EVD Surgical Guidance with Retro-Reflective Tool Tracking and Spatial Reconstruction using Head-Mounted Augmented Reality Device

TL;DR

The paper tackles accurate EVD guidance by correcting ToF depth sensor errors in AR-HMDs to non-invasively register pre-operative head images with the patient. It introduces a patient-specific depth distortion model and uses retro-reflective tools for robust head tracking, enabling dense surface reconstruction via TSDF and accurate intraoperative AR guidance. Experimental results show the depth-correction reduces errors by over $85\%$, achieving sub-millimetre surface reconstruction (e.g., $0.56$–$0.79\ \mathrm{mm}$) and a simulated-use accuracy of $2.09 \pm 0.16\ \mathrm{mm}$ entry point and $2.97 \pm 0.91^{\circ}$ orientation. The framework demonstrates high efficiency, non-invasiveness, and promising applicability to broader AR-guided neurosurgical tasks.

Abstract

Augmented Reality (AR) has been used to facilitate surgical guidance during External Ventricular Drain (EVD) surgery, reducing the risks of misplacement in manual operations. During this procedure, the key challenge is accurately estimating the spatial relationship between pre-operative images and actual patient anatomy in AR environment. This research proposes a novel framework utilizing Time of Flight (ToF) depth sensors integrated in commercially available AR Head Mounted Devices (HMD) for precise EVD surgical guidance. As previous studies have proven depth errors for ToF sensors, we first assessed their properties on AR-HMDs. Subsequently, a depth error model and patient-specific parameter identification method are introduced for accurate surface information. A tracking pipeline combining retro-reflective markers and point clouds is then proposed for accurate head tracking. The head surface is reconstructed using depth data for spatial registration, avoiding fixing tracking targets rigidly on the patient's skull. Firstly, $7.580\pm 1.488 mm$ depth value error was revealed on human skin, indicating the significance of depth correction. Our results showed that the error was reduced by over $85\%$ using proposed depth correction method on head phantoms in different materials. Meanwhile, the head surface reconstructed with corrected depth data achieved sub-millimetre accuracy. An experiment on sheep head revealed $0.79 mm$ reconstruction error. Furthermore, a user study was conducted for the performance in simulated EVD surgery, where five surgeons performed nine k-wire injections on a head phantom with virtual guidance. Results of this study revealed $2.09 \pm 0.16 mm$ translational accuracy and $2.97\pm 0.91$ degree orientational accuracy.

Figures (11)

• Figure 1: (a) The setup for accurate EVD surgical guidance utilized in the proposed framework. The retro-reflective tools are fixed on the headframe for tracking. (b) The sensor depth information is utilized to reconstruct the head surface and register pre-operative image with retro-reflective tool. (c) Retro-reflective tool tracking is used to provide visual guidance for drainage tube insertion. Here, we utilized k-wire injection to simulate EVD surgery.
• Figure 2: The proposed pipeline enabling accurate head tracking composites of three parts: 1) Multiple frames of AHAT sensor data are used to obtain patient-specific depth distortion parameters. 2) The undistorted point clouds are used to reconstruct the head surface and to register the pre-operative image with the retro-reflective tool rigidly fixed on the headframe. 3) The pose of the retro-reflective tool is extracted from dynamic sensor data for in-situ tracking.
• Figure 3: Testing structure to compare depth value provided by solving PnP and ToF sensor.
• Figure 4: Depth errors of different materials under HoloLens 2 AHAT camera. (a) The testing structure holds all the materials and retro-reflective markers at the same plane. (b), (c) Depth errors of different materials at different angles and depths, PC-ABS is excluded here due to high depth instability and data missing rate.
• Figure 5: AHAT sensor depth error model and patient-specific parameter identification method. (a) HoloLens 2 AHAT sensor depth value distortion model. (b) Depth value correction procedure during patient-specific parameter identification. Multiple retro-reflective markers are stuck on the patient's head during this procedure. (c) Evaluate the correctness of the corrected surface using the pre-operative image. (d) Evaluate the smoothness of the corrected surface near each marker.