AI and Cancer Research ── Breast ── 2026-10-25
Breast cancer: treatment, trials and AI
- Abstract
Tumor-infiltrating lymphocytes (TILs) are biomarkers of the breast cancer immune microenvironment associated with treatment response and prognosis, but their assessment relies mainly on surgical histopathology. We aimed to develop and validate a preoperative model combining multiphase tumoral–peritumoral DCE-MRI radiomics with conventional features to predict TIL levels. This retrospective two-center study included 495 patients who underwent preoperative breast DCE-MRI followed by surgical resection as initial treatment. TILs were assessed on surgical specimens using a prespecified 10% cutoff commonly adopted in previous breast MRI-based TIL studies. Radiomic features from tumoral and 3-mm peritumoral regions across six DCE-MRI phases were combined with clinical, semantic MRI, and biopsy-derived features. Model performance was assessed by discrimination, calibration, incremental value, and clinical utility, and Shapley additive explanations (SHAP) analysis was used to assess feature contributions. The combined model achieved the AUC of 0.882 (95% confidence interval [CI]: 0.840–0.920), 0.777 (95% CI: 0.672–0.866), and 0.738 (95% CI: 0.637–0.826) in the training, internal validation, and external validation sets, respectively. In external validation, the combined model showed significant integrated discrimination improvement (IDI) over the conventional model (IDI = 0.105, P = 0.005) and the tumoral–peritumoral model (IDI = 0.098, P = 0.001). Decision curve analysis showed greater net benefit across most evaluated threshold probabilities. SHAP analysis identified later-phase peritumoral radiomic features as the dominant predictive factors for TIL estimation. Integrating multiphase tumoral–peritumoral DCE-MRI radiomics with conventional preoperative features showed potential for predicting TIL levels, with late-phase peritumoral features dominating the predictive performance. This model provides a promising preoperative tool for immune-related risk stratification and individualized multidisciplinary treatment planning.
Journal IF-equivalent: 3.8 (OpenAlex 2-year mean citedness, value as of 2026-10-02, retrieved 2026-10-03; not the official Clarivate IF)Reference: Yang W, Chen M, Cui Y, Xu W, Tian W, Jiang Y, et al. Preoperative prediction of breast cancer tumor-infiltrating lymphocytes levels with multiphase tumoral–peritumoral DCE-MRI radiomics and conventional features. BMC Med Imaging. 2026 Oct 23 [Epub ahead of print]. doi:10.1186/s12880-026-02930-y.Checked: Abstract only - Abstract
Background: Accurate identification of residual axillary lymph node (ALN) disease after neoadjuvant chemotherapy (NAC) is essential for treatment planning and for evaluating potential surgical de-escalation. We assessed the diagnostic performance of conventional ultrasonography (US), shear-wave elastography (SWE), artificial intelligence (AI)/radiomics, and multimodal ultrasound-based models for post-NAC axillary restaging.
Methods: PubMed, Embase, the Cochrane Library, and Web of Science were searched from inception through December 2025. Histopathology after axillary surgery was the reference standard, and residual axillary nodal disease (ypN+) was uniformly defined as the positive target condition. Study-level 2×2 tables were harmonized to this direction. Risk of bias and applicability were assessed with the Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) tool. Bivariate random-effects models and hierarchical summary receiver operating characteristic (HSROC) curves were used for modality groups with a sufficient number of independent studies; the positive likelihood ratio (PLR) and negative likelihood ratio (NLR) were pooled with random-effects models. Potentially overlapping cohorts were examined in sensitivity analyses. The study protocol was registered in PROSPERO (International Prospective Register of Systematic Reviews; registration No. CRD420251271199).
Results: A total of 14 articles were included in the systematic review, 10 of which contributed 12 test-level datasets to the quantitative synthesis. For conventional US (six studies), the pooled sensitivity was 0.55 [95% confidence interval (CI): 0.36–0.73], the pooled specificity was 0.79 (95% CI: 0.73–0.84), the area under the HSROC curve (AUC) was 0.78, and the pooled PLR and NLR were 2.47 (95% CI: 1.42–4.28) and 0.54 (95% CI: 0.30–0.95), respectively. AI/radiomics models (three studies) showed a pooled sensitivity of 0.93 (95% CI: 0.80–0.98), a specificity of 0.82 (95% CI: 0.60–0.94), and an AUC of 0.94; however, likelihood-ratio CIs were wide, and the bivariate covariance estimate reached the boundary. Evidence for SWE was limited to two potentially overlapping reports, and only one multimodal combined-model study had a complete harmonized 2×2 table. At a 40% pretest probability, a negative conventional US result reduced the posttest probability of residual nodal disease to approximately 26%.
