To present a dynamic uncertainty-level assessment framework for real-time needle tracking in CT-guided interventions, addressing existing challenges and enabling surgeons to make informed intraoperative decisions.
Approach:
Key Findings:
The framework outputs estimated needle position with a quantitative uncertainty percentage, improving transparency and trust, with a reported accuracy improvement of X% over traditional methods.
A linear calibration links uncertainty to spatial accuracy, facilitating real-time decision-making, with a demonstrated correlation coefficient of Y.
The proposed methods were trained and evaluated on two publicly available datasets with diverse needle manipulation scenarios, achieving Z% accuracy in real-time tracking.
Interpretation:
The framework provides a clinically interpretable measure of uncertainty, transforming model-derived uncertainty into a percentage-based reliability scale that directly informs surgical decisions.
Limitations:
The classic method requires no training data but may lack the robustness of CNN-based approaches, potentially leading to higher error rates in complex scenarios.
The hybrid CNN's performance may depend on the amount of training data available, which could limit its applicability in low-data environments.
Conclusion:
The proposed framework enhances real-time needle tracking reliability in CT-guided procedures, supporting safer surgical workflows and potentially improving patient outcomes.
