Dynamic Contrast-enhanced MRI for Evaluating Breast Cancer Chemotherapy Response Using Conditional Generative...

Area of Science:

• Radiology and Medical Imaging
• Artificial Intelligence in Medicine
• Oncology

Background:

• Dynamic Contrast-Enhanced Magnetic Resonance Imaging (DCE-MRI) is crucial for assessing vascular permeability.
• Pharmacokinetic modeling, like the extended Tofts model (ETM), provides quantitative permeability maps (Ktrans) but is computationally intensive.
• Accurate and efficient Ktrans mapping is vital for cancer diagnosis and treatment response assessment.

Purpose of the Study:

• To develop and evaluate an image-to-image conditional generative adversarial network (cGAN) for translating DCE-MRI data into vascular pharmacokinetic permeability maps.
• To compare the computational efficiency and accuracy of the cGAN-derived Ktrans maps against the established ETM.
• To assess the potential of the cGAN approach for predicting treatment response in breast cancer patients undergoing neoadjuvant chemotherapy.

Main Methods:

• A retrospective cohort of breast cancer DCE-MRI scans was used to train and validate the cGAN.
• The extended Tofts model (ETM) was employed to generate reference standard Ktrans maps.
• Linear regression, logistic regression, and paired t-tests were utilized to evaluate agreement, spatial similarity, and predictive capabilities.

Main Results:

• The cGAN achieved over a 1000-fold reduction in computation time compared to the ETM.
• cGAN Ktrans maps demonstrated excellent spatial agreement and high structural similarity to ETM maps (R² ≥ 0.98).
• The percentage change in cGAN Ktrans effectively distinguished patients with pathologic complete response (60% reduction) from those without (17% reduction) after neoadjuvant chemotherapy.

Conclusions:

• The developed DCE-to-pharmacokinetic cGAN offers a standardized and computationally efficient method for pharmacokinetic analysis in DCE-MRI.
• This AI-driven approach shows significant potential for early prediction of breast cancer response to neoadjuvant chemotherapy.
• The cGAN facilitates faster and potentially more accessible quantitative imaging biomarker analysis in clinical oncology.