BackgroundArtificial intelligence (AI)-based systems have the potential to increase the efficiency and effectiveness of breast cancer screening programs but need to be carefully validated before clinical implementation.PurposeTo retrospectively evaluate an AI system to safely reduce the workload of a double-reading breast cancer screening program.Material and MethodsAll digital mammography (DM) screening examinations of women aged 40-74 years between August 2021 and January 2022 in Östergötland, Sweden were included. Analysis of the interval cancers (ICs) was performed in 2024. Each examination was double-read by two breast radiologists and processed by the AI system, which assigned a score of 1-10 to each examination based on increasing likelihood of cancer. In a retrospective simulation, the AI system was used for triaging; low-risk examinations (score 1-7) were selected for single reading and high-risk examinations (score 8-10) for double reading.ResultsA total of 15,468 DMs were included. Using an AI triaging strategy, 10,473 (67.7%) examinations received scores of 1-7, resulting in a 34% workload reduction. Overall, 52/53 screen-detected cancers were assigned a score of 8-10 by the AI system. One cancer was missed by the AI system (score 4) but was detected by the radiologists. In total, 11 cases of IC were found in the 2024 analysis.ConclusionReplacing one reader in breast cancer screening with an AI system for low-risk cases could safely reduce workload by 34%. In total, 11 cases of IC were found in the 2024 analysis; of them, three were identified correctly by the AI system at the 2021-2022 examination.

Mammogram visual question answering (VQA) integrates image interpretation with clinical reasoning and has potential to support breast cancer screening. We systematically evaluated the GPT-5 family and GPT-4o model on four public mammography datasets (EMBED, InBreast, CMMD, CBIS-DDSM) for BI-RADS assessment, abnormality detection, and malignancy classification tasks. GPT-5 consistently was the best performing model but lagged behind both human experts and domain-specific fine-tuned models. On EMBED, GPT-5 achieved the highest scores among GPT variants in density (56.8%), distortion (52.5%), mass (64.5%), calcification (63.5%), and malignancy (52.8%) classification. On InBreast, it attained 36.9% BI-RADS accuracy, 45.9% abnormality detection, and 35.0% malignancy classification. On CMMD, GPT-5 reached 32.3% abnormality detection and 55.0% malignancy accuracy. On CBIS-DDSM, it achieved 69.3% BI-RADS accuracy, 66.0% abnormality detection, and 58.2% malignancy accuracy. Compared with human expert estimations, GPT-5 exhibited lower sensitivity (63.5%) and specificity (52.3%). While GPT-5 exhibits promising capabilities for screening tasks, its performance remains insufficient for high-stakes clinical imaging applications without targeted domain adaptation and optimization. However, the tremendous improvements in performance from GPT-4o to GPT-5 show a promising trend in the potential for general large language models (LLMs) to assist with mammography VQA tasks.

To construct a multimodal ultrasound (US) radiomics model for predicting axillary lymph node metastasis (ALNM) in breast cancer and evaluated its application value in predicting ALNM and patient prognosis. From March 2014 to December 2022, data from 682 breast cancer patients from four hospitals were collected, including preoperative grayscale US, color Doppler flow imaging (CDFI), contrast-enhanced ultrasound (CEUS) imaging data, and clinical information. Data from the First Medical Center of PLA General Hospital were used as the training and internal validation sets, while data from Peking University First Hospital, the Cancer Hospital of the Chinese Academy of Medical Sciences, and the Fourth Medical Center of PLA General Hospital were used as the external validation set. LASSO regression was employed to select radiomic features (RFs), while eight machine learning algorithms were utilized to construct radiomic models based on US, CDFI, and CEUS. The prediction efficiency of ALNM was assessed to identify the optimal model. In the meantime, Radscore was computed and integrated with immunoinflammatory markers to forecast Disease-Free Survival (DFS) in breast cancer patients. Follow-up methods included telephone outreach and in-person hospital visits. The analysis employed Cox regression to pinpoint prognostic factors, while clinical-imaging models were developed accordingly. The performance of the model was evaluated using the C-index, Receiver Operating Characteristic (ROC) curves, calibration curves, and Decision Curve Analysis (DCA). In the training cohort (n = 400), 40% of patients had ALNM, with a mean age of 55 ± 10 years. The US + CDFI + CEUS-based radiomics model achieved Area Under the Curves (AUCs) of 0.88, 0.81, and 0.77 for predicting N0 versus N+ (≥ 1) in the training, internal, and external validation sets, respectively, outperforming the US-only model (P < 0.05). For distinguishing N+ (1-2) from N+ (≥ 3), the model achieved AUCs of 0.89, 0.74, and 0.75. Combining radiomics scores with clinical immunoinflammatory markers (platelet count and neutrophil-to-lymphocyte ratio) yielded a clinical-radiomics model predicting disease-free survival (DFS), with C-indices of 0.80, 0.73, and 0.79 across the three cohorts. In the external validation cohort, the clinical-radiomics model achieved higher AUCs for predicting 2-, 3-, and 5-year DFS compared to the clinical model alone (2-year: 0.79 vs. 0.66; 3-year: 0.83 vs. 0.70; 5-year: 0.78 vs. 0.64; all P < 0.05). Calibration and decision curve analyses demonstrated good model agreement and clinical utility. The multimodal ultrasound radiomics model based on US, CDFI, and CEUS could effectively predict ALNM in breast cancer. Furthermore, the combined application of radiomics and immune inflammation markers might predict the DFS of breast cancer patients to some extent.

