Visualization of Tracking Uncertainty in AR-based Surgical Guidance

Paper Title

Visualization of Tracking Uncertainty in AR-based Surgical Guidance

Publication Info

• Topic area: Augmented Reality (AR) for surgical guidance and uncertainty visualization.
• Keywords: Augmented Reality, surgical guidance, uncertainty visualization, tracking errors, Cone visualization, Circle visualization, Gauge visualization, Color visualization, AR-based drilling, user confidence.

Background and Problem

• Problem / challenge: Tracking errors in AR-based surgical systems can lead to misalignment between virtual overlays and the physical environment, affecting precision and reliability. Current solutions to improve tracking accuracy often require expensive hardware or extensive calibration, which are impractical in surgical workflows.
• Significance: Addressing tracking uncertainty is critical for improving surgical precision, reducing complications, and enhancing patient outcomes. Visualizing uncertainty may help surgeons adapt their strategies and improve task performance.
• Motivation and related work: Prior research has shown that visualizing uncertainty improves decision-making and trust in other domains, such as automated vehicles and forecasting. However, there is limited understanding of how uncertainty visualization impacts AR-guided surgical tasks, particularly tool tracking and manipulation.

Solution

• Proposed approach: Development and evaluation of four uncertainty visualization techniques (Cone, Circle, Gauge, and Color) for AR-based surgical drilling tasks.
• Novelty:
  1. Design of four uncertainty visualization techniques tailored to AR-based surgical drilling tasks.
  2. Empirical evidence showing that visualizing uncertainty improves drilling precision, trust, and confidence.
  3. Recommendations for integrating uncertainty visualization into AR surgical systems.
• Procedure and key techniques:
  • A two-part study was conducted: the first with 24 non-professional participants and the second with 4 surgeons.
  • Simulated tracking errors were introduced to evaluate the impact of the visualizations on task performance.
  • Objective measures (e.g., drilling accuracy, task completion time) and subjective measures (e.g., trust, confidence, preference) were collected and analyzed.

Results

• Concrete findings:
  • Uncertainty visualization improved drilling accuracy by 24% for non-professionals and 28% for surgeons.
  • Task completion time increased by 76% for non-professionals and 196% for surgeons.
  • Cone and Circle visualizations were the most effective and preferred techniques.
• Advantage over baselines:
  • All visualizations reduced tracking error compared to the baseline (no visualization), with Cone and Circle yielding the highest accuracy improvements.
  • Cone significantly improved overall drilling accuracy, while Circle minimized tracking error at the target point.
• Experiments / evaluation:
  • Experiment 1: 24 non-professionals performed 480 trials across five conditions (None, Cone, Circle, Gauge, Color).
  • Experiment 2: 4 surgeons performed 80 trials under the same conditions using a simulated knee bone.
  • Objective metrics: Total Added Error, Guided Deviation, Ground Truth Deviation, Task Completion Time.
  • Subjective metrics: trust, confidence, ease of use, preference rankings.
• Limitations and future work:
  • Simulated tracking errors were not influenced by user actions, which frustrated some participants.
  • Combining visualizations (e.g., Cone with Color) could provide complementary benefits but requires careful design to avoid cognitive overload.
  • The study included only four surgeons; larger expert cohorts are needed for robust validation.
  • Future research should explore real-time computation of uncertainty, adaptive visualizations, and cognitive workload assessments.

Summary

This study evaluated four uncertainty visualization techniques (Cone, Circle, Gauge, Color) for AR-based surgical drilling tasks. Results showed that visualizing uncertainty improved drilling accuracy (24–28%) and increased user confidence and trust, though it also increased task completion time. Cone and Circle were the most effective and preferred visualizations. The findings highlight the importance of integrating uncertainty visualizations into AR surgical systems, with Cone being particularly recommended for its ability to convey uncertainty along the entire trajectory. Future work should explore real-time uncertainty computation, combined visualizations, and larger-scale validation with surgeons.