Research Presentation Session
06:12M. Eriksson, Stockholm / SE
Purpose:
Mammography screening reduces breast cancer mortality but a large proportion of breast cancers are missed and detected at later stages, or develop in between screening intervals. We developed the KARMA model, taking the risk of breast cancer into consideration. This model identifies women who are likely to be diagnosed with breast cancer before or at the next screen.
Methods and materials:The study was based on the KARMA cohort, a prospective screening cohort including 70,877 participants. We identified 974 incident cancers and sampled 9,376 healthy individuals from the cohort. Risk scores were developed using mammographic features (density, masses, microcalcifications) and their asymmetry, age, menopausal status, family history of breast cancer, body-mass index, hormone replacement therapy, tobacco and alcohol, and a polygenic risk score including 313 SNPs. The KARMA score was developed from the risk scores using age stratified logistic regression.
Results:The full model reached an area under the curve of 0.77 (95% CI 0.76, 0.79) with good model fit (Hosmer-Lemeshow = 0.2). There was an 8-fold difference in risk between the 8% of women at high risk and at general risk. Women identified as high risk were more likely to be diagnosed with more aggressive cancers. The image-based model was validated in two independent cohorts.
Conclusion:By combining mammographic features, lifestyle factors, family history, and a polygenic risk score, we generated a model that identifies women with a high likelihood of being diagnosed with breast cancer within two years and in need of supplementary screening.
Limitations:Limited to model 1 validation.
Ethics committee approvalKarolinska Institutet dnr 2010/958-31/1, 2016/2600-31. Lund dnr 2009/770.
Funding:Märit and Hans Rausing’s Initiative Against Breast Cancer. The Kamprad Foundation.
05:27V. Pasqualino, Padova / IT
Purpose:
To compare contrast-enhanced spectral mammography (CESM) with breast MRI and in a women population at high and intermediate risk of developing breast cancer.
Methods and materials:59 women at high and intermediate risk for breast cancer, aged above 35 years, without known allergies to iodine and gadolinium contrast-agents, were prospectively recruited. CESM and MRI were performed with a minimum 72-hour interval to avoid nephrotoxicity. CESM images were evaluated independently by two breast radiologists, while MRI studies were interpreted by two different independent breast radiologists. Background parenchymal enhancement (BPE) and breast density (BD) were scored in 4 classes, while possible lesions were classified according to BIRADS scale. Sensitivity, specificity, positive-predictive-value (PPV), and negative-predictive-value (NPV) were compared by modality, and inter- and intra-modality agreement evaluated.
Results:Sensitivity was 100% for both modalities, while specificity was 81.1% with CESM and 71.6% with MRI. PPV was 76.6% with CESM and 66.4% with MRI, while NPV was 100% for both modalities. Neither specificity nor NPV differences were significant. The agreement for BPE and BD was from good to excellent for both modalities. The agreement between CESM and MRI for BPE was fair (k=0.34), probably due to the different phase of the ovarian cycle at the time of the two examinations.
Conclusion:CESM showed the same sensitivity as MRI, with some benefits in terms of specificity, even if not significant.
Limitations:Small sample size (preliminary results). BPE inter-modality disagreement likely influenced by the possible different phase of the ovarian cycle at the time of examinations.
Ethics committee approvalStudy approved by the IRB. Patient informed consent collected.
Funding:Research project.
05:54G. Gennaro, Padova / IT
Purpose:
To estimate the amount of radiation dose associated with contrast-enhanced spectral mammography (CESM).
Methods and materials:CESM is a dual-energy technique consisting in the sequential acquisition of two images, one at low-energy, equivalent to a standard mammogram, and the second at higher energy, both after the injection of iodine contrast agent. The two paired images are recombined to provide a “subtraction” image where possible lesions are contrast-enhanced. CESM was used within a clinical study comparing the clinical performance of CESM and breast MRI for the screening of women at high and intermediate risk of breast cancer. Patients enrolled in the study underwent both bilateral CESM in two views and breast MRI. Dosimetric data from the first 60 CESM examinations (486 images) was extracted from the DICOM header of all low- and high-energy image pairs acquired for each patient. Mean glandular dose (MGD) per-view was compared between low- and high-energy images and the increase in MGD compared to standard mammography was calculated.
