EZ-AR: A universal AR application for multi-model visualization and real-time tracking in healthcare settings

EZ-AR: A universal AR application for multi-model visualization and real-time tracking in healthcare settings | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article EZ-AR: A universal AR application for multi-model visualization and real-time tracking in healthcare settings Alicia Pose-Díez-de-la-Lastra, Gemma Arce-Alonso, José-Antonio Calvo-Haro, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7545942/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Augmented Reality (AR) offers significant potential to enhance surgical precision, improve clinical training, and support intraoperative decision-making. However, developing personalized AR applications for head-mounted displays such as Microsoft HoloLens 2 typically requires advanced technical expertise. Additionally, identifying a robust and accessible method for registering virtual models to the real world remains an open challenge. These limitations hinder the widespread adoption of AR in healthcare, particularly among non-technical clinical users. This work presents EZ-AR, a user-friendly AR application for Microsoft HoloLens 2 that automates complex tasks such as configuration and deployment, while ensuring low latency, high reliability, and smooth interaction. Ultimately, the system enables non-expert users to independently load and visualize personalized 3D models anchored to physical reference markers. The application accesses files from Google Drive in seconds and displays them in the physical environment using one of two integrated tracking methods: Vuforia or QR code detection. Once configured, the application no longer requires Internet access. The system's usability was assessed in terms of detection range and AR projection accuracy under varying surface inclinations and viewing angles. Results show a mean projection error of 3.3 ± 2.2 mm with QR code tracking and 3.9 ± 2.7 mm with Vuforia, both measured at a working distance compatible with typical surgical ergonomics. Further analysis revealed that accuracy is more influenced by viewing angle and surface inclination than by the tracking method itself. Overall, EZ-AR provides a robust and accessible AR framework to democratize AR adoption in clinical environments. Augmented reality Microsoft HoloLens 2 marker-based tracking surgical planning computer-assisted interventions. Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction Augmented Reality (AR) is revolutionizing medicine at an unprecedented rate by enabling users to visualize and interact with 3D digital content overlaid on the real world [1]. In clinical settings, AR has been successfully applied to medical education [2], simulation [3], clinical care [4], and intraoperative guidance [5], often demonstrating improved reproducibility, spatial awareness, and ergonomics compared to traditional methods. One of the major advantages of AR in surgery is its ability to display relevant 3D virtual information superimposed onto the patient, allowing the clinician to stay focused on the subject while enhancing spatial understanding [6], [7]. Head-mounted displays (HMDs) are the most widely used AR devices in surgical practice due to their hands-free interaction and enhanced visualization capabilities. Despite the rapid technological developments in recent years, the Microsoft HoloLens 2 (Redmond, WA, USA) remains the most commonly adopted platform in the clinical field. Indeed, among all commercially available systems, it is the one that has undergone the most rigorous validation process [8]. Since its first launch in 2019, it has been featured in over 220 clinical publications[1], with a steadily increasing number of studies supporting its use. As a result, HoloLens 2 is now widely regarded as the best commercially available head-mounted display for surgical applications [9]. Despite its potential, accurately registering virtual models to the patient's real anatomy remains one of the key technical challenges in AR-assisted surgery. Various approaches have been proposed, including manual alignment [10], [11], [12], SLAM-based anchoring [13], [14], and external tracking systems [15], [16] (commonly referred to as outside-in tracking). While these methods can achieve high precision, they are often limited by drift, complexity, the need for additional hardware, or a lack of reproducibility in dynamic clinical environments. Moreover, current implementations often rely on outdated or closed-source code, require external hardware, or offer limited documentation [17], [18], [19]. As a lightweight and accessible alternative, visual markers combined with recognition software and calibrated cameras