AI and Cancer Research ── Breast ── 2026-11-09

Journals covered and how papers are chosen: see the index

Breast cancer: treatment, trials and AI

• A multicenter clinicopathologic and MRI-based multimodal fusion model for predicting axillary lymph node metastasis in breast cancer
  Cancer Imaging · 2026-11-06 · Evidence: Observational study (two-center prediction model development with external validation)
  Abstract

  Background: Preoperative prediction of axillary lymph node (ALN) metastasis remains challenging. We developed and externally validated a DCE-MRI-based multimodal fusion model integrating clinical, radiomics, habitat, and deep learning features.

  Methods: A total of 404 patients from two centers were included. Center A ( n = 284) was split into training ( n = 198) and validation ( n = 86) sets; Center B ( n = 120) served as an independent external test cohort. Four sub-models (random forest for clinical variables, ExtraTrees for whole-tumor and habitat radiomics, ResNet50) were combined by weighted late fusion. Progressive fusion, a sequential ablation introducing modalities one at a time, tested whether each stream contributed non-redundant information; all 15 combinations were additionally evaluated. SHAP was applied to the feature-based sub-models and Grad-CAM to the CNN.

  Results: On external testing, the fusion model achieved an AUC of 0.847 (95% CI 0.771–0.921), significantly higher than the clinical model (0.754; DeLong P = 0.017). Differences versus the radiomics (0.797), deep learning (0.784), and habitat (0.817) sub-models were numerical rather than statistically significant ( P = 0.074, 0.066, and 0.248). Progressive fusion showed stepwise gains (0.754, 0.807, 0.812, 0.847), and exhaustive combination analysis indicated that each modality contributed non-redundant information. SHAP identified T stage as the dominant clinical predictor, and the most influential habitat features originated from three different subregions. Grad-CAM showed tumor-focused attention.

  Conclusion: This multimodal fusion model incorporating habitat radiomics achieved good discrimination for preoperative ALN prediction with external validation, and ablation confirmed each feature stream contributed non-redundant information.

  Journal IF-equivalent: 5.7 (OpenAlex 2-year mean citedness, value as of 2026-11-04, retrieved 2026-11-05; not the official Clarivate IF)
  Reference: Wang W, Li Y, Zou X, Ji G, Peng J. A multicenter clinicopathologic and MRI-based multimodal fusion model for predicting axillary lymph node metastasis in breast cancer. Cancer Imaging. 2026 Nov 6 [Epub ahead of print]. doi:10.1186/s40644-026-01152-3.
  Checked: Abstract only
• Fraction genome altered outperforms tumor mutational burden for machine learning-based mortality prediction across TP53, PIK3CA, GATA3, and MAP3K1 subtypes in invasive ductal carcinoma.
  Wiener medizinische Wochenschrift (1946) · 2026-11-06 · Evidence: Retrospective observational study (analysis of the AACR GENIE cohort)
  Abstract

  BACKGROUND: Invasive ductal carcinoma (IDC) shows marked heterogeneity in clinical outcomes, often not fully captured by standard molecular subtypes. Tumor mutational burden (TMB) has limited prognostic utility in IDC due to low mutation rates and failure to account for structural genomic alterations. Fraction genome altered (FGA), reflecting copy number changes, may better represent genomic instability and disease aggressiveness. METHODS: We analyzed 3396 IDC patients from the AACR GENIE cohort, focusing on four key gene mutations (TP53, PIK3CA, GATA3, MAP3K1). Machine learning models (logistic regression and support vector machine) were trained using FGA or TMB as the main predictors, adjusted for clinical variables, to predict mortality. RESULTS: Fraction genome altered consistently outperformed TMB in mortality prediction across all gene subtypes. In TP53-mutated patients, FGA improved the area under the receiver operating characteristic curve (AUROC; 0.70 vs. 0.69) and accuracy (67% vs. 64%) compared to TMB. In MAP3K1, FGA achieved a significant AUROC increase (0.75 vs. 0.49), indicating a shift from poor to strong discriminative capacity. High negative predictive values were observed with FGA, reaching 85% in GATA3 and 83.3% in MAP3K1, indicating strong reliability in identifying survivors. CONCLUSION: Fraction genome altered is a more robust and interpretable predictor of mortality than TMB in IDC. Its consistent performance across diverse genomic subtypes highlights its potential as a clinically actionable biomarker for risk stratification and personalized patient management in breast cancer. ZUSAMMENFASSUNG: HINTERGRUND: Das invasive duktale Karzinom (IDC) weist eine deutliche Heterogenität bei den klinischen Ergebnissen auf, die oftmals nicht vollständig durch die molekularen Standardsubtypen erfasst werden. Die Tumormutationslast (TMB) hat beim IDC nur begrenzten prognostischen Nutzen aufgrund niedriger Mutationsraten und fehlender Berücksichtigung struktureller genomischer Alterationen. Der Anteil des veränderten Genoms („fraction genome altered“, FGA) spiegelt die Kopienzahlveränderungen wider und könnte genomische Instabilität und Krankheitsaggressivität besser darstellen. METHODEN: Aus der AACR-GENIE-Kohorte wurden die Daten von 3396 IDC-Patienten ausgewertet, mit Fokus auf 4 entscheidenden Genmutationen (TP53, PIK3CA, GATA3, MAP3K1). Modelle des maschinellen Lernens (logistische Regression und Support Vector Machine) wurden anhand von FGA oder TMB als den wesentlichen Prädiktoren unter Berücksichtigung klinischer Variablen trainiert, um die Mortalität vorherzusagen. ERGEBNISSE: Dabei war FGS über alle Gensubtypen hinweg durchgehend aussagekräftiger als TMB bei der Mortalitätsvorhersage. Bei Patienten mit TP53-Mutation verbesserten sich mit FGA die Fläche unter der Receiver-Operating-Kurve (AUROC; 0,70 vs. 0,69) und die Genauigkeit (67% vs. 64%) im Vergleich zu TMB. Bei MAP3K1 erzielte FGA einen signifikanten AUROC-Anstieg (0,75 vs. 0,49), was auf eine Verlagerung von einer schwachen zu einer starken diskriminativen Leistung hinweist. Für FGA wurden hohe negative prädiktive Werte festgestellt, sie erreichten 85% bei GATA3 und 83,3% bei MAP3K1, was auf eine starke Reliabilität in Bezug auf die Erkennung von Überlebenden hinweist. SCHLUSSFOLGERUNG: Die FGA ist ein robusterer und besser interpretierbarer Prädiktor der Mortalität als TMB bei IDC. Ihre durchgehende Leistung über verschiedene genomische Subtypen hinweg betont ihr Potenzial als klinisch nutzbarer Biomarker für die Risikostratifizierung und die personalisierte Patientenversorgung bei Brustkrebs.

  Journal IF-equivalent: 1.2 (OpenAlex 2-year mean citedness, value as of 2026-11-07, retrieved 2026-11-08; not the official Clarivate IF)
  Reference: Kumari B, Lahariya R, Anand G. Fraction genome altered outperforms tumor mutational burden for machine learning-based mortality prediction across TP53, PIK3CA, GATA3, and MAP3K1 subtypes in invasive ductal carcinoma. Wien Med Wochenschr. 2026 Nov 6 [Epub ahead of print]. doi:10.1007/s10354-026-01202-2.
  Checked: Abstract only

← Back to this issue