# Unsupervised Beamforming with Optimized Coherence Loss for Clutter Suppression in Single Plane-Wave Ultrasound Imaging

## TL;DR

This paper introduces a new unsupervised beamforming method that improves ultrasound image quality by reducing clutter artifacts in real-time.

## Contribution

The novel approach uses optimized deep coherence loss with physics-based criteria for adaptive clutter suppression in single plane-wave ultrasound.

## Key findings

- UBF-DCLopt improved contrast-to-noise ratio by 22% in phantom experiments.
- In vivo results showed a 32% improvement in contrast-to-noise ratio compared to fixed coherence loss methods.
- The method maintains high spatiotemporal resolution while suppressing reverberation artifacts.

## Abstract

Background: Single plane-wave ultrasound imaging (SPWI) enables acquisition speeds exceeding 1000 Hz, making it suitable for real-time applications requiring high temporal resolution. However, SPWI suffers from clutter artifacts, such as multipath reverberations, which degrade image contrast and diagnostic reliability. Methods: To address this limitation, we propose an unsupervised beamforming approach based on optimized deep coherence loss (UBF-DCLopt), which adaptively performs signal coherence computation according to the inter-frame decorrelation of plane-wave data. In addition, optimal plane-wave frames for coherence loss calculation are adaptively determined by physics-based criteria that account for steering angle and broadband pulse characteristics. To evaluate the proposed method, simulation, phantom and in vivo studies were conducted. For training and validation, publicly available datasets and data acquired from a fabricated clutter phantom were employed. Results: Experimental results demonstrated that the proposed UBF-DCLopt achieved contrast-to-noise ratio (CNR) improvements of 22% in phantom experiments and 32% in the in vivo studies compared to an unsupervised beamforming method using fixed deep coherence loss (UBF-DCL). Conclusions: These results demonstrate that the physics-informed unsupervised approach significantly suppresses reverberation artifacts while maintaining high spatiotemporal resolution, thereby enabling enhanced diagnostic accuracy in real-time ultrasound imaging.