Denoising Echocardiography with an Improved Diffusion Model

Area of Science:

• Medical Imaging
• Computer Vision
• Artificial Intelligence

Background:

• Echocardiography is vital for cardiac diagnosis but often suffers from poor image quality due to acoustic interference and artifacts.
• Noise and artifacts in ultrasound images hinder accurate interpretation and diagnostic capabilities.
• Effective despeckling techniques are crucial for enhancing ultrasound image interpretability.

Purpose of the Study:

• To propose a diffusion model-based denoising method for enhancing ultrasound image quality.
• To improve the accuracy and diagnostic value of cardiac ultrasound imaging.
• To develop an unsupervised method for ultrasound image enhancement.

Main Methods:

• A diffusion model-based denoising approach utilizing an interpolation technique and a U-Net architecture.
• Generation of interim images by interpolating noise-free and noisy images at each diffusion step.
• Unsupervised training and validation on two benchmark datasets.

Main Results:

• The proposed method effectively reduces noise while preserving image texture.
• Experimental results demonstrate superior performance compared to other denoising approaches.
• Clinically relevant qualitative and quantitative visual metrics were improved.

Conclusions:

• The diffusion model-based denoising method significantly enhances ultrasound image quality.
• The approach offers a promising solution for improving cardiac diagnosis through clearer ultrasound imaging.
• The developed technique preserves essential image textures, leading to more reliable diagnostic information.