Research Presentation Session
05:51P. Kapetas, Vienna / AT
Diffusion-weighted imaging (DWI) and acoustic radiation force impulse (ARFI) elastography are techniques investigating microstructural tissue properties. We hypothesised that the information provided by ultrasound-acquired ARFI and magnetic resonance imaging (MRI)-acquired DWI for breast lesion characterisation is largely redundant, and that ARFI could be applied similarly as DWI, skipping the additional MRI scan.Methods and materials:
60 patients (18-84 years, mean 51.9), each with one breast lesion, underwent both breast ultrasound including ARFI and MRI including DWI. For each lesion, maximum shear wave velocity (SWV) and mean apparent diffusion coefficient (ADC) were recorded. Histology was the reference standard. Diagnostic performance was assessed with receiver operating characteristics (ROC) curve analysis. Independent predictors of malignancy were evaluated using multivariate logistic regression. Correlation between SWV and ADC was assessed by the Pearson correlation coefficient (r).Results:
Both ARFI and DWI showed high diagnostic performance in differentiating benign (n=15) from malignant (n=45) breast lesions as measured by the area under the ROC curve (0.817 and 0.863 respectively, p=0.535). The corresponding sensitivity and specificity for each modality was 88.9%, 73.3%, 80%, and 86.7%. Correlation analysis demonstrated a significant (p<0.001) moderate (r=-0.492) negative correlation between SWV and ADC. Multivariate logistic regression showed that the combination of ARFI and DWI did not improve the results of the single modalities.Conclusion:
There is significant correlation between ultrasound-acquired SWV and MRI-acquired ADC values of breast lesions. Since both modalities have a similar accuracy for the differentiation of benign from malignant lesions, application of ARFI elastography may obviate the need for an additional MRI examination including DWI for breast lesion characterisation.Limitations:
Limited patient number.Ethics committee approval
IRB-approved, retrospective study. The need for an informed consent was waived.Funding:
No funding was received for this work.
09:12E. Lee, Seoul / KR
To evaluate the diagnostic value of adding quantitative parameters of SWE and SMI to breast US in differentiating benign and malignant breast masses.Methods and materials:
200 pathologically proven breast lesions in 193 patients were retrospectively reviewed using breast US with greyscale, SWE, and SMI. The AUC in the combination of BI-RADS, Emax, Eratio, and SMIVI were compared using calculated cut-off values for two sets of cases. Set 1 included only downgrade C4a. Set 2 included upgrade of C3 and downgrade of C4a. The diagnostic performance was compared between greyscale US and US with the additional combination of quantitative parameters.Results:
There were 115 (57.5 %) benign lesions and 85 (42.5%) malignant lesions. Emax with a cut-off value of 84.45 kPa had the highest AUC value compared to Eratioof 3.35% or SMIVI of 3.57%. In set 1, the combination of BI-RADS with Emaxor SMIVI had a significantly higher AUC value compared to BI-RADS alone (p<0.05). BI-RADS with the addition of Emax, Eratio, and SMIVI had the best diagnostic performance compared to BI-RADS alone, with an accuracy of 84.0%. The specificity increased significantly from 46.1% (53 of 115) to 79.1% (91 of 115), and the sensitivity was slightly reduced but not statistically significant.Conclusion:
Combining all quantitative values of SWE and SMI with BI-RADS improved the diagnostic performance in differentiating between benign and malignant breast lesions.Limitations:
This study was a retrospective design and inter- or intraobserver variability was not evaluated. We only investigated the quantitative values of SWE and SMI.Ethics committee approval
This retrospective study was approved by the institutional review board for ethical issues in clinical research.Funding:
This work received Soonchunhyang University research funding.
