RPS 1002 - What is new in breast ultrasound?

Author Block: E. Nikolova, J. K. Weber, G. Zanetti, J. Wieler, T. Frauenfelder, M. Marcon; Zurich/CH
Purpose: To investigate the false positive (FP) and false negative (FN) rate over three years after implementation of an automated breast ultrasound (ABUS) system.
Methods or Background: In this IRB-approved retrospective study all women undergoing ABUS examination in our department between October 2015 and October 2018 have been revised (1995 women in total, mean age±SD, 48.9±13 years). We included all women who have undergone a follow-up of at least 24 months after ABUS examination or with BI-RADS category 4 or 5, who have undergone histological evaluation. FP cases were defined as cases classified BI-RADS 3, 4 or 5 in ABUS who have shown to be benign lesions during follow-up or after biopsy. FN cases were defined as all cases of cancer diagnosed during follow-up and that were retrospectively already visible in the ABUS examination. Descriptive statistics were used.
Results or Findings: 1219 women (51.9±11.0 years) were included and classification after ABUS examination was: BI-RADS 1 n=136 (11.1%), BI-RADS 2 n=890 (73.0%), BI-RADS 3 n=161 (13.2%), BI-RADS 4 n=15 (1.2%) and BI-RADS 5 n=17 (1.4%). Patients had a mean follow-up±SD of 54.3±16.7 months. A woman was diagnosed with an invasive carcinoma 14 months and another 24 months after ABUS examination classified BI-RADS 2; both lesions were already visible in previous ABUS and mammography and the "retraction phenomenon sign" was visible in the ABUS-coronal reconstruction. A malignancy was also diagnosed in 3/161(1.9%) BI-RADS 3 lesions, 7/15 (46.6%) BI-RADS4 lesions and 17/17 lesions (100%) BI-RADS 5.The FN rate was 6.9% (2/29 cases). The FP rate was 13.9% (166/1190 cases). Among the 14 malignant lesions, in one case the lesion was only visible in ABUS and not in mammography (7.1%); in three cases lesion was visible in both but multifocality only in ABUS (21.4%).
Conclusion: We found a FN rate of 6.9% and a FP rate of 13.9%. To avoid false negative cases ABUS coronal reconstruction should be accurately evaluated.
Limitations: This was a retrospective study.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethics committee approval was received; KEK 2016-00064.

Author Block: E. Nikolova, J. K. Weber, G. Zanetti, J. Wieler, T. Frauenfelder, M. Marcon; Zurich/CH
Purpose: The aim of this study was to investigate the learning curve after implementation of an automated breast ultrasound system (ABUS) over three years.
Methods or Background: In this IRB-approved retrospective study we included all women undergoing ABUS examination in our department between October 2015 and October 2018 who have undergone a follow-up of at least 24 months after ABUS examination or with BI-RADS category 4 or 5, who have undergone histological evaluation. The number of additional handheld ultrasound exams (AHHUS) for better definition of an ABUS finding and number of false positive (FP) cases were noted. FP were defined as cases classified BI-RADS 3, 4 or 5 in ABUS who have shown to be benign lesions during follow-up/after biopsy. AHHUSs and FPs were compared in the first 6 months versus second 6 months, yearly over the three years and considering presence/absence of previous ultrasound exam (UE). Chi-square test was applied.
Results or Findings: A total of 1223 women (mean age±SD, 51.9±11.0 years) were included: 346 (28.3%), 597 (48.8%) and 280 (22.9%) exams were performed respectively in the year 1, 2 and 3. 288/1223 (23.5%) had no previous UE. FP cases were 13/45 (28.9%) in the first 6 months and 70/301 (23.3%) in the second 6 months (p=.454). Cases with AHHUS were 16/45 (35.6%) in the first 6 months and 66/301 (21.9%) in the second 6 months (p=.038). Over three years FP were 83/346 (24.0%), 76/597 (12.7%) and 14/280 (5.0%) for year 1,2 and 3 respectively (p <.001). AHHUS were 82/346 (23.7%), 57/597 (9.5%) and 19/280 (6.8%) for year 1,2 and 3, respectively (p <.001). FP and AHHUS were more frequent in women without previous ultrasound exam (FP:21.2% versus 12.0%, p<.001 and AHHUS: 17.7% versus 11.4%, p <.001).
Conclusion: After ABUS implementation FP and AHHUS gradually reduced, especially after the first year; FP and AHHUS are more frequent in women without previous ultrasound exam.
Limitations: This was a retrospective study.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethics committee approval was given; KEK 2016-00064.

