Dual-camera Augmented Reality-Based Navigation System for Precision CT-guided Biopsy and Ablation

To develop and characterize a CT-mounted dual-camera navigation system for augmented reality (AR) in CT-guided needle procedures.

Two orthogonal cameras were mounted on the CT gantry. The position of an anthropomorphic liver phantom relative to the gantry was registered to the cameras. Phantom targets and anatomy were preprocedurally segmented and overlaid on a monitor. A needle was angled and manipulated to the targets with feedback based upon the axial and sagittal camera perspectives. In one iteration, an automatic needle segmentation with vector extension helped define virtual needle trajectories. Four interventional radiologists each performed 18 needle placements in three cohorts (n = 6 each): 1. Dual-camera-based navigation ("Cam-Nav"). 2. Dual-camera-based navigation plus automatic needle trajectory definition ("Path-Nav"), and 3. CT-guided "freehand". Needle placement accuracy (closest distance from needle to target center), angle error and placement time (from planning to final needle position) were measured.

Median needle placement accuracy using Cam-Nav, Path-Nav, and freehand was 5.3 mm (interquartile range 3.8-8.3), 5.1 mm (2.3-6.1), and 16.7 mm (7.7-26.6). Angle errors were 3.3 degrees (2.3-4.5), 2.6 degrees (2.3-3.8), and 9.7 degrees (6.1-15.3). Path-Nav demonstrated no difference in accuracy and angle error compared to Cam-Nav (p = 0.35, p = 0.43) but median errors were improved. Both camera techniques demonstrated better accuracy and angle errors (all p < 0.001) and showed shorter placement times (32 s [21-45], 36 s [26-60], p = 0.29) compared to freehand (86 s [65-114], both p < 0.001).

CT-mounted dual-camera navigation showed improved needle placement accuracy over freehand and may present a reliable, cost-effective, and rapid navigation option for percutaneous needle placements. No level of evidence, experimental study.