Results:The mean compressed breast thickness was 50.8±15.8 mm. The average total MGD per view was 2.22±0.44 mGy; 70% due to the low-energy image, 1.53±0.42 mGy, obtained with the same exposure parameters as for a standard mammogram, while the remaining 30% of the dose, 0.70±0.07 mGy, was due to the high-energy image. Compared to standard mammography, the mean dose increase associated with CESM was 48%. For any breast thickness per-view, CESM dose was either within or below MGD limiting values established for both standard mammography and tomosynthesis.
Conclusion:The modest increase in radiation dose by CESM should not be an obstacle for its use as a screening test for women at high and intermediate risk of breast cancer.
Limitations:Relatively small sample size.
Ethics committee approvalStudy approved by the Institutional Ethics Committee.
Funding:Regional Healthcare System.
07:40R. Faermann, Gan / IL
Purpose:
Recent studies have shown that BRCA mutation carriers are prone to earlier onset of DCIS and invasive breast cancer with a higher prevalence of DCIS in BRCA carriers than non-carriers. Most of BRCA-associated tumours have DCIS present, favouring the existence of a premalignant pathway. However, DCIS in those studies was diagnosed with mammography (calcifications).
MRI screening has become an important tool for screening and early detection in BRCA carriers. MRI can detect DCIS even if non-calcified on mammography.
The aim of our study was to analyse BRCA patients with DCIS on MRI biopsy and the MRI presentation.
Methods:A retrospective study of BRCA patients with pure DCIS diagnosed with MRI biopsy between 2015 until 2019 at the Sheba Medical Center. All MRI and mammography studies were analysed by a fellowship-trained radiologist.
Results:950 BRCA carriers underwent surveillance MRI. 22 had a diagnosis of DCIS on MRI biopsy. 14 patients (64%) were BRCA1 carriers, while 8 patients (36%) were BRCA2 carriers. The median age of BRCA1 patients was 40 years old, which was lower than of BRCA2 patients (66 years old) (p<.05). The most common MRI presentation was non-mass enhancement. BRCA 1 patients were more commonly hormone receptors negative (9 patients, 64%) and had high-grade DCIS (100%), while BRCA2 patients were more commonly hormone receptors positive (5 patients, 63%) and had more commonly intermediate grade DCIS (5 patients, 63%).
Conclusion:MRI detected non-calcified DCIS is more common in BRCA1 gene-mutation carriers and presents earlier than in BRCA2 patients, usually as a high-grade disease. This finding is opposite to calcified DCIS detected on mammography, which is more common in BRCA2 carriers.
Limitations:Single-centre only, retrospective study.
Ethics:n/a
Funding:No funding was received for this work.
07:19S. Astley, Manchester / UK
Purpose:
To quantify breast density from low-dose mammograms and hence facilitate screening stratification for young women.
Methods and materials:We analysed routine screening mammograms from women in the Predicting Risk of Cancer at Screening study (PROCAS) and from a study of Automated Low Dose Risk Assessment Mammography (ALDRAM) in which both standard and low-dose mammograms (one-tenth the original dose) were acquired. We simulated low dose mammograms in a PROCAS case-control set of 335 prior mammograms of women with interval and subsequent screen-detected cancers, each matched with three controls on age, BMI group, menopausal status, and parity. We developed an automated breast density method using a convolutional neural network trained on the average visual assessment of two readers who recorded percent mammographic density on visual analogue scales (VAS). With this, we predicted density in the 1,340 simulated low-dose PROCAS images and in the low-dose and standard mammograms from 148 women aged 31-45 in ALDRAM. For the case-control set, we computed odds ratios (OR) of developing breast cancer between the highest and lowest quintiles of predicted density. We also calculated the per-breast Pearson correlation between predicted density on standard and low dose mammograms in ALDRAM.
Results:The OR between the highest and lowest quintiles of mammographic density in the simulated low-dose case-control set was 3.57 (95% CI 2.25-5.68). In the ALDRAM cohort, the correlation between density in standard mammograms and their low-dose equivalents was 0.987.
