RF-ULM: Ultrasound Localization Microscopy Learned From Radio-Frequency Wavefronts

Area of Science:

• Medical Imaging
• Biophysics
• Signal Processing

Background:

• Ultrasound Localization Microscopy (ULM) requires precise particle localization for high-resolution imaging.
• Traditional delay-and-sum beamforming in ULM reduces valuable Radio-Frequency (RF) channel data, limiting localization potential.
• The implications of RF data reduction on localization accuracy are not well understood.

Purpose of the Study:

• To develop a method for directly localizing scatterers using raw RF channel data in ULM.
• To leverage the rich information within RF wavefronts (shape, phase) for improved localization accuracy.
• To assess the impact of beamforming on ULM and compare the proposed method against state-of-the-art techniques.

Main Methods:

• A custom super-resolution Deep Neural Network (DNN) was designed, incorporating feature channel shuffling and a semi-global convolutional block.
• Non-maximum suppression was employed for reliable wavefront localization.
• A geometric point transformation was introduced for accurate mapping to B-mode coordinates.

Main Results:

• The proposed RF-ULM method demonstrates high precision and reduced complexity compared to state-of-the-art techniques.
• The method effectively bridges the domain shift between synthetic and real-world ultrasound data.
• The first in vivo results of a wavefront-localizing DNN in ULM were successfully presented, showcasing practical applicability.

Conclusions:

• Directly localizing scatterers in RF channel data offers significant advantages for ULM.
• The developed DNN-based approach enhances localization precision and overcomes limitations of traditional beamforming.
• The open-source release of code and methods will benefit the wider research community.