Medical ultrasound image speckle reduction and resolution enhancement using texture compensated multi-resolution...

Area of Science:

• Medical Imaging
• Image Processing
• Biomedical Engineering

Background:

• Ultrasound (US) imaging is widely used in healthcare but suffers from noise and low resolution, degrading image quality, especially in low-cost systems.
• Speckle and other noise artifacts are inherent limitations of ultrasound imaging, impacting diagnostic accuracy.
• Existing methods may over-smooth images, losing crucial anatomical and textural information.

Purpose of the Study:

• To propose a novel method for enhancing ultrasound image quality.
• To simultaneously improve image resolution and suppress noise in ultrasound images.
• To preserve essential anatomical features and texture information during image enhancement.

Main Methods:

• A novel image enhancement method is proposed, incorporating texture compensation to retain anatomical details.
• The method simultaneously addresses resolution enhancement and noise suppression in ultrasound images.
• Ultrasound image formation physics knowledge is utilized to generate augmented datasets for training the enhancement model.

Main Results:

• The proposed method effectively enhances the overall quality of ultrasound images.
• Simultaneous resolution enhancement and noise suppression were achieved.
• Texture compensation successfully preserved important anatomical structures and texture information.
• The use of physics-based augmented datasets improved the training and performance of the proposed network.

Conclusions:

• The developed method significantly improves ultrasound image quality by enhancing resolution and reducing noise.
• Texture compensation is crucial for preserving diagnostic information in enhanced ultrasound images.
• Leveraging ultrasound physics for data augmentation is an effective strategy for training deep learning models in medical imaging.