Deep Learning-Driven Radiomic Feature Extraction for Predicting Complete Pathological Response to Neoadjuvant...

Area of Science:

• Oncology
• Medical Imaging
• Artificial Intelligence

Background:

• Accurate prediction of pathological complete response (pCR) after neoadjuvant chemotherapy (NAC) is crucial for optimizing breast cancer treatment.
• 18F-FDG PET/CT imaging offers functional and anatomical insights into tumor response.
• Advanced radiomic and texture features from PET/CT may enhance prediction accuracy.

Purpose of the Study:

• To assess the predictive potential of 18F-FDG PET/CT parameters, including texture features, for pCR after the first NAC course in breast cancer.
• To evaluate the combination of imaging features, clinical data, and breast cancer molecular subtypes for improved pCR prediction.
• To develop an AI-driven approach for early treatment response assessment.

Main Methods:

• 204 breast cancer patients undergoing 18F-FDG PET/CT for NAC assessment were retrospectively analyzed.
• Automated lesion segmentation using the nnUNet deep learning model.
• Machine learning classifiers (Random Forest, XGBoost, SVM) were employed using metabolic, radiomic, texture features, and molecular subtype data.

Main Results:

• The combined model integrating PET/CT features (baseline and follow-up) with breast cancer subtype achieved the highest prediction performance (mean balanced accuracy 0.76 ± 0.09).
• This combined approach outperformed models using only HER2 (0.67 ± 0.08) or TN (0.65 ± 0.06) subtype data.
• Features from baseline and follow-up scans, combined with subtype information, demonstrated strong predictive value for pCR.

Conclusions:

• Integrating baseline and follow-up 18F-FDG PET/CT radiomic and texture features with breast cancer subtype information significantly improves pCR prediction after the first NAC course.
• This AI-driven approach enables more accurate early assessment of treatment response.
• The findings support personalized therapy decisions, potentially avoiding ineffective treatments and improving patient outcomes.