RPS 1302 - The state of the art in breast MRI

Author Block: E. Rubino¹, C. M. L. Trombadori¹, C. Boldrini¹, A. Marra¹, A. D'Angelo¹, P. Belli¹, P. Clauser²; ¹Rome/IT, ²Vienna/AT
Purpose: In breast magnetic resonance imaging (MRI), T2-weighted (T2w) hyperintensity is typically observed in benign lesions but can also occur in some malignant tumors. The aim of our study was to retrospectively evaluate the association between T2 hyperintensity and histopathological and molecular features of breast lesions.
Methods or Background: We retrospectively analyzed 1387 breast MRIs performed between 2022 and 2024. For each lesion, a breast radiologist assessed signal intensity on T2w images using non-fat suppressed T2-TSE sequences, enhancement pattern, histology, molecular subtype (Luminal A, Luminal B, HER2+, triple-negative), and Ki67 proliferation index. Associations among these variables were analyzed using Pearson’s chi-squared test, Fisher’s exact test, Mann-Whitney U test, and Spearman’s correlation.
Results or Findings: Out of 1387 patients, 779 lesions were hyperintense on T2w images. No statistically significant association was found between T2 hyperintensity and lesion benignity or malignancy, with similar proportions of hyperintensity in both groups. Analysis of malignant histological subtypes revealed that tumors with an in situ component (DCIS and DCIS + IDC) were more frequent among non-hyperintense lesions, while rare tumor subtypes - such as mucinous, papillary, and metaplastic carcinomas - were significantly more frequent among T2-hyperintense lesions (p<0.001). Molecular subtype analysis showed that Luminal A tumors were underrepresented among T2-hyperintense lesions, whereas Luminal B and triple-negative tumors were more common (p<0.05). Spearman’s correlation indicated that higher T2 signal intensity was associated with higher Ki67 levels (p<0.001).
Conclusion: In our cohort, T2 hyperintensity showed significant associations with rare tumor subtypes (mucinous, papillary, and metaplastic carcinomas), Luminal B and triple-negative molecular subtypes, and high proliferative activity. These preliminary findings suggest that T2w signal characteristics may contribute to refining lesion interpretation in breast MRI and highlight their potential role in lesion characterization.
Limitations: Retrospective single-study center, single reader
Funding for this study: No
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethics Committee Territorial Lazio Area 3, protocol no. 00000883/24, ID 6699, non-profit study.

Author Block: A. Angelino¹, F. Ferrara¹, V. Longo¹, E. Farina¹, A. Carcagnì¹, C. M. L. Trombadori¹, P. Belli¹, A. D'Angelo¹, P. Clauser²; ¹Rome/IT, ²Vienna/AT
Purpose: To explore the role of high b-value 1400 DWI in breast lesion characterization comparing a quantitative assessment of the apparent diffusion coefficient (ADC) map with a qualitative visual analysis.
Methods or Background: This retrospective, IRB-approved, single-center study included women with at least one breast lesion classified as BI-RADS 2–6 at consecutive MRIs performed from January 2022 to February 2024. The reference standard was histology from image-guided biopsy or ≥2 years follow-up for non-suspicious lesions. MRI was performed on 1.5T scanners according to international guidelines, including DWI with b=0,1400 s/mm². ADC maps were generated using mono-exponential fitting. One of six breast radiologists (≥3 years’ experience), blinded to clinical data, evaluated lesions quantitatively (ADC measured with a ROI in the darkest area) and qualitatively (5-point visual scale: hyperintense/non-suspicious to markedly hypointense/highly suspicious). ROC analysis determined sensitivity and specificity; the highest Youden index on the ROC curve defined the optimal ADC threshold.
Results or Findings: A total of 773 lesions in 760 women (mean age 51.2 years) were analyzed: 597 malignant (77.2%) and 176 benign (22.8%). Malignant lesions had lower ADC values (0.85×10⁻³ mm²/s, SD 0.21) than benign lesions (1.23×10⁻³ mm²/s, SD 0.34; p<0.001). Quantitative analysis achieved an AUC of 0.81 (95% CI: 0.78–0.87). ADC threshold for distinguishing malignant from benign lesions was of 1.03×10⁻³ mm²/s (85% sensitivity, 72% specificity). Visual assessment showed an AUC of 0.73 (95% CI: 0.68–0.78), with high sensitivity (89%) but low specificity (49%).
Conclusion: Quantitative and qualitative ADC map assessment effectively differentiated benign from malignant breast lesions but remained less accurate than CE-MRI (p=0.001 and p=0.027). Visual ADC analysis achieved good sensitivity yet low specificity.
Limitations: Retrospective, single-center study; high cancer prevalence, inter-reader variability, subjective scoring.
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: This study was approved by the local ethics committee (ID 6699).

