Utilizing Longitudinal DCE-MRI and Tumor Microenvironment Data for Predicting Neoadjuvant Therapy Outcomes in Breast Cancer Through Deep Learning Techniques - Summary - MDSpire
Advertisement
Utilizing Longitudinal DCE-MRI and Tumor Microenvironment Data for Predicting Neoadjuvant Therapy Outcomes in Breast Cancer Through Deep Learning Techniques
To develop and validate a multimodal fusion model for early and accurate prediction of pathological complete response (pCR) to neoadjuvant therapy (NAT) in breast cancer, enhancing treatment planning.
Key Findings:
The combined model achieved an AUC of 0.90 and specificity of 95% in the validation cohort, indicating its strong predictive capability.
The Post-2nd-NAT DL model outperformed the Pre-NAT DL model (AUC = 0.85 vs. AUC = 0.75), highlighting the importance of early treatment imaging.
The immune-inflammation model showed independent predictive capability (AUC = 0.73), suggesting its potential role in clinical assessments.
Interpretation:
The multimodal fusion model significantly enhances early prediction of pCR to NAT, providing a potential tool for personalized treatment planning in breast cancer, especially compared to existing predictive methods.
Limitations:
Retrospective study design may introduce bias; future studies should consider prospective designs.
Limited generalizability due to single-center data; multicenter studies are needed to validate findings.
Conclusion:
Integrating deep learning features from DCE-MRI, dynamic inflammatory indicators, and baseline TILs improves prediction accuracy for treatment response in breast cancer patients undergoing NAT, emphasizing the need for personalized treatment strategies.
