Clinical Report: Predicting Neoadjuvant Therapy Outcomes in Breast Cancer

Overview

This study presents a multimodal fusion model that integrates deep learning features from longitudinal DCE-MRI, peripheral blood inflammatory indices, and baseline tumor-infiltrating lymphocytes to predict pathological complete response (pCR) in breast cancer patients undergoing neoadjuvant therapy. The combined model demonstrated superior predictive performance compared to single-modality models.

Background

Neoadjuvant therapy (NAT) is crucial for patients with locally advanced breast cancer, influencing surgical decisions and patient outcomes. Accurate early prediction of treatment response is essential to optimize individualized therapy and minimize unnecessary toxicity. Current clinical practice lacks reliable non-invasive biomarkers for predicting treatment efficacy early in the treatment course.

Data Highlights

Key Findings

• The combined model achieved an AUC of 0.90 and specificity of 95% in predicting pCR.
• The Post-2nd-NAT DL model outperformed the Pre-NAT DL model (AUC 0.85 vs. 0.75).
• The immune-inflammation model independently predicted pCR with an AUC of 0.73.
• Deep learning features from early DCE-MRI are critical for enhancing prediction accuracy.
• Multimodal fusion strategies can aid in personalized treatment planning for breast cancer patients.

Clinical Implications

The integration of deep learning features from imaging and inflammatory markers can significantly enhance the prediction of treatment response in breast cancer. This approach may facilitate more personalized treatment strategies, allowing clinicians to tailor neoadjuvant therapy based on predicted outcomes.

Conclusion

The multimodal fusion model represents a promising advancement in the early prediction of pCR to neoadjuvant therapy in breast cancer, potentially improving treatment planning and patient outcomes.

References

1. European Radiology, 2024 -- Evaluation of Anti-HER2 Treatment Response for Tailoring Therapy in Early HER2-Positive Breast Cancer Utilizing an Innovative Deep Learning Radiomics Approach
2. European Radiology, 2025 -- Radiomics Utilizing Deep Learning Fails to Enhance Prediction of Residual Cancer Burden Following Chemotherapy in the LIMA Breast MRI Study
3. European Radiology, 2025 -- A Deep Learning Approach for Classifying Grade 2 Nottingham Histologic Breast Tumors Using Dynamic Contrast-Enhanced MRI
4. asco ai in oncology -- Improved Immunotherapy Response Prediction in NSCLC With Deep-Learning Radiomic Biomarker
5. NCCN Guidelines® Insights: Breast Cancer, Version 5.2025 - PubMed
6. Early prediction of neoadjuvant therapy response in breast cancer using MRI-based neural networks: data from the ACRIN 6698 trial and a prospective Chinese cohort - PubMed
7. Tumor-Infiltrating Lymphocyte Scoring in Neoadjuvant-Treated Breast Cancer
8. NCCN Guidelines® Insights: Breast Cancer, Version 5.2025 - PubMed
9. Early prediction of neoadjuvant therapy response in breast cancer using MRI-based neural networks: data from the ACRIN 6698 trial and a prospective Chinese cohort - PubMed
10. Tumor-Infiltrating Lymphocyte Scoring in Neoadjuvant-Treated Breast Cancer