have been widely used to establish robust spatial registration (marker-based, inside-out tracking techniques). Numerous software solutions for AR image marker tracking are already accessible in the market, including ARToolKit, ArUco, Vuforia, and QR code detection. These technologies have been extensively explored in a variety of clinical applications, such as assisting robotic laparoscopic surgeries [20], visualizing virtual models during thoracotomy procedures [21], guiding complex reconstructive surgeries [22], and supporting ultrasound training [23]. Despite the efficiency, robustness, and accuracy offered by these ready-to-use tracking alternatives, developing personalized AR applications typically requires advanced programming skills and lacks the support of flexible, user-friendly frameworks. As a result, these barriers hinder the development and broader adoption of this transformative technology, limiting its full potential in clinical and research settings. To address these challenges, this work pursues two main objectives: To develop a lightweight, intuitive, and fully customizable AR application for Microsoft HoloLens 2 that enables non-expert users to independently select, configure, and visualize custom 3D models anchored to a physical reference marker. The application should not require programming skills and rely on a cloud-based workflow for model storage and retrieval. To implement and evaluate two alternative marker-based tracking methods, comparing the AR projection accuracy achieved with each. Among the commercially available options, we selected Vuforia due to its widespread use in clinical applications, and QR code detection, owing to its native integration into the HoloLens 2 system. By removing technical barriers, this tool is intended to support the broader adoption of AR in routine clinical workflows and to democratize the use of HoloLens 2 for personalized surgical planning and intraoperative visualization. Moreover, while most previous studies report task-specific tracking errors, we aim to provide a robust, reproducible, and generalizable assessment of projection accuracy using a custom-designed validation platform. To characterize the intrinsic performance of each tracking method, we examine the influence of surface inclination, viewing angle, and distance to the AR marker. [1] This quantification was obtained using the Citation Report tool in Web of Science (https://www.webofscience.com/wos/woscc/basic-search), by performing a keyword search with “HoloLens 2 ” and “medicine” in June 2025. 2. Materials and Methods The following subsections provide a detailed explanation of the methodology employed in this work. 2.1 offers an overview of the system architecture. 2.2 describes the functionalities of the 3D Slicer module created to support user-personalized initialization. 2.3 introduces EZ-AR , the developed universal AR application for HoloLens 2. 2.4 outlines the AR reference markers and tracking approaches implemented. Finally, 2.5 details the technical and functional evaluation of the system. All components presented in this work, including the AR application, the 3D Slicer module, and the 3D models used for system validation, are freely available in a public GitHub repository at https://github.com/BSEL-UC3M/EZ-AR. 2.1. System overview Our proposal comprises two clearly differentiated components. The primary one is the Microsoft HoloLens 2 augmented reality headset, which runs a generic AR application called EZ-AR , specifically designed to operate natively on this device. Upon launch, the application presents a user interface that enables the download and visualization of up to eight personalized, not predefined, virtual 3D models. The second key component is a custom-developed module within the 3D Slicer platform [24]. This module provides an intuitive interface for selecting and configuring the desired 3D models for visualization in the headset. The outputs generated by the module must be uploaded to a Google Drive folder, from which they are retrieved by the headset to be displayed in the physical environment. Thanks to EZ- AR, users without prior experience in AR development can display their own 3D models in Microsoft HoloLens 2. A video tutorial showing how to utilize the system step by step is provided in SuppVideo-1_SetupInstructions . 2.2. 