Author Block: L. A. Incardona¹, M. Franchini², P. Belli³, G. M. Giuseppetti⁴, G. P. Scaperrotta⁵, E. Cassano³, C. Di Maggio⁶, E. Montrucchio⁷, J. Nori¹; ¹Florence/IT, ²Pisa/IT, ³Rome/IT, ⁴Ancona/IT, ⁵Vimodrone/IT, ⁶Padua/IT, ⁷La Spezia/IT
Purpose: The P.I.N.K. study seeks to evaluate the enhanced diagnostic precision in breast cancer screening across Italy, achieved through various imaging technology combinations, with a particular emphasis on the added diagnostic value of Ultrasound (US) in detecting breast lesions, especially those undetected by Mammography (MX).
Methods or Background: Initiated in 2018, the study engaged 30,023 patients aged 40 and above, undergoing MX and at least US and/or Tomosynthesis (TS). Inclusion criteria encompassed patients with histopathologically-confirmed B3-B5 lesions and a positive US assessment (U4-U6), utilizing a complete diagnostic pathway of MX+TS+US or MX+US+TS. The primary objective was to ascertain the additional detection rate of US compared to other imaging techniques, stratifying cases detected solely through US by the MX risk category and age at diagnosis/mammographic breast density.
Results or Findings: Of the recruited, 883 had proven breast B3-B5 lesions, with 408 and 27 undergoing MX+TS+US and MX+US+TS pathways respectively. US added detection rate in the MX+TS+US pathway was 14.7%, increasing to 21.5% in benign MX (R2) cases and decreasing to 12% in negative MX (R1) cases. US demonstrated an elevated diagnostic rate in the youngest (40-49 years) and oldest (over 70 years) subgroups, and a higher detection rate (26.1%) in women with dense breasts.
Conclusion: The findings suggest that US could be a significant addition to the current breast cancer screening approach, especially where MX may not detect lesions, potentially improving early detection and treatment, thereby enhancing patient outcomes.
Limitations: No limitations were identified.
Funding for this study: No funding was provided for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: This was a retrospective study.

Author Block: G. Irmici¹, C. Depretto¹, A. Cozzi², G. Della Pepa¹, E. D’Ascoli¹, C. De Berardinis¹, A. Bonanomi¹, S. Marziali¹, G. P. Scaperrotta¹; ¹Milan/IT, ²Lugano/CH
Purpose: The aim of this study was to investigate if an artificial intelligence-based decision-support system (Koios DS) influences diagnostic performance of hand-held breast US performed by radiologists of different experience and can reduce the number of unnecessary breast biopsies.
Methods or Background: This prospective monocentric study enrolled consecutive patients referred for breast biopsy in a tertiary-level centre between May 2022 and January 2023. A junior (2 years of breast imaging experience) and a senior reader (20 years experience) assigned US BI-RADS categories before and after Koios application. Differences in diagnostic performance, taking biopsy as the reference standard, were assessed by comparing areas (AUC) under the receiver-operating characteristic curves (DeLong's test). The number of unnecessary biopsies (i.e. yielding benign results) before and after the application of Koios was compared with the McNemar's test.
Results or Findings: We enrolled 222 patients (median age 58 years, interquartile range 46–72) with 226 lesions (at biopsy: 137 malignant, 89 benign). The diagnostic performance of the junior reader (AUC 0.785, 95% confidence interval [CI] 0.736–0.836) was significantly improved (p <0.001) by Koios (AUC 0.864, 95% CI 0.819–0.908). Likewise, the diagnostic performance of the senior reader (AUC 0.823, 95% CI 0.777–0.869) was significantly improved (AUC 0.868, 95% CI 0.825–0.912, p <0.001). The diagnostic performance of the junior reader assisted by Koios (AUC 0.864, 95% CI 0.819–0.908) was higher than that of the senior reader alone (AUC 0.823, 95% CI 0.777–0.869), albeit non-significantly (p=0.053). The application of Koios was able to significantly reduce the number of unnecessary biopsies both for the junior (from 51.6% to 32.6%, p<0.001) and the senior reader (from 46.1% to 33.7%, p<0.001).
Conclusion: The Koios decision-support system is able to significantly improve the performance of junior and senior readers and reduce the number of unnecessary biopsies.
Limitations: This was a single-centre study within a high-experience setting.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: No information provided by the submitter.

