Ultrasound Slice-to-Volume Registration via Artifact-Suppressed Ultra-Feature Learning.

Area of Science:

• Medical Imaging
• Surgical Navigation
• Ultrasound Technology

Background:

• Slice-to-volume registration (SVR) is critical for ultrasound (US)-guided liver interventions.
• Current SVR methods struggle with US imaging artifacts like rib shadowing and poor vasculature visualization, impacting accuracy.
• Liver interventions require precise targeting to ensure patient safety.

Purpose of the Study:

• To develop an advanced SVR method that overcomes limitations of current techniques in liver US imaging.
• To enhance pose estimation accuracy by simultaneously suppressing artifacts and improving feature extraction robustness.
• To improve the safety and efficacy of US-guided liver interventions through superior registration accuracy.

Main Methods:

• Proposed Artifact Suppressed Ultra-Feature guided SVR (ASUF-SVR) framework.
• Integrated Ultra-Feature Extraction (UFE) module with a dual-branch design to handle low signal-to-noise and contrast US images.
• Incorporated Artifact-Suppressed Evaluation (ASE) module to supervise similarity measurement and focus on anatomical structures.

Main Results:

• ASUF-SVR demonstrated consistent outperformance against state-of-the-art baselines across quantitative and qualitative evaluations.
• The method showed robustness on datasets with varying initial offsets and across different organs.
• Validated improved accuracy in SVR, crucial for minimizing mis-targeting risks.

Conclusions:

• ASUF-SVR significantly enhances pose estimation accuracy for liver US-guided interventions.
• The proposed method effectively suppresses artifacts and improves feature extraction in challenging US images.
• ASUF-SVR contributes to increased clinical safety in liver interventions by improving registration precision.