3D Slicer module The 3D Slicer module ( EZ-AR setup ) can be downloaded from the GitHub repository: https://github.com/BSEL-UC3M/EZ-AR. The module is organized into four main sections: Initialization, Positioning, Save Models, and Customize Model Info . Further details regarding each section and user instructions are provided in the SuppDocument_EZ-AR_InstructionsAndDetails . This module guides users through a sequential workflow: first, the tracking method to be used in the AR application (either Vuforia or QR tracking) is selected. Then, the operator chooses up to eight virtual 3D models from the local storage of their PC. This limitation of eight 3D models was established to avoid exceeding the computational capabilities of the Microsoft HoloLens 2 headset. Using a simple slider-based interface, users can define the models' pose relative to the AR reference marker. Finally, they assign a name and display color to each model for visualization in the Microsoft HoloLens 2 application. The output of the 3D Slicer module is a configuration file containing all these settings. This configuration file, along with the selected 3D models, must be uploaded to a Google Drive folder with access permissions set to “Anyone with the link”. The Microsoft HoloLens 2 headset subsequently accesses this folder to semi-automatically retrieve, import, and render the virtual content. 2.3. Augmented reality application The application was developed using Unity LTS 2022.3.20f1, the C# programming language, and Visual Studio 2022. The application is ready for download and use in the GitHub repository mentioned above. Detailed information about the AR application's functionality and user interface is provided in SuppDocument_EZ-AR_InstructionsAndDetails . 2.4. AR reference markers and tracking This work proposes the use of physical AR reference markers as a means of spatially aligning real and virtual environments. The Microsoft HoloLens 2 detects the pose of these markers in space, enabling the projection of user-selected 3D virtual models anchored to specific physical locations. Ultimately, a patient-specific surgical guide could be designed to incorporate a physical support for the AR marker, thereby permitting the automatic registration of virtual anatomical structures to their counterparts in the patient [22], [25]. A demonstration video has been recorded to showcase sample use cases of EZ-AR. These are available at SuppVideo-2_SampleUseCases. The models used in the video recording were extracted from the supplementary data of [26], available at https://zenodo.org/records/10958624; and from patient 1.3.6.1.4.1.9328.50.4.0001 in [27], available at https://github.com/MIRACLE-Center/CTPelvic1K. Two alternative tracking methods were implemented and made available within the application: Vuforia and QR code detection. To facilitate understanding and visual distinction, two clearly differentiated AR reference markers were employed (Figure 1). In our implementation, users can choose either of the two tracking methods. Here we provide a detailed comparison of both tracking alternatives, focusing on their projection accuracy and ease of use. Regardless of the chosen alternative, the AR app displays a virtual frame surrounding the AR marker to visually verify the alignment between the virtual and real worlds. Since QR codes can vary in dimensions, the frame size is dynamically adjusted to match the detected marker. Additionally, the frame transitions from green to red when tracking is lost. All 3D models employed in the validation experiments, including the AR reference markers, can be found in the GitHub repository at https://github.com/BSEL-UC3M/EZ-AR. 2.4.1. Vuforia SDK Vuforia software development kit (SDK) (Parametric Technology Corporation Inc., Boston, MA, USA) uses computer vision techniques to identify and track predefined image targets and basic 3D objects in real time [28], [29]. Vuforia consistently ranks at the top of most “Top AR” rankings, mainly due to its ease of implementation and broad compatibility across multiple devices, without requiring manual calibration of camera parameters [30]. In the clinical context, numerous research groups have employed Vuforia-based tracking on Microsoft HoloLens 2 for a variety of purposes, including needle insertion guidance [31], instrument alignment [32], surgical suturing assistance [33], delineation of virtual cutting planes [34], and pre-radiotherapy posture correction [35]. In this work, the Vuforia SDK (version 10.25.4) was integrated into Unity following the official setup guidelines [36]. At runtime, Vuforia compares these extracted features to those captured by the HoloLens 2 front-facing RGB camera and calculates the pose of the tracked object. Vuforia requires prior definition of the marker through the Vuforia developer portal, which generates a Unity-compatible package containing the marker’s features and size. The AR marker employed in this work consisted of a custom black-and-white square (55 × 55 mm) rated with the highest quality score by the Vuforia platform (Figure 1A). Despite its widespread adoption, Vuforia has its limitations. It functions as a closed-source "black box," meaning that users have no access to its internal processes or image processing pipelines and thus cannot customize or fine-tune the tracking algorithms. Furthermore, as a proprietary platform developed and maintained by an external company, long-term compatibility and continued support across devices cannot be guaranteed, posing a potential risk for critical clinical applications. 