Introduction
An innovative, patient-specific, hip implant was developed for canine patients suffering from developmental hip dysplasia. In a first cohort study, the post-operative positions of the implants deviated with an average of 5 mm when compared to the pre-operatively planned positions. A novel augmented reality (AR) based intra-operative navigation method (Holoma) was proposed to decrease implant positioning errors to a maximum of 3 mm. Holoma provides ArUco marker-based, mixed reality surgical navigation on the Hololens 2. The goal of this study was to quantify the error associated with each technical component in Holoma, to determine its navigational accuracies.

Methods
Holoma’s technical errors were quantified in a laboratory setting within a surgical operation room. The accuracy of ArUco marker localization and the amount of jittering (falsely detected marker movement) were tested by performing stationary and dynamic experiments. The precision of three image-to-patient registration methods (marker pose, point-to-point, point cloud) was determined by calculating the euclidean distance between multiple registration attempts. Lastly, implant navigation accuracy was quantified by positioning an implant on a precisely predetermined position and recording Holoma’s navigation instructions.

Results
Holoma’s ArUco marker localization accuracy had a median error of -0.97 mm (min; max: -2.04; -0.27) with a marker-to-camera distance of 60 cm, and median jittering was 0.65 mm (min; max: 0; 2.07). Median image-to-patient registration precision was 2.20 mm (min; max: 0.39; 6.00), 3.52 mm (0.21; 6.92), and 3.79 mm (0.55; 8.83), for marker pose, point-to-point, and point cloud registration, respectively. Lastly, implant navigation instructions showed median errors of 2 mm (min; max: 0; 7), 3 mm (0; 7), and 4 mm (0; 9) after marker pose, point-to-point, and point cloud registration, respectively.

Discussion and conclusion
The experiments suggest an accumulation of technical errors in Holoma’s navigation application. The accuracy and precision of the navigation software were determined to be inadequate to decrease patient specific hip implant placement errors, at the time of investigating. Future improvements within Holoma will determine its feasibility for accurate implant navigation. Other navigation techniques should be considered in the search for optimal implant positioning guidance.

Introduction
Augmented Reality (AR) and Mixed Reality (MR) can be used for surgical navigation to execute the pre-operative plan. HoloMA is a novel AR/MR application which can guide the user to place surgical instruments on the planned location within the patient. In this pilot study, the AR/MR-guidance of HoloMA was used to place personalized canine acetabular roof implants on the pre-planned location on the iliac bone.

Methods
Dedicated tools to perform the AR/MR patient registration and surgical guidance were developed. An in silico patient registration test was conducted to assess if the available bony surface during the acetabular roof surgery was suitable to perform the patient registration accurately. Pilot tests to place implants using the AR/MR-guidance of HoloMA were conducted on phantoms, a cadaver and in an in vivo dog patient. The translational and angulation error between the planned and the post-operative implant positions were determined. The aim was to achieve implant placement with a maximum translational error of 4.0 mm and a maximum angulation error of 5.0° relative to the pre-operative plan.

Results
The in silico patient registration test demonstrated a mean translational error of 0.94 ± 0.23 mm and a mean angulation error of 2.49 ± 0.34°. In the phantom tests, implants (n=6) were placed with a mean translational error of 1.94 ± 0.79 mm. The mean angulation errors in this test were: 2.87 ± 1.81° (transversal plane), 1.72 ± 1.64° (dorsal plane) and 3.10 ± 2.52° (sagittal plane). Two of the implants of the phantom test and both implants of the cadaveric test (n=2) were positioned with a translational error exceeding 4.0 mm and/or angulation error exceeding 5.0° from the planned positions. No implants were placed using AR/MR-guidance in the in vivo dog patient test.

Conclusion
The results of the in silico patient registration test hold promise for the use of AR/MR-guidance in positioning personalized acetabular roof implants. However, the moderate outcomes observed in the phantom and cadaveric test suggest the need for further testing and improvements before deploying this AR/MR technology in a clinical setting.