A Priori Prediction of Breast Cancer Response to Neoadjuvant Chemotherapy Using CT Radiomics.

Area of Science:

• Medical Imaging
• Oncology
• Artificial Intelligence

Background:

• Neoadjuvant chemotherapy (NAC) response is crucial for breast cancer prognosis.
• Current evaluation relies on post-surgery pathology, delaying treatment adaptation.
• Predicting NAC response pre-treatment is essential for personalized therapy.

Purpose of the Study:

• To develop and validate a machine learning model integrating radiomic features from pre-treatment CT scans and clinical variables.
• To predict both pathologic complete response (pCR) and clinical response to NAC in breast cancer patients.
• To improve early prediction of treatment efficacy for personalized breast cancer management.

Main Methods:

• Extracted 214 radiomic features from intratumoral and peritumoral regions of pre-treatment contrast-enhanced CT scans.
• Incorporated 7 baseline clinical variables into the predictive model.
• Developed an XGBoost-based model and evaluated its performance using accuracy, AUC, and other metrics across ten independent data partitions.

Main Results:

• The combined clinical-radiomic model significantly outperformed models using only radiomic or clinical features for both pCR and clinical response prediction.
• For pCR classification, the combined model achieved 82.8% accuracy and an AUC of 0.846.
• For clinical response classification, the combined model achieved 71.7% accuracy and an AUC of 0.725.

Conclusions:

• Integrating CT-derived radiomic features with clinical data enhances the prediction of neoadjuvant chemotherapy response in breast cancer.
• This approach supports earlier and more personalized therapeutic decision-making.
• The findings highlight the potential of AI-driven imaging analysis in optimizing breast cancer treatment strategies.