2.4.2. QR tracking Microsoft HoloLens 2 includes native software that utilizes the device's stereo cameras to detect QR codes in the scene. This open-source QR code detection system can be integrated and accessed in Unity with a plugin [37]. Developers can freely access the detection parameters and extract the 3D world coordinates associated with the QR code, allowing it to be an AR reference marker to which virtual 3D models can be anchored. Examples of QR code tracking in clinical environments include skill and clinical assessment [38], [39]. The QR code tracking method overcomes the limitations previously described for Vuforia. Since it is developed and maintained by Microsoft, long-term compatibility within the HoloLens ecosystem is more likely to be ensured. Moreover, unlike Vuforia, this method does not require prior definition of the marker’s visual content or size. Instead, the system detects any valid QR code in real-time and establishes a coordinate system at its location. This can be particularly convenient in a universal AR application designed for deployment across varied healthcare environments. Nevertheless, QR code detection presents its own constraints: according to the official documentation [37], only versions 1–10 are supported, a minimum size of 50 × 50 mm is required, and markers must be flat, logo-free, and viewed within an in-plane orientation of ±45°. Detection performance is highly sensitive to lighting conditions and contrast, and a quiet zone of at least four modules surrounding the code is mandatory. These requirements, along with the chosen QR version, directly influence the physical size of the marker and its detection range. Although lower versions allow for longer working distances due to larger quiet zones, they may be suboptimal in surgical environments where minimizing the physical footprint of the marker is critical. To balance usability and performance, a version 3 QR code measuring 55 mm per side was selected (Figure 1B). This size enabled a reliable working distance of over 40 cm between the headset and the reference marker, appropriate for use in operating rooms, considering the average length of a human forearm [40]. 2.5. System evaluation The system was evaluated based on three main factors: EZ-AR performance, detection range, and AR projection accuracy. 2.5.1. EZ-AR performance To analyze the performance of the proposed application, we measured the time required to load and display virtual 3D models on the HoloLens 2 when varying numbers of models and model file sizes are requested. These measurements were carried out both during the initial download from Google Drive and during subsequent loads from the headset’s local storage. Each trial was performed three times. 2.5.2. Detection range The versatility of the two proposed tracking methods, Vuforia and QR code detection, was also evaluated. Specifically, we analyzed the minimum and maximum distances between the AR marker and the headset required for reliable tracking as a function of the marker’s physical size. These metrics are essential to ensure compatibility with clinical environments, where workspace is limited and marker dimensions must remain unobtrusive while still providing a comfortable and reliable detection range for surgical use. To evaluate the detection range of each tracking method, eight versions of each marker type (Vuforia and QR code), ranging from 3 to 10 cm per side in 1 cm increments, were printed on paper. For each situation, the AR application was launched, and the marker was first placed beyond the detection range. Then, the marker was gradually moved closer until it was recognized, annotating the minimum detection distance. In a separate trial, the marker was positioned within detectable range and slowly moved away until tracking was lost or became unstable, establishing the maximum detection distance. Each measurement was repeated three times for every marker size and tracking method. 