<i>"Just Accepted" papers have undergone full peer review and have been accepted for publication in <i>Radiology: Artificial Intelligence</i>. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content</i>. The MammosighTR dataset, derived from Türkiye's national breast cancer screening mammography program, provides BI-RADS-labeled mammograms with detailed annotations on breast composition and lesion quadrant location, which may be useful for developing and testing AI models in breast cancer detection. ©RSNA, 2025.

Tumor-infiltrating lymphocytes (TILs) are key immune biomarkers associated with prognosis and treatment response in early-stage breast cancer (BC), particularly in the triple-negative subtype. This study aimed to evaluate whether [18F]FDG PET/CT imaging and routine cell blood count (CBC)-derived biomarkers can serve as non-invasive surrogates for TILs, using machine-learning models. We retrospectively analyzed 358 patients with biopsy-proven early-stage invasive BC who underwent pre-treatment [18F]FDG PET/CT imaging. PET-derived biomarkers were extracted from the primary tumor, lymph nodes, and lymphoid organs (spleen and bone marrow). CBC-derived biomarkers included neutrophil-to-lymphocyte ratio (NLR) and platelet-to-lymphocyte ratio (PLR). TILs were assessed histologically and categorized as low (0-10%), intermediate (11-59%), or high (≥ 60%). Correlations were assessed using Spearman's rank coefficient, and classification and regression models were built using several machine-learning algorithms. Tumor SUVmax and tumor SUVmean showed the highest correlation with TIL levels (ρ = 0.29 and 0.30 respectively, p < 0.001 for both), but overall associations between TILs and PET or CBC-derived biomarkers were weak. No CBC-derived biomarker showed significant correlation or discriminative performance. Machine-learning models failed to predict TIL levels with satisfactory accuracy (maximum balanced accuracy = 0.66). Lymphoid organ metrics (SLR, BLR) and CBC-derived parameters did not significantly enhance predictive value. In this study, neither [18F]FDG PET/CT nor routine CBC-derived biomarkers reliably predict TILs levels in early-stage BC. This observation was made in presence of potential scanner-related variability and for a restricted set of usual PET metrics. Future models should incorporate more targeted imaging approaches, such as immunoPET, to non-invasively assess immune infiltration with higher specificity and improve personalized treatment strategies.

