Denoising Plane Wave Ultrasound Images Using Diffusion Probabilistic Models.

Area of Science:

• Medical Imaging
• Ultrasound Technology
• Image Processing

Background:

• High frame-rate ultrasound imaging, particularly plane wave (PW) imaging, offers advanced visualization capabilities.
• A significant challenge in high frame-rate ultrasound is the high noise level, which compromises image quality and limits clinical adoption.
• Effective denoising methods are crucial for improving the utility of PW ultrasound images.

Purpose of the Study:

• To develop and evaluate a novel denoising method for plane wave (PW) ultrasound imaging.
• To enhance the quality of high frame-rate ultrasound images by reducing noise.
• To adapt denoising diffusion probabilistic models (DDPMs) for beamformed radio frequency (RF) data in ultrasound.

Main Methods:

• Adapted a denoising diffusion probabilistic model (DDPM) to process beamformed radio frequency (RF) data.
• Trained the model using a dataset of 400 simulated ultrasound images.
• Utilized natural image segmentation masks as intensity maps for generated images, enabling accurate denoising across diverse anatomical structures.

Main Results:

• The proposed method effectively reduced noise in simulated plane wave (PW) ultrasound images.
• Demonstrated significant improvements in image quality for phantom and in vivo ultrasound data.
• Comparative analysis showed superior performance against existing denoising techniques across multiple evaluation metrics.

Conclusions:

• The DDPM-based denoising approach successfully enhances the quality of high frame-rate ultrasound images.
• The method is effective for simulated, phantom, and in vivo data, addressing noise challenges in PW imaging.
• This technique holds promise for broader adoption of high frame-rate ultrasound imaging in clinical practice.