Interpretable Dynamic MRI-ITH Model for Predicting Neoadjuvant Chemotherapy Response in Breast Cancer: A Multicenter...

Area of Science:

• Oncology
• Radiology
• Medical Imaging

Background:

• Neoadjuvant chemotherapy (NAC) is crucial for breast cancer treatment.
• Accurate early prediction of pathological complete response (pCR) is vital for treatment optimization.
• Intratumoral heterogeneity (ITH) assessed by MRI may offer insights into treatment response.

Purpose of the Study:

• To develop and validate an interpretable model for early pCR assessment in breast cancer patients receiving NAC.
• To integrate dynamic MRI-derived intratumoral heterogeneity (ITH) scores with clinicopathological features.
• To evaluate the model's performance across multiple validation sets.

Main Methods:

• Prospective enrollment of 400 breast cancer patients across three centers.
• Calculation of baseline (ITH0) and dynamic changes (MRI-ΔITH1, MRI-ΔITH2) in MRI-ITH scores.
• Development of seven predictive models using logistic regression and interpretation with SHAP analysis.

Main Results:

• The best performing model integrated clinical features and MRI-ΔITH2, achieving AUCs of 0.940 (training), 0.873 (internal validation), and 0.917 (external validation).
• SHAP analysis identified MRI-ΔITH2 (31.7%), PR status (24.3%), and HER-2 status (18.8%) as key predictive factors.
• The model demonstrated robust performance in predicting pCR.

Conclusions:

• A predictive model combining dynamic MRI-ITH scores and clinicopathological features shows promise for early pCR assessment in breast cancer patients post-NAC.
• The model's interpretability aids in understanding key predictive factors.
• Further multicenter validation is recommended prior to clinical implementation.