Visualization in 2D/3D registration matters for assuring technology-assisted image-guided surgery.

Area of Science:

• Medical Imaging
• Surgical Technology
• Human-Computer Interaction

Background:

• Image-guided navigation and surgical robotics are advancing minimally invasive surgery.
• Ensuring safety is paramount for deploying these technologies in clinical settings.
• 2D/3D registration is crucial for aligning preoperative and intraoperative data, but verification methods are needed.

Purpose of the Study:

• To develop and evaluate novel visualization paradigms for human assessment of 2D/3D registration accuracy.
• To investigate the impact of different visualization approaches on user performance in identifying registration errors.
• To enhance the safety assurance of image-guided surgery systems.

Main Methods:

• Developed novel visualization paradigms (Neutral, Attention-Guiding, Correspondence-Suggesting).
• Employed a sampling method based on approximate posterior distribution to simulate registration offsets.
• Conducted a user study with 22 participants evaluating 12 pelvic fluoroscopy images.

Main Results:

• All visualization paradigms improved performance over random guessing.
• Novel paradigms outperformed the neutral approach in differentiating acceptable/unacceptable registrations.
• Highest accuracy (70.4%) and F1 score (65.7%) were achieved with Attention-Guiding and Correspondence-Suggesting paradigms.

Conclusions:

• Visualization paradigms significantly influence human assessment of 2D/3D registration errors.
• Further research is required to optimize these methods for enhanced accuracy.
• This work is a key step towards improving safety in technology-assisted image-guided surgery.