Author Block: E. N. Korkmaz, H. Aydin, M. Kamışlık; Ankara/TR
Purpose: In breast MRI of non-mass enhancement (NME), BI-RADS internal enhancement patterns—clumped and clustered ring (CRE)—have not been reported to transform across dynamic phases. We observed that lesions appearing clumped on early dynamics often demonstrated CRE on later or high-resolution images, suggesting that limited spatial resolution may obscure early CRE. This study evaluated the malignancy association of this phase-dependent transformation and its added diagnostic value combined with ADC.
Methods or Background: This single-centre retrospective study (2022–2025) included 80 patients with histopathology-proven NME. MRI was performed at 1.5 T with a dedicated breast coil. Early and late post-contrast DCE-MRI, a high-resolution sagittal sequence, and diffusion-weighted imaging were obtained. Enhancement patterns were assessed by two blinded radiologists. Statistical analysis included χ²/Fisher’s tests, Mann–Whitney U for age, ROC analysis, and logistic regression.
Results or Findings: Among 80 lesions, 47 were benign and 33 malignant. Malignant cases were significantly older (p = 0.015). Late-phase and high-resolution images showed a stronger association of CRE with malignancy. Phase-dependent transformation occurred in 43.8% of lesions and was associated with increased malignancy risk. Washout kinetic curves were more frequent in malignant lesions. ADC showed moderate diagnostic performance, and logistic regression identified CRE and wider distribution as independent predictors.
Conclusion: Phase-dependent transformation from clumped to CRE is a frequent and diagnostically valuable feature of NME. Incorporating late-phase, high-resolution imaging and diffusion assessment into breast MRI protocols may improve specificity and guide biopsy decisions.
Limitations: This study has several limitations. Its retrospective, single-centre design may limit generalisability. Not all benign lesions had histological confirmation or complete follow-up, possibly underestimating malignancy. Interobserver variability remains a factor despite consensus reading, and ultrafast MRI was not directly evaluated.
Funding for this study: This study received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the Institutional Review Board of Gülhane Training and Research Hospital, approval number: 2025/126

Author Block: J. Zhou, M. Du, M. Wang; Wenzhou/CN
Purpose: Tumor-infiltration lymphocytes (TILs) is a key prognostic factor for breast cancer. It is a crucial component of the tumor microenvironment associated with the metabolism of tumor cells and the local immune response. The study aimed to evaluate the clinicopathological parameters associated with TILs, and MRI features corresponding to TILs expression.
Methods or Background: A total of 785 patients were retrospectively investigated in this study, with 676 as training dataset and 109 as testing dataset. The percentage of TILs in the stroma adjacent to the tumor cells was assessed on H&E-stained slides, and they were separated into low (<10%) and high (≥10%) groups. MRI features were reviewed by two radiologists by consensus using BI-RADS lexicon descriptors. Clinicopathological and MRI features were compared between low and high TILs groups, and parameters with P<0.05 were used to build logistic regression model. The analysis of disease-free survival (DFS) was performed to assess the model’s performance.
Results or Findings: Of 676 patients, 43 (6.4%) underwent recurrence with the DFS 54.1±17.4 (mean±SD) months, compared to non-recurrence group with DFS 29.0±20.4 months. Histological grade, ER, PR, HER2 over-expression, low-HER2, Ki-67≥20%, molecular subtype and the presence of peri-tumor edema showed significant difference between low and high TILs groups. Nomogram was build with above parameters and the AUC was 0.727 in training and 0.710 in testing dataset. Kaplan-Meier survival analysis showed patients with high TILs level had better DFS than who with low TILs, but without showing significant difference (P=0.052, HR=1.80).
Conclusion: Certain important pathological parameters and edema presence on MRI are associated with high TILs expression, which may help identify patients who may be candidate for immunotherapy and have better response to neoadjuvant therapy and better survival.
Limitations: The cutoff of TILs level hasn’t been standardized.
Funding for this study: This study was supported by National Natural Science Foundation of China (No. 82572172).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study has been approved by the ethics committee in the hospital, and approved document is YS2025-423.