Conclusions: Conventional US alone does not reliably exclude residual axillary nodal disease after NAC. AI/radiomics approaches show promising diagnostic performance, but the evidence for this is sparse, heterogeneous, and imprecise. The currently available data do not support definitive modality ranking or omission of axillary surgery on the basis of imaging alone, and prospective multicenter external validation is required.
Journal IF-equivalent: 1.9 (OpenAlex 2-year mean citedness, value as of 2026-10-12, retrieved 2026-10-13; not the official Clarivate IF)Reference: Liu T, He Y, Yuan X, Wang J, Liu Y, Cao C, et al. Diagnostic accuracy of ultrasound-based imaging modalities for assessing residual axillary lymph node disease after neoadjuvant chemotherapy in patients with breast cancer: a systematic review and meta-analysis. Quant Imaging Med Surg. 2026 Nov;16(11):868-8. doi:10.21037/qims-2026-1427.Checked: Abstract only - Abstract
Background/Objectives: Accurate molecular subtype characterization is essential for breast cancer management, but currently relies on invasive biopsy procedures with limited ability to capture tumor heterogeneity and the surrounding microenvironment. In this study, we investigated the feasibility of predicting breast cancer molecular subtypes directly from three-dimensional (3D) digital breast tomosynthesis (DBT) volumes using deep learning, while also evaluating the influence of different volumetric standardization strategies on classification performance.
Methods: Using the multicenter OPTIMAM database, a 3D ResNet-18 was trained to perform five-class molecular subtype classification (Luminal A, Luminal B1, Luminal B2, HER2-enriched, and triple-negative breast cancer (TNBC)) as well as one-vs-rest binary classification. To assess the impact of volumetric representation, different standardization strategies, including interpolation-based resampling and zero-padding approaches, were evaluated. In addition, Gradient-weighted Class Activation Mapping (Grad-CAM) was used to investigate the image regions contributing to model predictions.
Results: Classification performance varied according to the volumetric standardization strategy. Among the evaluated approaches, central padding provided the most consistent performance across repeated holdout evaluations, achieving a mean multiclass AUC of 64.07%, with TNBC reaching 70.39%. In the binary setting, HER2-enriched tumors achieved the highest AUC (76.02%). Furthermore, Grad-CAM analysis demonstrated that model activations frequently extended beyond the lesion itself and involved surrounding tissue regions.
Conclusions: Overall, these findings demonstrate that volumetric DBT combined with 3D deep learning enables non-invasive prediction of breast cancer molecular subtypes. The results further indicate that volumetric standardization influences classification performance and suggest that both intratumoral and peritumoral imaging information contribute to subtype discrimination, although generalization to previously unseen patients remains unestablished.
Journal IF-equivalent: 5.4 (OpenAlex 2-year mean citedness, value as of 2026-10-23, retrieved 2026-10-24; not the official Clarivate IF)Reference: Tavares R, Carrasquinha E, Mota AM. Impact of Volumetric Representation on Deep Learning-Based Molecular Subtype Prediction from 3D Digital Breast Tomosynthesis. Biomedicines. 2026 Oct 22;14(11):2411. doi:10.3390/biomedicines14112411.Checked: Abstract only - Abstract
This study aimed to develop a machine learning framework for objective ductal carcinoma in situ (DCIS) grading and to investigate intratumoral heterogeneity as a predictive biomarker. The grading framework was developed using 267 DCIS resections and validated on two external test sets (n=116 and n=69). A single-cell grading model graded individual nuclei using quantitative measures of nuclear size and pleomorphism. An integrated grading model then combined single-cell with architectural features in a support vector machine classifier. Separately, an invasion prediction model was trained on 288 intermediate- and high-grade DCIS cases and validated on 54 cases to predict upstaging to invasive carcinoma >1 mm. Compared to low-grade nuclei, high-grade nuclei were significantly larger and less uniformly oriented (all P 1 mm over the clinical-only model, increasing the area under the receiver operating characteristic curve from 0.704 to 0.827. These findings identify quantitatively measured intratumoral heterogeneity as a biologically relevant feature that may explain diagnostic ambiguity in borderline cases and help predict extensive invasive upstaging.
Journal IF-equivalent: 2.9 (OpenAlex 2-year mean citedness, value as of 2026-10-23, retrieved 2026-10-24; not the official Clarivate IF)Reference: Yang L, Chen J, Lei C, Yang M, Cai P, Fu J, et al. Machine learning-based quantitative pathology reveals intratumoral heterogeneity in ductal carcinoma in situ and predicts invasive upstaging. Am J Cancer Res. 2026;16(9):4102-14. doi:10.62347/szjy1956. PMID: 42908734.Checked: Abstract only