Early Prediction of Breast Cancer Therapy Response using Multiresolution Fractal Analysis of DCE-MRI Parametric Maps

Area of Science:

• Radiology
• Medical Imaging
• Oncology

Background:

• Neoadjuvant chemotherapy (NACT) is a standard treatment for breast cancer.
• Predicting treatment response early is crucial for optimizing patient outcomes.
• Dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) provides insights into tumor vascularity.

Purpose of the Study:

• To evaluate the efficacy of multiresolution fractal analysis of DCE-MRI parametric maps for early prediction of breast cancer response to NACT.
• To compare the predictive performance of multiresolution fractal analysis with conventional texture analysis methods.

Main Methods:

• 55 breast cancer patients underwent DCE-MRI before, during, and after NACT.
• Parametric maps were generated using the shutter-speed model.
• Multiresolution fractal analysis and conventional methods (single-resolution fractal, GLCM, RLM) were applied to extract features.
• Support vector machine classified responses based on features from the first NACT cycle.

Main Results:

• Multiresolution fractal features demonstrated superior predictive performance compared to conventional methods.
• Area under the curve (AUC) values reached 0.91 (all parameters) and 0.80 (Ktrans) in training and testing sets, respectively.
• Statistically significant differences (P < .05) were observed for several parametric maps.

Conclusions:

• Multiresolution fractal analysis effectively captures changes in tumor vascular heterogeneity from DCE-MRI.
• This method offers a promising approach for early prediction of NACT response in breast cancer.
• Enhanced texture decomposition at various scales improves predictive accuracy.