Author Block: V. Marisi¹, M. C. Torrione², S. Hannan², F. Ricciardella², G. Piccolo², A. Di Credico², A. Figorilli², M. Muzi², M. Caulo²; ¹Vasto/IT, ²Chieti/IT
Purpose: The aim of the study was to analyse the concordance between radiological and histopathological characteristics of breast lesions, demonstrate the high predictive value of the BIRADS (Breast Imaging-Reporting and Data System) classification and asses managment, surgical treatment and follow-up.
Methods or Background: In this retrospective study, we analysed the discrepant findings between imaging and pathological findings in patients undergoing breast biopsy between January 2022 and April 2023. 857 patients with 957 lesions undergoing ultrasound-guided breast biopsy were included in this study. Exclusion criteria were BIRADS4 b having been performed on patients or biopsies with another first-instance method (stereotaxis or MR-guided biopsy).
Results or Findings: 52 discordant biopsies were evaluated by two experts in breast radiology. 32 classified as BIRADS3 or 4a were found to be B5 on histological examination (2 infiltrating lobular carcinoma, 8 ductal carcinoma in situ, 16 invasive carcinoma NST, 2 infiltrating ductal carcinoma, 1 papillomatosis, 1 infiltrating papillary carcinoma and 2 invasive G2 carcinoma); 20 biopsies classified BIRADS4c or 5 were B1 (4), B2 (10) and B3 (6). According to discordance radiological and histopathological findings, 2 patients with B1 results underwent a new biopsy or VAAB: the result was B5 (1) and B3 (1); 5 patients with B2 result underwent VABB or surgical excision: the result was B5 (4 patients) and B2 (1 patients).
Conclusion: The BIRADS classification is fundamental in breast cancer diagnosis. When there are discrepancies between imaging and histological results a discussion by an interdisciplinary team is mandatory: this is crucial in clarifying why they exist and to re-evaluate the most suitable methodology in continuing the diagnostic process. In our series, 52% of suspicious lesions with a benign result were found to be malignant after rebiopsy or surgery.
Limitations: No limitations were identified.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: No information provided by the submitter.

Author Block: P. Kapetas¹, C. Fürböck¹, R. Aggarwal², B. Altuwayjiri³, P. Clauser¹, T. H. Helbich¹, G. Langs¹, P. A. Baltzer¹; ¹Vienna/AT, ²Southend-On-Sea/UK, ³Riyadh/SA
Purpose: The study aimed to evaluate whether pre-therapeutic breast ultrasound (US) can serve in the prediction of breast cancer (BC) patients non-responding to neoadjuvant chemotherapy (NAC) and to compare two different models.
Methods or Background: This retrospective, IRB-approved study included 245 patients with histologically confirmed BC undergoing NAC. The dataset was divided into a training (165 cases) and a validation set (80 cases). A representative B-mode US image of each tumour from the pre-treatment examination was selected. Two experienced breast fellows independently evaluated the lesions using standard BI-RADS descriptors. Logistic regression was used to identify independent predictors of NAC response in the training set and create a model. Additionally, a Resnet18-based neural network with Dropout layers to decrease the amount of overfitting was trained to predict the treatment outcome. The performance of both models was evaluated on the validation set using descriptive statistics. Postoperative histology was the standard of reference for treatment response.
Results or Findings: 145 patients (59.2%) did not achieve a pathological complete response. From the BI-RADS descriptors, oval or round shape, microlobulated or spiculated margin and the presence of calcifications or edema proved to be independent predictors of pCR. A model using these showed an accuracy, sensitivity, specificity, positive and negative predictive value of respectively 65%, 82%, 41%, 67% and 61% for the prediction of non-responders to NAC. Compared to that, the DL-based model achieved an accuracy, sensitivity, specificity, positive and negative predictive value of 76%, 88%, 59%, 76% and 78%.
Conclusion: Breast US can accurately predict lack of response to NAC for BC patients prior to its initiation. A DL model using images from the baseline US examination demonstrates an increased diagnostic performance as compared to standard B-mode BI-RADS descriptors.
Limitations: Monocentric study, small patient number.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethics committee - additional information: This study was approved by the Medical University of Vienna (Ref. Number: 1793/2013)

Author Block: C. Lee, G. Hesley, M. Urban, M. Piltin; Rochester, MN/US
Purpose: The purpose of this clinical trial is to evaluate the safety and twinkling signature of the marker after months of NST. Ultrasound detection of commercial breast biopsy markers is challenging, particularly in the axilla in patients with node-positive breast cancer responding favorably to neoadjuvant systemic therapy (NST). We have developed a biopsy marker that demonstrates robust ultrasound color Doppler twinkling.
Methods or Background: In an ongoing, prospective, single-arm trial involving 10 patients with node-positive breast cancer, a twinkling marker was implanted near the conventional biopsy marker (control) in the positive axillary lymph node before the start of NST. Ultrasound was performed at baseline, during NST, and preoperatively to assess for twinkling and B-mode conspicuity of the twinkling and commercial markers. Based on the preoperative ultrasound, the surgeon determined whether I-125 seed localisation of the positive node was necessary or if intraoperative ultrasound guidance, as part of their surgical practice, would be used to retrieve the marked positive node.
Results or Findings: Eight of 10 patients with a mean age of 49.5 years (SD 14.1) have enrolled. Each patient has a different commercial biopsy marker in the positive node. Thus far, four of the eight patients had ultrasounds during NST, and twinkling of the twinkling marker was readily present on two different ultrasound scanners (GE Logiq and Fujifilm Arietta). No adverse events. Three patients have completed targeted axillary dissection, and the surgeon successfully used ultrasound twinkling intraoperatively to identify the positive node. None of the three patients required preoperative axillary radioactive seed localisation.
Conclusion: Early preliminary results are promising for identifying a marker by ultrasound Doppler twinkling after NST, possibly obviating preoperative localisation.
Limitations: The small study size and involvement of a single surgeon are limitations.
Funding for this study: Funding was received from Mayo Clinic President's Discovery Translation Program.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study received approval from the Mayo Clinic Institutional Review Board, and written consent was obtained from all participants. This study is registered under ClinicalTrials.gov NCT0554347.

