Abstract
PURPOSE: Neurosurgical biopsies require high accuracy and precision and are executed with image-guided surgical navigation. The current state-of-the-art techniques require markers, are displayed on a 2D screen, and have a time-consuming setup. We propose an AR-driven surgical navigation method that automatically projects a 3D virtual overlay onto a patient in real-time, without the use of any markers. METHOD: Baseline accuracy of the proposed system and the StealthStation S8 was measured on a 3D printed human head phantom in a lab-based setting. For the measurements in the operating room, seventeen participants who underwent a neurosurgical biopsy with the StealthStation S8 were included. Prior to the clinical procedure, our proposed markerless AR system provided an automated three-dimensional virtual overlay onto the patient to the surgeon. By measuring the difference in the planned biopsy trajectory between the state-of-the-art StealthStation S8 and our experimental system, a comparison was made between the two systems. RESULTS: The average clinical error for the entry point of the proposed system was 4.5 ± 2.2 mm, which is lower than the total error of the current clinical gold standard found in literature. CONCLUSION: The total error of the system proposed in this study reaches the gold standard for image-guided neuronavigation, in both lab-controlled and clinical settings. These initial results highlight the potential and advantages of AR over other methods, offering promising AR opportunities for future clinical applications.