rates and recall rates in digital and analogue mammography

21158 screening mammograms were obtained, 10024 acquired using full field digital mammography (FFDM) and 11134 acquired using film-screen mammography For each mammogram, data were collected on recall for further assessment due to detection of microcalcification, use of needle biopsy, and presence of microcalcifications in biopsy specimens 61.5% of women who had a core biopsy following digital mammography had microcalcifications detected, compared with 65.8% for analogue mammography but this difference was not significant (p=0.71) Rates of detection of microcalcifications in women screened by the two methods were similar It was also found that the recall rate for assessment for women screened digitally (6.1%) was significantly higher than the recall rate for those screened by analogue mammography (3.3%), 95% confidence interval 2.2% - 3.4% Screening using digital mammography leads to a higher recall rate for assessment than analogue mammography, although similar rates of detection of microcalcifications occur with both imaging modalities.

Mammography reading with computer-aided detection (CAD): single view vs two views

Two-view mammography is known to be more effective than one-view in increasing breast cancer detection and reducing recall rates. In addition, there is evidence that computer aided detection (CAD) systems are able to prompt malignant abnormalities that have been overlooked by a human reader. Using data from the UK NHS Breast Screening Programme (NHSBSP) we compared double reading with single reading using a CAD system, to assess the relationship between CAD and number of views in terms of the sensitivity of the screening regime to cancer detection and the recall rate of normal cases. CAD appeared to contribute to an increased cancer detection rate with single-view mammography without significantly increasing the recall rate. For two-view mammography, there was no significant change in sensitivity using CAD but a significantly higher recall rate. However, single-view mammography was used in incident rounds in which previous mammograms were available whereas two-view mammography was used in the prevalent round where no previous mammograms were available.

Reader performance in visual assessment of breast density using visual analogue scales: are some readers more predictive of breast cancer?

Mammographic breast density is one of the strongest risk factors for breast cancer, and is used in risk prediction and for deciding appropriate imaging strategies. In the Predicting Risk Of Cancer At Screening (PROCAS) study, percent density estimated by two readers on Visual Analogue Scales (VAS) has shown a strong relationship with breast cancer risk when assessed against automated methods. However, this method suffers from reader variability. This study aimed to assess the performance of PROCAS readers using VAS, and to identify those most predictive of breast cancer. We selected the seven readers who had estimated density on over 6,500 women including at least 100 cancer cases, analysing their performance using multivariable logistic regression and Receiver Operator Characteristic (ROC) analysis. All seven readers showed statistically significant odds ratios (OR) for cancer risk according to VAS score after adjusting for classical risk factors. The OR was greatest for reader 18 at 1.026 (95% Cl 1.018-1.034). Adjusted Area Under the ROC Curves (AUCs) were statistically significant for all readers, but greatest for reader 14 at 0.639. Further analysis of the VAS scores for these two readers showed reader 14 had higher sensitivity (78.0% versus 42.2%), whereas reader 18 had higher specificity (78.0% versus 46.0%). Our results demonstrate individual differences when assigning VAS scores; one better identified those with increased risk, whereas another better identified low risk individuals. However, despite their different strengths, both readers showed similar predictive abilities overall. Standardised training for VAS may improve reader variability and consistency of VAS scoring.

Inter and intra observer variability in a semi-automatic method for measuring volumetric breast density

If breast density is to be incorporated into breast cancer risk prediction models, the technique used for measurement must be quantitative, accurate, objective and reproducible We present a semi-automated method that has been used by three independent operators to measure glandular volume from the digitised mammograms of 29 women (116 images) Additionally, one operator used the method on 10 separate occasions on a sample of 24 images Intra-observer variability was found to be acceptably low, with coefficients of variation ranging from 3.5 – 5.7% depending on mammographic view (intra-class correlation coefficient close to 1 in all cases) However, inter-observer variability was greater with significant differences in glandular volume recorded between observers This was attributed to the method of breast edge detection The development of a new automatic breast edge detection algorithm has resolved the issue The average difference in glandular volume measurement between two independent operators in the cranio-caudal view is -0.89cm3 (95% confidence interval -2.77 – 0.99 cm3) using the new method, compared to 5.99cm3 (95% confidence interval 2.72 – 9.76 cm3) using the old method.