05:43S. Heywang-Koebrunner, Munich / DE
Population-based centralised screening programs to date rely on mammography-screening only. However, even with latest digital technology, sensitivity drops strongly in the upper 5-15% density range. Feasibility and outcome of quality-assured density-adapted mammographic-sonographic screening is tested.Methods and materials:
Density is prospectively calculated on approximately 240,000 mammograms to select 30,000 patients with a density exceeding the upper 15% density range. These women are offered participation. Mammographic 1st, 2nd, and consensus readings are followed by independent combined mammographic-sonographic 1st, 2nds and consensus readings.Results:
For women with breast density > upper 15% range, incremental cancer detection, incremental false-positive rate, biopsy rate, and costs will be recorded. Acceptance by patients, feasibility for screeners, and quality assurance will be tested. Incremental detection of at least 60 cancers is expected.Conclusion:
A significant impact on screening effect is expected for women with very high breast densities.Limitations:
Due to wide availability of ultrasound, the ethical requirement of patient information in a randomised controlled trial design is impossible. Two different control groups are possible: either a historical group or a prospective parallel control group (completely anonymised) from women in the same SUs (selected by mammography unit or date).
Effect on interval cancer rate and stage distribution in the follow-up round will be evaluated after completion of the follow-up phase.Ethics committee approval
The study is payed through the National German Innovation Fund.
06:44S. Messina, Palermo / IT
To assess the role of digital breast tomosynthesis (DBT) and MRI in detection and loco-regional presurgical staging (LoPreSS) of breast cancer (BC).Methods and materials:
78 women (age: 32-76), undergone both DBT and breast MRI for biopsied suspicious breast lesions or for equivocal findings at conventional imaging, were divided into two groups on the basis of mammographic density (MD), in non-dense (A-B) and dense (C-D) breasts, and two groups according to the background parenchymal enhancement (BP), 1-2 and 3-4. Sensitivity, specificity, negative and positive predictive values (NPV and PPV), and accuracy in LoPreSS in unifocal, multifocal, multicentric and bilateral BC, were calculated.Results:
28/78 lesions were benign and 48/78 malignant (17/48 unifocal, 9/48 multifocal, 20/48 multicentric, 2/48 bilateral). MRI showed higher sensitivity (100% vs 87,5%, p=0.0013), specificity (76,7% vs 46,7%, p=0.0001), PPV (87,3% vs 72,4%, p=0.0208), and NPV (100% vs 70%, p<0.0001) than DBT. MRI showed a significantly higher accuracy in LoPreSS than DBT for multicentric (100% vs 45%, p=0.0001), multifocal (100% vs 55%, p=0.0274), and bilateral (100% vs 50%, p=0.3173), but not for unifocal (82,3% vs 64.7%, p=0.2520) BC. 20/48 BC patients showed non-dense and 28/48 dense breasts. DBT showed higher accuracy in LoPreSS in the non-dense group (17/20- 85% -vs 9/28-32.1%-p=0.0081). With MRI, no differences in LoPreSS between non-dense and dense breast were noted (95% vs 92,8 %, p=0), as well as between BPE 1-2 group and 3-4 group (93,5% and 94,1%, p=0.93) respectively. Among non-dense breasts, MRI and DBT showed no significant differences in preoperative staging accuracy (95% vs 85%; p=0.30).Conclusion:
In BC, LoPreSS breast MRI showed higher accuracy in dense breasts than DBT. No significant differences in non-dense breasts were noted.Limitations:
Single-centre study.Ethics committee approval
No funding was received for this work.