2.5.3. AR projection accuracy Finally, we analyzed how the AR projection accuracy is affected by the surface orientation on which the virtual model is projected. To this end, we designed a custom evaluation platform consisting of multiple inclined planes at known angles: 90°, 75°, 45°, 30°, 15°, and 0°, with varying orientations. In addition to the inclined planes, the platform features 12 conical holes (Ø 4 mm × 3 mm depth) distributed across its surface for point-based registration (Figure 2). The platform was 3D printed in basic PLA using an Ender 3 3D printer (Creality 3D, Shenzhen, China). In this procedure, an OptiTrack V230 Duo motion capture system (OptiTrack, Oregon, USA), connected to 3D Slicer via PLUS and OpenIGTLink [37], served as the gold standard for all measurements. This system simultaneously tracked the retroreflective spheres attached to a handheld pointer and those mounted on the validation platform through a rigid body. The rigid body in the validation platform was secured using a custom 3D-printed PLA adapter manufactured with the same Ender 3 printer. Additionally, one physical copy of each type of AR reference marker was 3D printed in black and white using a Bambu Lab P1S printer with the AMS unit. Each marker was produced as a 55 mm square. Eight users with varying levels of AR experience (three experts, two intermediates, and three novices) followed the same evaluation procedure. Each user launched the AR application, selected one of the two tracking methods (depending on the AR reference marker being evaluated), and downloaded two virtual 3D models to the HoloLens 2 local storage. One model represented the physical validation platform, and the other consisted of fourteen spheres (Ø 3 mm) virtually placed on its surfaces—three on the 90° and 0° planes, and two on each inclined surface (red spheres in Figure 2). Once the marker was detected, users stood at a fixed distance of 53 cm, aligned at eye level. This distance was selected based on the optimal detection range determined in the previous experiment, taking into account the size of the AR markers and the average length of a human forearm [40]. Without changing head position or viewing angle, each user sequentially collected the perceived positions of all virtual spheres as projected on the validation platform using the tracked pointer The recorded positions were later compared with their known reference coordinates to compute the projection deviation (AR projection accuracy), following the same procedure as in previous works [25], [41], [42]. Each volunteer repeated the procedure twice for both markers, with the order randomized across participants. Although Microsoft officially states that QR codes remain detectable even when tilted up to ±45° [37], we observed that tracking was lost at relatively small angles. Consequently, QR-based measurements were only recorded from a perpendicular (90°) perspective to the marker. In contrast, Vuforia tracking remained stable under inclined viewpoints, allowing two additional trials per user, tilting the platform 45°. During the analysis, Mann-Whitney U tests were conducted to identify statistically significant differences between AR tracking methods and between repetitions. In addition, Kruskal-Wallis tests were used to assess whether user variability or plane inclination resulted in significantly different outcomes. When significant differences were detected, post hoc Conover tests were performed to determine which specific groups were most dissimilar. 3. Results 3.1. EZ-AR performance Upon clicking the Load Models button in the AR application's control panel, all models specified in the configuration file were sequentially loaded into the virtual environment. Table 1 summarizes the average time required to load and display the virtual models, whether retrieved from Google Drive or the headset’s local storage. Models were downloaded from Google Drive using a stable Wi-Fi connection provided by a standard home router with 1 Gb fiber-optic internet service. The measurements cover the entire process, starting when the Load Models button was pressed, and finishing when the last model was fully rendered with its designated color and position. Table 1. Time required to load and display virtual models from either Google Drive or local storage, based on the number of models and the total file size. Number of models Total size of all models (KB) Loading time from Google Drive, in seconds (mean ± std) Loading time from local storage, in seconds (mean ± std) 1 59,161 24.7 ± 2.0 10.5 ± 0.3 1 2,724 4.3 ± 0.2 0.9 ± 0.1 1 395 3.3 ± 0.2 0.5 ± 0.1 3 7,633 10.9 ± 0.7 1.7 ± 0.1 3 24,720 14.1 ± 1.0 2.9 ± 0.0 6 63,826 31.9 ± 1.3 8.9 ± 0.2 8 11,835 23.0 ± 1.3 2.6 ± 0.2 3.2. Detection range Figure 3 illustrates the detection range for each AR tracking method. In both cases, the distance to the marker was considered insufficient or excessive when tracking was completely lost or became too unstable to be considered reliable. All measurements were taken with the user looking at the marker perpendicularly. Vuforia exhibited a wider detection range, and the system could track the marker even when the platform was tilted up to 45°, as long as it remained within the valid detection distance. On the other hand, QR code tracking was only reliable when the marker was viewed perpendicularly, and any deviation from this angle resulted in immediate loss of tracking. Nevertheless, under ideal conditions (i.e., within the detection range and from a perpendicular perspective), QR code tracking provided a greater sense of stability in projecting virtual 3D models, whereas minor jitter was consistently observed with Vuforia. 