Sparse channel sampling for ultrasound localization microscopy (SPARSE-ULM).

Area of Science:

• Biomedical imaging
• Ultrasound technology
• Neuroscience

Background:

• Ultrasound Localization Microscopy (ULM) offers high-resolution *in vivo* cerebral vasculature mapping.
• ULM generates large datasets, limiting acquisition, transfer, storage, and processing speeds.
• Current data transfer rates are a significant bottleneck for ULM technology.

Purpose of the Study:

• To develop a novel ULM reconstruction framework to reduce data acquisition.
• To decrease hardware complexity by randomly subsampling linear probe channels.
• To evaluate method performance and optimize parameters for efficient ULM.

Main Methods:

• Developed a reconstruction framework using random channel subsampling of a linear probe.
• Conducted *in silico* evaluations using SIMUS simulation software with realistic phantoms.
• Performed *in vivo* comparisons using rat brain acquisitions post-craniotomy.

Main Results:

• Reducing active channels slightly degrades signal-to-noise ratio and increases false microbubble detection rates.
• Localization accuracy is minimally affected by channel reduction, maintaining high precision.
• False positive rates increased from 3.7% (128 channels) to 11% (16 channels) in simulations.
• Average localization accuracy ranged from 9.93 μm (128 channels) to 10.6 μm (16 channels).

Conclusions:

• A balance exists between channel count and reconstructed vascular network quality in ULM.
• ULM is feasible with reduced channel counts, enabling high-resolution vascular mapping.
• This approach paves the way for developing cost-effective, high-resolution vascular mapping devices.