05:56P. Bakic, Philadelphia / US
Mechanical imaging (MI) maps tissue stiffness, which increases in breast tumours. The simultaneous acquisition of digital breast tomosynthesis (DBT) and MI combines the information about the anatomical structure and tissue stiffness of breast findings. This combination has the potential to reduce the number of false-positives and help avoid biopsies by resolving ambiguous DBT findings based upon their MI data. In simultaneous imaging, the MI pressure sensor is present during the DBT acquisition, producing visible artefacts. These artefacts may be reduced by preprocessing DBT projections, modifying the reconstruction algorithm, or designing radiolucent sensors. In this paper, we characterised artefacts seen in wide-angle DBT images with and without preprocessing.Methods and materials:
We have used a clinically available wide-angle DBT system to image a structured physical breast phantom on top of a commercial MI sensor. DBT images were acquired using the standard clinical protocol and automatic exposure control. We preprocessed DBT projections by flat fielding, using sensor images with and without the phantom. Images obtained with and without flat fielding were assessed visually by experienced medical physicists.Results:
Simultaneous DBTMI acquisition shows sensor artefacts in reconstructed images within ~1/3 of the phantom thickness above the detector. These artefacts are less prominent as compared to our previous study of simultaneous DBTMI using a narrow-angle DBT system. Flat-fielded images show substantially reduced appearance of artefacts, with slightly increased noise.Conclusion:
The simultaneous DBTMI imaging removes the requirement for two separate acquisitions. However, it produces noticeable artefacts due to the presence of the MI sensor. These artefacts overlap breast anatomical structures and may compromise diagnosis. Flat fielding of DBT projections substantially eliminates artefacts. Quantitative validation and method optimisation is ongoing.Limitations:
n/aEthics committee approval
Supported by the H2020 Marie Skłodowska-Curie Fellowship #846540.
05:58N. Anandan, Leeds / UK
The impact of DBT on dose when assessing masses.Methods and materials:
All women recalled between 13/11/15 and 29/07/16 following an abnormal screening mammogram had triple assessment and biopsy if required.
A DBT study was read within 6 weeks of assessment clinic. The number of additional mammographic views, doses (mGy), and DBT examination was recorded.
Statistics: Doses analysed in a mixed design, 2x2 ANOVA, looking at dose with and without DBT (within cases).Results:
1,470 women attended screening assessment and 835 women consented. 810 had complete data on dose and screening outcome.
353 cases recalled for a mass (351 women) and 316 biopsies performed (314 women). 74 cancers identified.
Tomosynthesis/ultrasoud recommended 97 biopsies and 73/74 cancers would have been identified with one case of DCIS missed.
Significant effect of use of DBT, F(1,339) = 4.80, p = .029, ω = .11 (within cases): mean dose without DBT was higher than with DBT.
When a biopsy was undertaken that tomosynthesis/ultrasound indicated should not have taken place, mean dose was higher without DBT.
When a biopsy was taken that tomosynthesis/ultrasound agreed should have taken place, there was no significant difference in dose without/with DBT.
Reduction in the number of films used, associated with the use of DBT, was analysed in a one way, between cases ANOVA looking at whether cases were biopsied or not. The overall reduction in the number of films used, associated with the use of DBT, was 3.61 (SD = 2.12).Conclusion:
DBT in assessment has the potential to reduce the dose and number of additional mammographic images in diagnostic workup of masses.Limitations:
This was a single-centre study.Ethics committee approval
IRB approved the prospective study.Funding:
No funding received for this work.
08:42M. van Lier, Amsterdam / NL
Mammographic breast compression using compression paddles equipped with a pressure indicator (pressure-based compression) has not yet been reported for flexible paddles in digital breast tomosynthesis (DBT). We aimed to evaluate the effect of a new DBT pressure-based flexible paddle (DBT-PBFP) on the technologist and patient experience. The impact on radiation dose and compression parameters variability was also addressed.Methods and materials:
Women with a DBT appointment, agreeing with study participation, received PBFP-guided breast compression on a GE-Senographe-Pristina. Mean compression pressure (force over contact area) was visualised during compression to the technologist and participant using 8 LED’s (2kPa per LED), aiming for an 8-14kPa target range. Compression parameters were retrospectively obtained from the study examinations and priors. The participant and technologist experiences were assessed using a questionnaire.Results:
From 94 participants, four-views conventional prior and study DBT examinations (total: 376 paired-views) and completed questionnaires were available. Technologists indicated that DBT-PBFP use eased explaining breast-compression (100%) and improved their workflow (97%). From all women, 89% would recommend DBT-PBFP-compression to friends. The compression force (mean±standard deviation) did not differ in CC-views (prior:82.3±20.6N, PBFP:86.4±34.2N, p=0.11), but increased in MLO-views (prior:85.2±21.9N, PBFP:103.1±31.5N, p<0.0001). Compression pressure variability decreased with 50% (p<0.0001) for CC-views and 34% (p<0.0001) for MLO-views, along with a decreased breast thickness (CC-view:8%, p<0.0001, MLO-view:10%, p<0.0001). Sub-group analysis (100 paired-views) showed overall average glandular dose reduction (prior:1.65±0.38mGy, PBFP:1.59±0.37mGy), with MLO-views as main contributor.Conclusion:
Using the new DBT-PBFP improved the participant and technologist experience and workflow while significantly reducing compression pressure variability, breast thickness, and average glandular dose.Limitations:
In some patients, a different DBT-system and/or technologist performed the study compared to prior examination, which may impact on the patient-comfort and radiation-dose comparability.Ethics committee approval
Institutional review-board approved. Written informed-consent.Funding:
No funding was received for this work.