Author Block: S. Kılıçarslan Özdemir, G. Sezgin, M. F. Tumer, M. Yağtu, K. A. Sinci, B. GENÇER; Izmir/TR
Purpose: Although many studies have reported an association between breast density and breast cancer, this relationship remains debated. Dense parenchyma is typically defined by visual mammographic assessment. This study aimed to quantitatively evaluate the relationship between breast density and breast cancer using three-dimensional volumetric MR images.
Methods or Background: This retrospective study reviewed MR exams from January to December 2022. Of 411 examinations, cases with bilateral malignancy (n=32), breast implants (n=59), prior contralateral mammoplasty (n=74), asymmetric parenchyma (n=3), or pure DCIS (n=2) were excluded. The final cohort included 121 patients with unilateral breast cancer and 121 controls without malignancy; none in the control Group developed cancer during two-year follow-up. MR scans were obtained using a 3T Lumina system with an 18-channel breast coil.Breast density was measured semi-automatically on non–fat-suppressed 3D T1-weighted images. The thoracic region was cropped along the pectoral muscle, and total breast and fat volumes segmented using a bright fat thresholding technique. Parenchymal volume was calculated by subtraction. Two observers performed the measurements. Based on BI-RADS 4th edition, densities ≥50% were classified as “dense.”
Results or Findings: Among malignant cases, 83 were invasive ductal, 28 invasive lobular, and 10 other types. The mean age was 52.8 years in the malignant group and 45.2 in controls. Dense parenchyma was found in 15.7% of malignant and 3.3% of controls. Breast density was significantly higher in the malignant group (p=0.000). Logistic regression indicated malignancy risk increased with fat content and age.
Conclusion: Quantitative MR analysis confirmed a strong association between increased breast density and breast cancer. Three-dimensional MR provides reliable volumetric assessment, though semi-automatic segmentation is time-consuming.
Limitations: Single-center design, limited sample size, and interobserver variability. Future multicenter studies using automated deep learning and radiomics methods may improve reproducibility.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: January 2024 / ID:647

Author Block: E. Tallamona¹, P. Francesco²; ¹Gela/IT, ²Misterbianco/IT
Purpose: To evaluate the role of the Kaiser Score (KS) applied to pre-treatment breast MRI in reducing unnecessary second-look ultrasound examinations and biopsies in patients scheduled for neoadjuvant chemotherapy (NACT).
Methods or Background: This retrospective single-center study included consecutive patients with biopsy-proven breast cancer evaluated for NACT between January 2023 and December 2024. All patients underwent baseline breast MRI. For each MRI-detected lesion, the KS was calculated based on morphology, dynamic enhancement, and additional diagnostic features. Lesions with KS ≥5 were considered suspicious, while those with KS ≤4 were considered likely benign. Second-look ultrasound was performed for lesions without sonographic correlates or not already biopsied, and biopsies were conducted according to institutional protocols. Final histopathological results were used as the reference standard. The primary endpoints were the number of second-look ultrasounds and biopsies that could be avoided by applying the KS. Diagnostic performance metrics, including sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV), were calculated.
Results or Findings: We included in the study 56 patients. Only the "second-look" lesions evaluated as suspicious and biopsied by a team of at least 5 year experienced breast radiologists were included in the study. Histological results were collected and classified according the Easton Ellis classification from B1 to B5 and then compared to the given imaging score. By adopting a KS threshold of 5 both second-look ultrasounds and biopsies could have been safely avoided with a low grade of missing "B3" and malignant lesion.
Conclusion: The KS is a reliable tool to stratify MRI-detected lesions in patients undergoing NACT. Its systematic use before second-look ultrasound could substantially reduce unnecessary imaging and biopsies, streamlining pre-treatment workup and minimizing patient burden.
Limitations: This study is limited by its retrospective, single-center design,
Funding for this study: No funding needed
Has your study been approved by an ethics committee? Not applicable
Ethics committee - additional information:

Author Block: J. Qu, Y. Liu, S. Lui; Cheng Du/CN
Purpose: To develop and evaluate an interpretable model for distinguishing early breast cancer (BC) from benign lesions and generating structured reports.
Methods or Background: Preoperative MRI scans and corresponding radiological reports from patients with breast lesions were retrospectively collected across five institutions. Lesion descriptors extracted from MRI reports were translated into standardized BI-RADS lexicon terms (hereafter termed “concepts”) by radiologists. These concepts, alongside multiparametric MRI sequences, were input into a Concept Bottleneck Model (CBM) designed to differentiate BC from benign lesions and to generate structured reports. Classification performance of the CBM was compared against a black-box model. CBM-derived concept accuracy was also assessed. Additionally, a two-phase multi-reader study was conducted to evaluate the clinical utility of the CBM.
Results or Findings: A total of 1,695 pathology-confirmed breast lesions (857 malignant and 838 benign) from 1,634 patients were included. In the test set, the CBM achieved an AUC of 0.92 (95%CI 0.90-0.93), comparable to the benchmark black-box model (AUC: 0.93 [95%CI 0.92-0.94]). The accuracy of CBM-generated concepts ranged from 0.62 to 1.00. In the multi-reader study, the CBM yielded an accuracy of 0.89, performing on par with one radiologist and outperforming the other seven (all P < 0.05). With CBM assistance, radiologists correctly downgraded 22.1% of the lesions to benign. Diagnostic accuracy improved for three radiologists, increasing from 0.71-0.72 to 0.82-0.91 (all P < 0.05). Furthermore, CBM support enhanced inter-reader agreement in both concept recognition and BI-RADS classification, with Gwet’s AC1 increasing from 0.27-1.00 to 0.55-1.00.
Conclusion: The CBM enhances intrinsic interpretability through image-concept alignment, effectively discriminates early BC from benign lesions, potentially reduces unnecessary benign breast lesion biopsies, and improves reporting consistency among radiologists.
Limitations: The limited size of the dataset warrants further validation in larger, preferably prospective cohorts.
Funding for this study: This study was supported by (1) the National Natural Science Foundation of China (Project Nos. 82441007, 82120108014); (2) 1.3.5 project for disciplines of excellence, West China Hospital, Sichuan University (Project No. ZYGD23003); (3) 1.3.5 projects for Artificial Intelligence (Project No. ZYAI24010), West China Hospital, Sichuan University; (4) Chengdu Science and Technology Office, major technology application demonstration project (Project No.2022-GH03-00017-HZ); and (5) the Postdoctor Research Fund of West China Hospital, Sichuan University (2025HXBH040).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This retrospective study was approved by the Ethics Committee on Biomedical Research, West China Hospital of Sichuan University, Chengdu, Chinal (approval No. 2025-949).

Author Block: S. Ham¹, B. K. Seo¹, S. Wang², J. Y. Lee¹, M. S. Bae¹, K. R. Cho³, O. Woo³, S. E. Song³, S-Y. Kim³; ¹Ansan/KR, ²Cambridge/UK, ³Seoul/KR
Purpose: Ductal carcinoma in situ (DCIS) often presents on MRI as nonmass enhancement with variable enhancement than invasive cancer. We aimed to develop and validate a deep learning–based automatic segmentation algorithm for biopsy-proven DCIS on preoperative MRI, achieving robust performance across lesion types, background parenchymal enhancement (BPE), and invasive cancer upgrade.
Methods or Background: This retrospective multicenter study included 842 DCIS lesions (Institution A, 342; Institution B, 500) collected over 11 years from nine 3T scanners (five vendors). SAM 2.1-L was initially trained on 292 cases from Institution A, with 50 held out for testing. Active learning was applied to Institution B in three iterations (250, 150, and 50 cases), using expert corrections for retraining. Final evaluation was performed on an independent test set of 100 lesions (50 per institution). Subgroup analyses included lesion type, size, BPE, and invasive upgrade. Segmentation performance was assessed with the dice similarity coefficient (DSC) and compared with training from scratch. Annotation times were compared between manual and active learning–assisted labeling.
Results or Findings: Overall DSC improved from 0.68 in the initial model to 0.73 after multiple iterations (p<0.05). DSC did not differ by lesion type (mass, 0.72 vs. nonmass enhancement, 0.73) or invasive upgrade (pure DCIS, 0.77 vs. invasive upgrade, 0.71; both p>0.05), but was significantly higher in tumors >30 mm and with lower BPE (minimal or mild) (all p<0.05). Annotation time decreased from 453 to 35 seconds per lesion (p<0.05).
Conclusion: Active learning with SAM 2.1-L enabled robust DCIS segmentation on MRI regardless of lesion type or invasive upgrade, while dramatically reducing annotation time. Gains in accuracy and efficiency support its potential for clinical integration in breast MRI analysis.
Limitations: External multicenter datasets is needed to confirm generalizability.
Funding for this study: This research was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (RS-2024-00347290).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Approved by the Institutional Review Board (IRB) of Korea University Ansan Hospital.

