Robotic Compounding for Whole-Brain Non-Invasive 3D Ultrasound Localization Microscopy

Abstract

3D ultrasound localization microscopy (ULM) allows the extraction of anatomical and functional representations of vascular networks with a spatial resolution beyond the diffraction limit ( $\sim \lambda $ /10) by localizing injected microbubble (MB) contrast agents and tracking their positions over time. To advance this technology towards clinical diagnostics, the ability to obtain a large field of view (FOV) becomes a pressing necessity. One solution for large FOV imaging is automated stitching/compounding of multiple volume acquisitions. This is challenging for full-brain imaging, as the acquisition of images through the skull requires parallel positioning of the transducer surface relative to the skull to optimize ultrasound transmission. Herein, we demonstrate an automated positioning system that relies on predefined optimized orientations, enabling fast acquisition and positioning for rapid full-brain imaging. As an example of expanded FOV application, we achieved non-invasive full-brain imaging of an 8-week-old rat by collecting data across 11 transducer positions. To ensure optimal acoustic penetration through the intact skull, the transducer orientation was robotically positioned. We compared this approach with pure transducer translation. Additionally, we acquired whole-brain vasculature images from 4-week-old rats using 24 100-second scans of optimized transducer positions, comparing these results to a single-position 2400-second scan. Automated robotic compounding enabled the acquisition of full-brain vascular information while minimizing acquisition dead time. Optimized transducer angles enhanced the vascular network visualization across the brain, including challenging areas such as the cerebellum (10x improvement) and hindbrain (3.5x improvement). Moreover, our multi-position acquisition method allowed us to capture approximately four times more vascular volume transcranially, covering the entire rat brain, compared to the ~1 cm3 typically obtained with single-position acquisitions using the same transducer. This work demonstrates the benefit of automated robot-assisted multi-angle/multi-position acquisitions in ULM to acquire a volumetric field of view larger than otherwise possible with a single position acquisition, especially those through the skull.