Entropy based detection of molecularly targeted nanoparticle ultrasound contrast agents in tumors

Abstract

In this study, we demonstrate the use of “joint entropy” of two random variables (X,Y) can be applied to markedly improve tumor conspicuity (where X = f(t) =backscattered waveform and Y = g(t) = a reference waveform; both differentiable functions). Previous studies have shown that a good initial choice of reference is a reflection of the original insonifying pulse taken from a stainless-steel reflector. Using this choice, joint entropy analysis is more sensitive to accumulation of targeted contrast agents than conventional gray-scale or signal energy analysis by roughly a factor of 2[Hughes, M. S., et al., J. Acoust. Soc. Am., 133(1), p 283, 2013]. We now derive an improved reference that is applied to three groups of (MDA-435, breast tumor) flank tumor-implanted athymic nude mice to identify tumor vasculature after binding perfluorocarbon nanoparticles (~250 nm) to neovascular avb3 integrins. Five mice received i.v.avb3-targeted nanoparticles, five received nontargeted nanoparticles, and five received saline at a dose of 1 ml/kg, which was allowed to circulate for up to two hours prior to imaging. Three analogous groups of nonimplanted mice were imaged in the same region following the same imaging protocol. Our results indicate an improvement in contrast by a factor of 2.5 over previously published results. Thus, judicious selection of the reference waveform is critical to improving contrast-to-noise in tumor environments when attempting to detect targeted nanostructures for molecular imaging of sparse features.