CUSTrack: Causality-Inspired Liver Ultrasound Tracking With Periodic Motion Bias Mitigation.

Area of Science:

• Medical Imaging
• Ultrasound Technology
• Biomedical Engineering

Background:

• Real-time liver tissue tracking is crucial for ultrasound applications.
• Historical trajectories offer priors for localization but can introduce periodicity bias.
• Existing methods often over-rely on periodic motion, neglecting appearance matching.

Purpose of the Study:

• To develop a novel liver tracking method mitigating periodicity bias.
• To improve target localization accuracy and robustness in liver ultrasound.
• To leverage causal inference for enhanced trajectory analysis.

Main Methods:

• Proposed Causal Ultrasound-guided tracking (CUS-Track) method.
• Decomposition and correction of causal effects from historical trajectories.
• Counterfactual reasoning to eliminate periodicity bias.
• Incorporated deconfounding module for latent confounder removal.

Main Results:

• CUS-Track demonstrated superior tracking accuracy compared to existing methods.
• The method showed enhanced robustness under challenging ultrasound conditions.
• Effective suppression of periodicity bias while preserving useful trajectory priors.

Conclusions:

• CUS-Track offers a robust solution for real-time liver tracking.
• Causal inference provides a powerful framework for addressing bias in motion tracking.
• The proposed method advances the state-of-the-art in ultrasound-guided interventions.