Author Block: T. Spiegel¹, V. Romeo², D. A. Resch¹, O. Lafcı¹, S. Rasul¹, A. Stiglbauer-Tscholakoff¹, P. Clauser¹, K. Pinker-Domenig³, T. H. Helbich¹; ¹Vienna/AT, ²Naples/IT, ³New York, NY/US
Purpose: To evaluate whether multiparametric 18F-Fluorodeoxyglucose (18F-FDG)-PET/MR-imaging biomarkers provide prognostic value for predicting overall survival (OS) in breast cancer (BC) patients.
Methods or Background: This retrospective analysis of a prospectively performed single-center study includes 202 BC patients (median age 52 years, range 25-82 years), who underwent simultaneous 18F-FDG-PET/MRI of the breast and whole body using a Biograph mMR system (Siemens, Germany) for staging purposes. Follow-up was available for up to 103 months (median 69 months). Imaging parameters of the index lesion included tumor size, SUVmax, ADC and perfusion parameters (Plasma Flow, Volume Distribution, Mean Transit Time). Histopathological and clinical variables included lymph node (LN) status, presence of distant metastases, and molecular subtype. Kaplan-Meier survival curves with log-rank tests, univariate and multivariate Cox regression analyses using hierarchical modeling to assess incremental prognostic value, were performed; p<0.05 was considered significant.
Results or Findings: During follow-up, 28 (13.9%) patients passed away. Tumor size predicted poorer OS (HR 1.224 (95% CI 1.044-1.435), p=0.013). Presence of distant metastases was strongly predictive for OS (HR 3.609 (95%CI 1.632-7.982), p=0.002). OS differed significantly among molecular subtypes (log rank p<0.001). LN status showed only a trend (HR 2.330 (95%CI 0.990-5.482), p=0.053). SUVmax, ADC, and perfusion parameters were not significant in predicting OS. Adding distant metastasis to tumor size improved the model (p=0.013). Further inclusion of LN information (p=0.191), molecular subtype (p=0.084) and imaging parameters (p>0.225) contributed minimally.
Conclusion: Among several histopathologic and clinical variables tumor size and distant metastases remain the strongest predictors for OS in BC patients. Several 18F-FDG-PET/MR-imaging biomarkers did not improve prognostic models significantly. This indicates that FDG-PET/MR-imaging parameters do not yet surpass conventional assessments of tumor size and metastases in predicting OS.
Limitations: Single-center design, limited number of events, and heterogeneous follow-up duration.
Funding for this study: This study is supported by the WWTF (Vienna Science and Technology Fund), grant number LS20-065.
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: EK Nr. 510/2009