06:04I. Akin, Izmir / TR
Breast conserving surgery (BCS) has become a routine procedure in early-stage breast cancer. Because of the risk of local recurrence, the most important aim of BCS is the removal of a malignant mass with no tumour margin. However, it has been shown that surgical margin negativity rates are still low. The sensitivity of specimen mammography (SMG) in terms of surgical margin negativity is low. Specimen tomosynthesis (STS) has increased the diagnostic efficacy lesion detection. The aim of this study was to determine the efficacy of STS added to SMG in the intraoperative process to evaluate the surgical margin in excised tissue.Methods and materials:
SMG and STS findings of BIRADS 4-5 specimens were evaluated retrospectively by two different observers. The observers were unaware of the histopathological outcome during the evaluation process. The SMG and the STS findings were compared in terms of BIRADS classification and histopathology. Inter- and intraobserver correlation analyses were performed. Chi-square test and compatibility Kappa (K) analysis were performed.Results:
89 patients were included in this study. SMG-STS findings for observer 1 was K 0,80 (excellent) and for observer 2 was K 0,77 (good). Interobserver Ks were 0.48 and 0.43 (moderate), respectively. Sensitivities in SMG were 45% and 23%, the specificities were 90% and 74%, PPVs were 88% and 70%, NPVs were 50% and 42%, respectively. The sensitivities in the STS were 53% and 43%, specificities were 90% and 78%, PPVs were 87% and 48%, and NPVs were 61% and 49%, respectively.Conclusion:
In BCS, using STS in the intraoperative process provides the optimal evaluation of the surgical margin, a shortening of the duration of surgery, and a decrease in re-excision rates.Limitations:
Limited patients number.Ethics committee approval
No funding was received for this work.
06:38S. Iles, Halifax / CA
To identify which mammographic image positioning errors are associated with patient and physical acquisition parameters in the general screening population.Methods and materials:
Image quality assessments were performed on mammograms of 9259 subjects included in a population-based case-control study. The following mammography positioning errors were assessed (densitas qualityaiTM) on MLO and CC images from each subject's most recent screening exam: exaggeration, skin folds, portion cut-off, patient-related artefacts, posterior tissues missing, improper pectoralis muscle positioning, inadequate IMF, sagging, and proper height on image receptor. Logistic regression was used to model the presence of positioning errors as a function of patient/physical acquisition parameters including age, breast volume, and compression pressure partitioned into deciles. The Mantel-Haenszel test for linear trend was used to assess a dose-response relationship between the risk of errors and patient/physical acquisition parameters.Results:
When examining individual patient/physical acquisition parameters, patient age was found to be linearly associated with 10 of the 13 errors (p < 0.05), and breast volume and compression pressure were each found to be associated with 12 of the 13 errors (p < 0.05).Conclusion:
This study shows that patient and physical acquisition parameters, such as age, breast volume, and compression pressure, are related to elevated or diminished risk of mammography image positioning errors during a screening exam.Limitations:
This study requires external validation.Ethics committee approval
REB Approval was received and a waiver of consent was granted.Funding:
There was no funding received for this specific study, but the images used in this study were obtained from a previously funded study.
