Comparative analysis of motion correction algorithms for contrast-enhanced ultrasound imaging

Ultrasound imaging is widely used in medical diagnostics due to its speed, safety, and cost-effectiveness. However, restrictions like speckle noise and limited scan lines might cause a loss of image detail. Contrast-enhanced ultrasound (CEUS) overcomes these drawbacks by using microbubbles to improve tissue perfusion imaging and tumor identification. Despite its advantages, CEUS is highly prone to motion artifacts, particularly in dynamic applications such as Dynamic Contrast-Enhanced Ultrasound (DCE-US), where patient or probe movement can significantly impact image quality. In conventional DCE-US, motion artifacts are frequently addressed by discarding the motion-affected frames, resulting in the loss of valuable data. To preserve image integrity and improve clinical analysis, recent advances have focused on developing motion correction algorithms. This study assesses four such algorithms—Convolutional Neural Network (CNN), U-Net, sub-pixel registration, and 2-tier image registration—using quantitative metrics such as Root Mean Square Error (RMSE), Correlation Coefficient (CC), and Structural Similarity Index Measure (SSIM). The purpose of this study is to find the most effective way to reduce motion artifacts while improving the qualitative assessment of medical image analysis.