Author Block: G. Chiffi¹, Ö. Lafci², A. Santonocito³, N. Pirringer-Pötsch², A. Stiglbauer-Tscholakoff², M. Costantini¹, T. H. Helbich², P. A. Baltzer², P. Clauser²; ¹Rome/IT, ²Vienna/AT, ³Turin/IT
Purpose: To evaluate the accuracy of tumor size and T-stage estimation in IBC by Contrast-Enhanced Mammography (CEM) and Magnetic Resonance Imaging (MRI), compared with pathology.
Methods or Background: This retrospective single-centre study included 100 patients (mean age 57 years) with biopsy-proven IBC who underwent pre-surgical CEM and MRI, each showing at least a component of mass enhancement. Two radiologists independently measured the maximum invasive mass component. CEM lesions were measured on low-energy and recombined images, MRI measurements on early post-contrast dynamic series (MRI-dyn) and Maximum Intensity Projection reconstructions (MRI-MIP). Pathology was the reference standard. Agreement was evaluated with intraclass correlation coefficients (ICC), Bland–Altman analysis using a ±2.5 mm margin. Impact on T-stage classification was assessed.
Results or Findings: A total of 112 invasive lesions were analysed (mean pathological size 13.8 mm). All modalities showed good correlation with pathology (ICC 0.825–0.876, p < 0.001). Mean lesion sizes were close to pathology, but case-wise agreement varied. Bland–Altman analysis showed for CEM mean differences ranged from –0.9 to +0.2 mm across reader, with limits of agreement (LOA) –11.4 to +9.6 mm; for MRI-dyn, –1.1 to 0.0 mm, LOA –11.5 to +9.3 mm. MRI-MIP showed differences –1.6 to +0.5 mm with wider LOA (–12.2 to +11.2 mm). Discrepancies increased with tumor size, trending toward underestimation. T-stage accuracy was highest with MRI-dyn (84%), followed by CEM (80%), while MRI-MIP consistently overstaged. Small but consistent inter-reader shifts were observed.
Conclusion: CEM and MRI-dyn showed good correlation with pathology, but accuracy declined for larger tumors. For preoperative size and T-stage assessment, MRI-dyn should be preferred, CEM as an alternative. Standardized measurement rules may reduce inter-reader variability.
Limitations: Single-centre retrospective design, modest sample size, exclusion of non-mass enhancement, which-though likely irrelevant for T stage-may influence surgical planning.
Funding for this study: Not applicable
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This study was approved by the Institutional Review Board (IRB), which ruled that written informed consent is unnecessary. Data were collected within a prospective study comparing the diagnostic value of CEM to MRI in a problem-solving setting (Ethics Review Board number 2282/2019).

Author Block: J. Azcona Sáenz¹, J. Molero-Calafell¹, U. Lalji², T. Wegman², P. Santiago Díaz¹, J. M. Maiques Llacer¹, R. Alcantara Souza¹, M. d. M. Vernet Tomás¹, T. Van Nijnatten²; ¹Barcelona/ES, ²Maastricht/NL
Purpose: To compare contrast-enhanced mammography (CEM) and MRI in the preoperative diagnosis of pure ductal carcinoma in situ (pDCIS), and to evaluate their performance in detecting additional lesions (ALs).
Methods or Background: In this international two-center retrospective study, 46 pDCIS lesions at Hospital del Mar and 18 lesions at Maastricht UMC+, all of whom underwent preoperative CEM and MRI, were reviewed. The sensitivity, morphology, and extent of pDCIS on each modality, as well as their ability to detect ALs, were assessed.
Results or Findings: CEM and MRI combined with low-energy images (LEI) showed the highest sensitivity for pDCIS detection at both centers (100.0%), followed by MRI (97.5% at Hospital del Mar; 93.3% at Maastricht UMC+), recombined images (RCI) (90.0% and 86.7%), and LEI (80.0% and 86.7%) of CEM. Calcifications were the most frequent LEI descriptor (~60%). On RCI and MRI, pDCIS predominantly presented as non-mass enhancement (60.0–75.0). LEI tended to underestimate tumor size, particularly at Maastricht UMC+, whereas MRI provided the most accurate estimations, closely followed by RCI. MRI detected more ALs than CEM; however, most corresponded to false positives (Hospital del Mar: 21 vs. 12 [MRI vs. CEM]; Maastricht UMC+: 15 vs. 11). CEM failed to identify one additional malignant lesion.
Conclusion: CEM and MRI + LEI showed slightly higher sensitivity than MRI for detecting pDCIS. MRI provided the most accurate preoperative size assessment, closely followed by RCI, with both modalities overestimating size. MRI was superior to CEM in detecting additional malignant lesions.
Limitations: -Small sample size in both cohorts.
-Retrospective design.
-Lesion size assessment may have been biased in cases where VAB were performed.
-Diagnostic specificity and disease prevalence could not be evaluated.
-Not all patients underwent bilateral mastectomies. Therefore, assessment of true false-negative rates was precluded.
Funding for this study: No
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: This retrospective study was approved by the Institutional Ethics Committees of both centers, with a waiver of informed consent.
Maastricht UMC+: 2025-0114
Hospital del Mar: 2025/11910