FastCF: A Fast Coupled Filtering Method for Ultrasound-Based Tissue Deformation Analysis

In emergency rooms or intensive care units (ICUs), ultrasound imaging is the primary 3-D real-time imaging modality adopted for tissue deformation analysis. However, the accuracy of ultrasound-based analysis tools is severely affected by a phenomenon called feature motion decorrelation, making the feature-based tracking methods unable to faithfully reveal the underlying tissue deformation. We have developed a coupled filtering method to compensate for feature motion decorrelation analytically, and it has achieved much higher accuracy than existing methods. However, its 3-D implementation is computationally too expensive.

In this paper, we propose an accelerated version of the 3-D coupled filtering method named FastCF for near-real-time applications.

FastCF is 4.23 billion times faster and can finish the analysis within 135 s on one NVIDIA A100 GPU.

FastCF can be directly combined with existing ultrasound imaging systems to facilitate quantitative diagnosis since it does not require specialized ultrafast ultrasound imaging systems.