Multiparametric Microvascular Ultrasound to Classify Tumor Sensitivity to Anti-Angiogenic Treatment: Application to Multiple Cell Lines

The potential of microvascular ultrasound techniques for detecting anti-angiogenic treatment response in cancer models is widely recognized. This study assesses the usefulness of the perfusion parameters from SVD-based optimal shrinkage clutter filtered power Doppler (PD) and statistical histogram-based contrast-enhanced ultrasound (CEUS) images, compared to perfusion parameters from conventional PD and mean-intensity-based CEUS images, in improving the antiangiogenic tumor treatment response classification of cell lines with differing sensitivity level. Eleven perfusion parameters and the tumor volume feature are fed into a multivariable logistic regression learning model to classify antiangiogenic treatment responses of renal cell carcinoma cell lines, Caki-1 (very sensitive) and ACHN (resistant), engrafted on a chicken embryo assay. Results indicated that the model with selected parameters from statistical CEUS analysis, SVD-filtered 2D PD images outperformed a model using all microvascular features and a classifier with features from conventional analysis of CEUS and PD only. Therefore, the perfusion parameters from SVD-based PD and statistical CEUS method are best used as a supplement to conventional PD and CEUS analysis.