This study aimed to explore the diagnostic performance of ultrasound S-Detect in differentiating Breast Imaging-Reporting and Data System (BI-RADS) 4 breast nodules ≤ 20 mm and > 20 mm. Between November 2020 and November 2022, a total of 382 breast nodules in 312 patients were classified as BI-RADS-4 by conventional ultrasound. Using pathology results as the gold standard, we applied receiver operator characteristics (ROC), sensitivity (SE), specificity (SP), accuracy (ACC), positive predictive value (PPV), and negative predictive value (NPV) to analyze the diagnostic value of BI-RADS, S-Detect, and the two techniques in combination (Co-Detect) in the diagnosis of BI-RADS 4 breast nodules ≤ 20 mm and > 20 mm. There were 382 BI-RADS-4 nodules, of which 151 were pathologically confirmed as malignant, and 231 as benign. In lesions ≤ 20 mm, the SE, SP, ACC, PPV, NPV, and area under the curve (AUC) of the BI-RADS group were 77.27%, 89.73%, 85.71%, 78.16%, 89.24%, 0.835, respectively. SE, SP, ACC, PPV, NPV, and AUC of the S-Detect group were 92.05%, 78.92%, 83.15%, 67.50%, 95.43%, 0.855, respectively. SE, SP, ACC, PPV, NPV, and AUC of the Co-Detect group were 89.77%, 93.51%, 92.31%, 86.81%, 95.05%, 0.916, respectively. The differences of SE, ACC, NPV, and AUC between the BI-RADS group and the Co-Detect group were statistically significant (P < 0.05). In lesions > 20 mm, SE, SP, ACC, PPV, NPV, and AUC of the BI-RADS group were 88.99%, 89.13%, 88.99%, 91.80%, 85.42%, 0.890, respectively. SE, SP, ACC, PPV, NPV, and AUC of the S-Detect group were 98.41%, 69.57%, 86.24%, 81.58%, 96.97%, 0.840, respectively. SE, SP, ACC, PPV, NPV, and AUC of the Co-Detect group were 98.41%, 91.30%, 95.41%, 93.94%, 97.67%, 0.949, respectively. A total of 166 BI-RADS 4 A nodules were downgraded to category 3 by Co-Detect, with 160 (96.4%) confirmed as benign and 6 (all ≤ 20 mm) as false negatives. Conversely, 25 nodules were upgraded to 4B, of which 19 (76.0%) were malignant. The difference in AUC between the BI-RADS group and the Co-Detect group was statistically significant (P < 0.05). S-Detect combined with BI-RADS is effective in the differential diagnosis of BI-RADS 4 breast nodules ≤ 20 mm and > 20 mm. However, its performance is particularly pronounced in lesions ≤ 20 mm, where it contributes to a significant reduction in unnecessary biopsies.

Breast cancer, the most prevalent cancer among women globally, is classified into molecular subtypes (luminal, HER2-enriched, and triple-negative) to guide treatment and prognosis. Traditional subtyping methods, such as gene profiling and immunohistochemistry, are invasive and limited by intratumoural heterogeneity. MRI radiomics analysis offers a non-invasive alternative by extracting quantitative imaging features, yet its application in diverse, multi-ethnic populations remains underexplored. This study aimed to identify predictive radiomic features from multiple MRI sequences to classify breast cancer subtypes, compare the performance of four MRI sequences, and determine the optimal machine learning (ML) model for this task. A total of 162 retrospective breast cancer MRI cases were semi-automatically segmented, and 256 radiomic features were extracted. A multimodal ML framework integrating random forest and recursive feature elimination was developed to identify the most predictive features based on the area under the receiver operating characteristic curve (AUROC). Key predictive features included age, tumour size, margin characteristics, and intensity patterns within the tumour. Among MRI sequences, inversion recovery and T1 post-contrast performed best for subtyping. In addition, texture-based ML models effectively emulated visual assessment, demonstrating the potential of radiomics in non-invasive breast cancer subtyping. With the top ten features, the AUROC values are 0.735, 0.630, and 0.747 for luminal, HER2-enriched, and triple-negative, respectively. These findings highlight the role of MRI-based texture features and advanced ML in enhancing breast cancer diagnosis, offering a non-invasive tool for personalised treatment planning while complementing existing clinical workflows.

