Detection of early therapeutic response with dynamic contrast enhanced ultrasound using a perfusion clustering algorithm

Dynamic contrast enhanced ultrasound (DCE-US) has been applied in clinical and pre-clinical studies, to monitor microvascular tumoral response to antiangiogenic therapies. Flow quantification in 2D+t DCE-US is generally performed by averaging the signal in large regions of analysis (ROAs), followed by the regression of a parametric model to the obtained echo-power versus time curves. However, this approach can lead to a loss of spatial information about the microvascular flow distribution throughout the tumor if this distribution is spatially heterogeneous. This work adapts a clustering algorithm (DynClust) developed previously for evaluation of contrast-enhanced computed tomography to DCE-US data. The sensitivity and specificity of the algorithm are assessed on a bi-compartmental simulated data set, the relevance of the identified perfusion clusters is assessed on DCE-US acquisition in murine kidneys. Its capacity to detect differences between a group of 27 mice treated with placebo and a group of 20 mice treated with an anti-angiogenic is assessed. The number of clusters Kclust derived from the clustering clustering algorithm are related to the heterogenity of the perfusion in the tumour. Results demonstrate ability to significantly discriminate (p = 0.011, Mann-Whitney U test) between the therapy groups while, among the conventional DCE-US parameters, only the AUC provided significant discrimination (p = 0.049).