Enhancing Ultrasound Molecular Imaging: RPCA-Based Filtering to Differentiate Tumor-Bound and Free Microbubbles

In early cancer detection, ultrasound molecular imaging (UMI) offers enhanced accuracy regarding the presence, location, and extent of disease by utilizing targeted contrast agents such as microbubbles, exceeding the capabilities of conventional ultrasound. One challenge in UMI is differentiating between bound microbubbles attached to cancer biomarkers and free-floating microbubbles. A common method for detecting bound microbubbles involves a differential targeted enhancement technique, where a strong ultrasound pulse bursts the bubbles to highlight them by comparing pre- and post-burst images. However, this approach destroys the targeted contrast agents, thereby preventing further imaging of the region, and requires precise knowledge of the tumor’s location, which limits the application of UMI.In this work, we propose a non-destructive method to distinguish between bound and free microbubbles using robust pricipal component analysis (RPCA). We validated this approach on ten different simulated ultrasound time-series datasets, with results indicating that an RMS error of approximately 0.1 and a Dice coefficient ranging from 0.75 to 0.9, using between 12 and 26 image frames, can be achieved in reconstructing the image showing only the bound microbubbles. We present initial results from in vivo experiments using mouse models of breast cancer development with B7-H3-targeted microbubbles.