Geometric Concordance of Deterministic Tract Planning With Operator-Selected Skin Entry Sites in CT-Guided Lung...

To test the hypothesis that a deterministic tract-planning framework generates anatomically feasible skin entry locations spatially concordant with operator-selected entry sites during CT-guided lung biopsy.

This single-center retrospective study included 199 consecutive CT-guided lung biopsies performed between January 2022 and June 2025 (mean age, 66.6 ± 9.5 years; 20.6% female). Procedural CT images were analyzed using a deterministic tract-planning system implemented in 3D Slicer with AI-assisted whole-thorax segmentation (TotalSegmentator) and predefined anatomic and geometric constraints. Literature-reported odds ratios for pneumothorax and pulmonary hemorrhage were encoded as fixed coefficients to characterize procedural burden across feasible trajectories. No model training, outcome-based fitting, or patient-specific adverse-event prediction was performed. Operator-used skin entry sites were identified on procedural CT images and compared with system-derived feasible entry locations under identical geometric assumptions. Concordance was assessed using Euclidean distance and patient-position agreement.

Diagnostic samples were obtained in 169 of 199 procedures (84.9%). Any adverse event occurred in 125 procedures (62.8%), including clinically relevant adverse events in 25 (12.6%). The median Euclidean distance between operator-used skin entry sites and the nearest tract among the 10 lowest-burden system-derived trajectories was 15.2 mm (IQR, 6.34-30.85 mm; range, 0.3-181.4 mm). Positioning agreement was moderate (κ = 0.534; 95% CI, 0.432-0.632; p < .001).

A deterministic tract-planning framework generated anatomically feasible entry locations spatially concordant with operator-selected skin entry sites. These findings support reproducible visualization of feasible access alternatives within a case-specific decision space for planning review, case review, and operator training.