05:33S. Sampangi, Bangalore / IN
To identify and characterise breast lesions with micro-calcifications using ultrasound elastography.Methods and materials:
All patients with breast lesions and micro-calcifications on digital mammogram were subjected for ultrasound and elastography for further characterisation of the lesion. Siemens ACUSON S2000 machine with 14L5 and 9L4 probes were used for the ultrasound scan. Greyscale images of the lesions were described as per ACR BIRADS category. Elastography with VTI images were obtained in and around the lesion. VTI showed the lesion as soft or hard compared to adjacent parenchyma. Elastography/B Mode Ratio was calculated. E/B Ratio of 1 or <1 was considered as benign and > 1 was considered as malignant. Micro-calcifications with hypoechoic lesions were visualised, however in absence of hypoechoic mass, micro-calfications werenot seen in greyscale. VTI images showed micro-calcifications as bright dot like appearence in a hard tumour background along with E/B Ration > 1.Results:
Out of 105 breast lesions with micro-calcifications on a mammogram, the ultrasound scan could identify micro-calcifications in only 20 lesions. However, elastography identified 102 cases. On elastography, the micro-calcifications showed as bright spots in a dark or hard tumour background. Another 3 lesion showed it as a harder area. This gives confidence for radiologists to identify lesions even in the absence of a hypoechoic mass in the breast. Cases of DCIS without mass lesion are generally difficult to identify on the greyscale, but VTI will detect it as a hard area with bright specks of calcifications, like a starry sky appearance. These are also seen along the duct wall and cyst wall in fibro-cystic disease.Conclusion:
Elastography will detect micro-calcifications even in the absence of mass lesions on greyscale.Limitations:
Necrotic changes in breast lesion will be mask micro-calcification on elastography.Ethics committee approval
No funding was received for this work.
07:24 Neha, Moradabad / IN
Fibrocystic breast disease or fibroadenosis is a benign disease of the breast characterised clinically by mastalgia, firm palpable regions in the breast, and a suspicious palpable lump. Mastalgia is characteristically premenstrual and is partially or completely relieved with menstrual flow associated with decreased firmness. High-resolution ultrasonography (HRUS) is being utilised for differentiating benign from malignant breast diseases. Recently, sonoelastography, especially acoustic radiation forced impulse (ARFI) imaging, is being evaluated in breast imaging.Methods and materials:
This pilot study aimed to evaluate sonoelastography in fibrocystic breast disease and ARFI in differentiating normal from fibrocystic breast parenchyma.
30 breasts, 15 each with normal and clinically-suspected fibrocystic disease, were evaluated by sonoelastography using a high-resolution, linear-array transducer. Data from virtual touch tissue quantification (VTTQ) was collected from fibroglandular parenchyma in the axillary tail and superolateral quadrant of the breast in both normal and clinical disease groups, as fibrocystic breast disease is more common in the superolateral quadrant. The data obtained from the control group was then compared with that from clinically suspected cases of fibrocystic breast disease. The patients with a diagnosis of fibrocystic breast disease on sonoelastography were followed-up clinically or with tissue diagnosis.Results:
Sonoelastography picked up 10 out of 12 clinically/tissue-proven cases with elasticity values lower than that of the normal breast (usually less than 2m/sec)Conclusion:
Sonoelastography (both VTI and VTTQ) shows highly accurate results allowing for the confident diagnosis of fibrocystic disease, thus obviating the need for tissue diagnosis. Given the high prevalence of benign fibrocystic disease of the breast and high accuracy of sonoelastography, imaging can be used with high confidence in not only reaching the diagnosis but also in monitoring the treatment of disease, thus optimising the cost and duration of treatment.Limitations:
n/aEthics committee approval
No funding was received for this work.
Monique van Lier
Leeds, United Kingdom
Isil Başara Akin
Sian E. Iles
Eun Ji Lee
Seoul, Korea, Republic of
Sylvia H. Heywang-Köbrunner
Philadelphia, United States