Conclusion:Automated assessment of breast density in mammograms taken with one-tenth of the usual dose is feasible and has the potential to be used for stratification of young women in personalised breast screening programmes.
Limitations:Only GE mammograms were used.
Ethics committee approvalNHS-REC ref:19/NW/0037.
Funding:MRC-CiC award.
06:29J. Browne, Barcelona / ES
Purpose:
Previous studies of breast density (BDen) suggest a 4 to 6-fold increase in breast cancer risk (BCR) for high density relative to non-dense breasts. However, many incidental factors have to be accounted for in these studies: family and personal history, BMI, obstetric history, hormonal treatment, and age.
Our objective was to compare glandular volume (GVol), breast volume (BVol), and breast density (BDen) in each breast of women with unilateral breast cancer (UBC) versus no cancer contralateral breast (NCCB), and determine if BDen is an independent BCR.
Methods and materials:We reported the results of GVol, BVol, and BDen measured during clinical practice in 216 women with unilateral invasive cancers from 4/2014 to 12/2015.
All women had mammography performed with photon counting equipment. The detector separates photons in low and high energy photons, allowing spectral determination of GVol, BVol, and calculates BDen.
Cancers were detected using mammography, ultrasound, or MRI.
Results:Of the 216 cases, 104 were right NCCB/left UCB and 112 were left NCCB/UCB. Mean BDen was 30.1% in UBC vs 28.8% for NCCB and mean GVol was 160.3cc vs 147,4cc, respectively (both P<0.0001). BVol differences were not significant (P= 0.199). In 117 cases, BDen was higher in UCB than NCCB, in 59 cases lower, and in 40 cases the same. In 131 cases, GVol was higher in UCB, while lower in 85.
Conclusion:Our results indicate that even with identical hormonal status, immunologic system, genetic background, BMI, and habits, higher BDen and higher GVol are significantly associated with a higher BCR. Our study shows that even clinically non-appraisable BDen and GVol differences are independently associated with a higher BCR.
Limitations:n/a
Ethics committee approvaln/a
Funding:No funding was received for this work.
10:39R. Lo Gullo, New York / US
Purpose:
To verify whether radiomics features extracted from MRI of BRCA-positive patients with breast masses smaller than 1 cm can differentiate benign from malignant lesions using model-free parameter maps.
Methods and materials:In this retrospective study, 96 BRCA mutation carriers (mean age at the time of biopsy = 45.5±13.5 years) as assessed with genetic testing and who had an MRI from November 2013–February 2019 that led to a biopsy (BI-RADS 4) or imaging follow up (BI-RADS 3) were included. Two radiologists assessed all lesions independently and in consensus according to the breast imaging and reporting and data system (BI-RADS) lexicon. Radiomics features, based on first-order statistics, the grey level co-occurrence matrix (GLCM), run length matrix (RLM), size zone matrix (SZM), neighbourhood grey level dependence matrix, and neighbourhood grey tone difference matrix, were calculated
Results:Consensus BI-RADS classification assessment achieved a diagnostic accuracy of 53.4%, a sensitivity of 73.1%, a specificity of 42.1%, a positive predictive value (PPV) of 40.5%, and a negative predictive value (NPV) of 76.2%. The model combining 5 parameters (age, lesion location, GLCM-based correlation from the pre-contrast phase, first-order coefficient of variation from the 1st post-contrast phase, and SZM-based grey level variance from the 1st post-contrast phase) achieved a diagnostic accuracy of 81.5%, a sensitivity of 63.2% (24/38), a specificity of 91.4% (64/70), PPV of 80.0% (24/30), and NPV of 82.1% (64/78).
Conclusion:Radiomics improves diagnostic accuracy compared to qualitative morphological assessment with BI-RADS classification alone in BRCA mutation carriers.
Limitations:Using only single-centre data. Relatively small sample size. We included small breast masses which constitute few pixels in the final ROI.
Ethics committee approvalIRB approved.
Funding:This work was partially supported by the NIH/NCI Cancer Center Support Grant (P30 CA008748) and the Breast Cancer Research Foundation.