Author Block: C. Boldrini, R. Dattoli, A. Marra, S. Amodeo, P. Belli, R. Manfredi; Rome/IT
Purpose: This study aims to evaluate, in a population of pediatric patients diagnosed with premature thelarche (PT) and central precocious puberty (CPP), the correlation existing between the clinical stage of breast development according to Tanner and the ultrasonographic stage of the mammary gland, both at the diagnosis and after pubertal “blocking” therapy, the latter evaluated in CPP girls. Under hormonal stimulation from birth until puberty, various changes in the mammary gland occur.
Methods or Background: A population of 42 pediatric female patients with evidence of breast “bump” at an age below the threshold of eight years was selected. The girls clinically ascertained to have breast development (Tanner stage B2 and subsequent) underwent routine tests: measurement of basal levels of FSH, LH and E2, GnRH stimulation test, determination of bone age using hand-wrist X-ray, vitamin D dosage. The ultrasonographic stage was established by taking into consideration the dimensionally largest breast and assigned from one to five. Patients diagnosed with CPP were candidates for carrying out blocking therapy (Decapeptyl®); after six months of therapy they were recalled to evaluate again growth speed, bone age, and US breast development.
Results or Findings: It was observed that the average number of patients at the diagnosis of CPP was stage 3 at the US examination (versus Tanner stage B2), and that after six months of blocking therapy the average was stage 2 at US. Clinical evaluation according to Tanner stages tended to overestimate breast development.
Conclusion: The assessment of the ultrasonographic stage of breast development represents a novel method to help clinicians in the evaluation of little girls affected with PT e CPP, either in the diagnosis or follow-up, and might become routine very soon in the future.
Limitations: Limitated number of patients
Funding for this study: The study did not receive any funding.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: No information provided by the submitter.

Author Block: L. Duijm¹, R. O'Leary², L. Boersma², M. van der Sangen³, S. Siesling⁴, R-J. Schipper³, A. Voogd²; ¹Nijmegen/NL, ²Maastricht/NL, ³Eindhoven/NL, ⁴Utrecht/NL
Purpose: The aim of this study was to provide updated figures on the risk of invasive ipsilateral breast cancer (iIBC) after breast conserving surgery (BCS) of ductal carcinoma in situ (DCIS) with or without adjuvant radiotherapy (RT). A second aim was to analyse the association between DCIS grade and the risk of iIBC following BCS.
Methods or Background: In this population-based, retrospective cohort study, the Netherlands Cancer Registry collected patient information on 25,719 women who received a new diagnosis of DCIS in the Netherlands from 1989-2021 and who underwent BCS, of which 19,034 (74%) received adjuvant RT. Kaplan-Meier analyses and Cox multivariable regression models were used in the analyses.
Results or Findings: A total of 1,135 patients experienced an iIBC. The 10-year cumulative iIBC incidence rates for patients diagnosed in the periods 1989-1998, 1999-2008 and 2009-2021 and undergoing BCS only, were 12.6%, 9% and 5% (P<0.001), respectively. For those undergoing BCS with RT these figures were 5.7%, 3.7% and 2.2%, respectively (P<0.001). In the multivariable analyses, DCIS grade was not significantly associated with the risk of iIBC.
Conclusion: Since 1989 the risk of iIBC has decreased substantially. No significant association of DCIS grade with the risk iIBC was found, stressing the need for more powerful prognostic factors to guide the loco-regional treatment of DCIS.
Limitations: The well known poor reproducibility among pathologists in grading DCIS lesions will haveresulted in misclassification of tumour grade and is a potential source of information bias.
Funding for this study: No funding was received for this study.
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information: Informed consent was not necessary for this observational study. Patients are informed that their data is part of the Netherlands Cancer Registry, and there is an optional opt-out procedure available. No approval for this study was required from a Medical Ethics Committee.