Marker-less Body Surface Registration with 3D imaging for Percutaneous Intervention with Smartphone Augmented Reality in Phantoms and in vivo.
Authors
Affiliations (3)
Affiliations (3)
- Center for Interventional Oncology, National Institutes of Health Clinical Center, Bethesda, MD, USA; Henri Mondor's Institute of Biomedical Research - Inserm, U955 Team N°18, Creteil, France. Electronic address: [email protected].
- Center for Interventional Oncology, National Institutes of Health Clinical Center, Bethesda, MD, USA; Philips Research North America, Cambridge, MA, USA.
- Center for Interventional Oncology, National Institutes of Health Clinical Center, Bethesda, MD, USA.
Abstract
To evaluate the accuracy of automatic surface tracking registration with a smartphone augmented reality (AR) guidance system for percutaneous needle insertion in phantoms and in vivo. An AR application for needle guidance was developed using smartphone platform with an integrated needle guide. Automatic registration using body surface tracking based on deep learning obviated the need for fiducials with no additional sensors or hardware. Multiplanar CT images were volumetrically rendered to enable direct overlay on the body without segmentation. Accuracy was assessed on an abdominal phantom with eight operators of varying experience. An in vivo study was conducted in three swine (N=15 targets), where embolization coils implanted in liver, kidney and muscle served as targets. Needle tip-to-target distance and angular error were measured on post-procedural CT. In phantom experiments, the median accuracy was 4.8 mm (IQR 3.3-7.7mm) with a median angular error of 2.2° (IQR 1.4-4.3°). In vivo, the mean accuracy was 8.9±4.3mm and the mean angular error was 4.2±2.2°. Accuracy varied by organ (p=0.012), with best results in muscle (5.1mm), followed by kidney (8.7mm) and liver (13.0mm), associated with the degree of respiratory motion. Surface tracking compensated for external body shift but not internal organ motion. Smartphone AR with automatic surface tracking enabled guidance for needle insertion in vivo without fiducial markers or manual segmentation. The technology is feasible for simplifying or supplementing AR workflows and may support portable image guidance for minimally invasive needle-based therapies in low-resource settings.