BackgroundBreast microcalcification diagnostics are challenging due to their subtle presentation, overlapping with benign findings, and high inter-reader variability, often leading to unnecessary biopsies. While deep learning (DL) models - particularly deep convolutional neural networks (DCNNs) - have shown potential to improve diagnostic accuracy, their clinical application remains limited by the need for large annotated datasets and the "black box" nature of their decision-making. PurposeTo develop and validate a deep learning model (DCNN) using a double transfer learning (d-TL) strategy for classifying suspected mammographic microcalcifications, with explainable AI (XAI) techniques to support model interpretability. Material and methodsA retrospective dataset of 396 annotated regions of interest (ROIs) from full-field digital mammography (FFDM) images of 194 patients who underwent stereotactic vacuum-assisted biopsy at the Womens Hospital radiological department, Helsinki University Hospital, was collected. The dataset was randomly split into training and test sets (24% test set, balanced for benign and malignant cases). A ResNeXt-based DCNN was developed using a d-TL approach: first pretrained on ImageNet, then adapted using an intermediate mammography dataset before fine-tuning on the target microcalcification data. Saliency maps were generated using Gradient-weighted Class Activation Mapping (Grad-CAM) to evaluate the visual relevance of model predictions. Diagnostic performance was compared to a radiologists BI-RADS-based assessment, using final histopathology as the reference standard. ResultsThe ensemble DCNN achieved an area under the ROC curve (AUC) of 0.76, with 65% sensitivity, 83% specificity, 79% positive predictive value (PPV), and 70% accuracy. The radiologist achieved an AUC of 0.65 with 100% sensitivity but lower specificity (30%) and PPV (59%). Grad-CAM visualizations showed consistent activation of the correct ROIs, even in misclassified cases where confidence scores fell below the threshold. ConclusionThe DCNN model utilizing d-TL achieved performance comparable to radiologists, with higher specificity and PPV than BI-RADS. The approach addresses data limitation issues and may help reduce additional imaging and unnecessary biopsies.

Enhancing the accuracy of tumor response predictions enables the development of tailored therapeutic strategies for patients with breast cancer. In this study, we developed deep radiomic models to enhance the prediction of chemotherapy response after the first treatment cycle. 18F-Fludeoxyglucose PET/CT imaging data and clinical record from 60 breast cancer patients were retrospectively obtained from the Cancer Imaging Archive. PET/CT scans were conducted at three distinct stages of treatment; prior to the initiation of chemotherapy (T1), following the first cycle of chemotherapy (T2), and after the full chemotherapy regimen (T3). The patient's primary gross tumor volume (GTV) was delineated on PET images using a 40% threshold of the maximum standardized uptake value (SUVmax). Radiomic features were extracted from the GTV based on the PET/CT images. In addition, a squeeze-and-excitation network (SENet) deep learning model was employed to generate additional features from the PET/CT images for combined analysis. A XGBoost machine learning model was developed and compared with the conventional machine learning algorithm [random forest (RF), logistic regression (LR) and support vector machine (SVM)]. The performance of each model was assessed using receiver operating characteristics area under the curve (ROC AUC) analysis, and prediction accuracy in a validation cohort. Model performance was evaluated through fivefold cross-validation on the entire cohort, with data splits stratified by treatment response categories to ensure balanced representation. The AUC values for the machine learning models using only radiomic features were 0.85(XGBoost), 0.76 (RF), 0.80 (LR), and 0.59 (SVM), with XGBoost showing the best performance. After incorporating additional deep learning-derived features from SENet, the AUC values increased to 0.92, 0.88, 0.90, and 0.61, respectively, demonstrating significant improvements in predictive accuracy. Predictions were based on pre-treatment (T1) and post-first-cycle (T2) imaging data, enabling early assessment of chemotherapy response after the initial treatment cycle. Integrating deep learning-derived features significantly enhanced the performance of predictive models for chemotherapy response in breast cancer patients. This study demonstrated the superior predictive capability of the XGBoost model, emphasizing its potential to optimize personalized therapeutic strategies by accurately identifying patients unlikely to respond to chemotherapy after the first treatment cycle.

Complete tumour removal is vital in curative breast cancer (BCa) surgery to prevent recurrence. Recently, [¹⁸F]FDG micro-PET-CT of lumpectomy specimens has shown promise for intraoperative margin assessment (IMA). To aid interpretation, we trained a 2D Residual U-Net to delineate invasive carcinoma of no special type in micro-PET-CT lumpectomy images. We collected 53 BCa lamella images from 19 patients with true histopathology-defined tumour segmentations. Group five-fold cross-validation yielded a dice similarity coefficient of 0.71 ± 0.20 for segmentation. Afterwards, an ensemble model was generated to segment tumours and predict margin status. Comparing predicted and true histopathological margin status in a separate set of 31 micro-PET-CT lumpectomy images of 31 patients achieved an F1 score of 84%, closely matching the mean performance of seven physicians who manually interpreted the same images. This model represents an important step towards a decision-support system that enhances micro-PET-CT-based IMA in BCa, facilitating its clinical adoption.