3.3. AR projection accuracy All eight users completed the experiments without difficulty, regardless of their prior experience with AR technology. A Kruskal-Wallis test revealed no statistically significant differences among users. Similarly, a Mann-Whitney U test showed no statistically significant differences between repeated trials by the same user. However, statistically significant differences were found based on the orientation of the surfaces onto which the spheres were projected. Interestingly, a Conover test with Holm correction revealed no statistically significant overall differences in projection accuracy across the three experimental conditions: QR tracking (perpendicular view), Vuforia tracking (perpendicular view), and Vuforia tracking (tilted view). A Conover test was performed to identify which platform inclination planes exhibited the most significant differences between them. The results revealed two main clusters. No statistically significant differences were found among the surfaces within each cluster, but clear differences emerged when comparing surfaces across the two groups. The first group included the planes inclined at 90°, 75°, and 45°, while the second cluster comprised the 30°, 15°, and 0° planes. Figure 4 presents a density heatmap showing the distribution of all projection errors grouped by surface orientation and tracking method used from each perspective. Data collected using Vuforia (tilted view) over the 30° inclined surface was systematically excluded, as that portion of the platform was not visible from the user’s point of view due to its steep angle. Additionally, no data could be collected from a tilted view using QR code tracking, as the system became unstable whenever the marker was not viewed perpendicularly. In [43], the authors conducted a similar experiment, analyzing axis-specific errors across different tracking methods, distances, and viewing angles. Table 2 compares their reported error values with those obtained in this study under the angle and distance conditions that most closely match our setup. In their case, rather than measuring the projection error of virtual points, the authors assessed the tracking error associated with marker detection, which was always performed with the marker oriented perpendicularly to the headset. Accordingly, only the results from this work corresponding to projections on the vertical surface of the platform (90°) are included in the comparison. Table 2. Mean Euclidean distance and absolute error considering targets oriented at 90°, per axis. Data is structured according to the AR tracking method employed and the tracking perspective, and compared to the values provided in [43]. Tracking method Source Angle to AR marker, in ° Distance to AR marker, in mm Euclidean distance, in mm (mean ± std) Abs error in R, in mm (mean ± std) Abs error in S, in mm (mean ± std) QR [43] 90 300 - 3.1 ± 0.2 0.8 ± 0.2 QR This work 90 530 2.5 ± 1.4 1.7 ± 1.4 1.5 ± 1.2 QR [43] 90 600 - 0.9 ± 0.1 2.5 ± 0.1 Vuforia [43] 90 300 - 3.1 ± 0.1 1.3 ± 0.1 Vuforia This work 90 530 2.6 ± 1.3 1.9 ± 1.2 1.4 ± 1.2 Vuforia [43] 90 600 - 2.1 ± 0.0 0.9 ± 0.1 Vuforia [43] 30 300 - 6.3 ± 0.1 1.8 ± 0.1 Vuforia This work 45 530 2.7 ± 1.6 1.9 ± 1.4 1.6 ± 1.3 Vuforia [43] 30 600 - 4.7 ± 0.1 1.8 ± 0.1 Discussion In this work, we have developed EZ-AR , a user-friendly, fully customizable AR application for Microsoft HoloLens 2, readily available for download and use at https://github.com/BSEL-UC3M/EZ-AR. Our primary objective was to equip non-expert users, such as clinicians, with the ability to display their personalized 3D virtual models in the headset without requiring any software development skills. In this case, they simply need to follow the instructions provided in the 3D Slicer module to adjust and upload the desired structures to Google Drive. The AR application will semi-automatically access them and display them integrated into the real world. Setup instructions are provided in the form of a video tutorial SuppVideo-1_SetupInstructions . The application offers a high level of versatility with respect to the virtual 3D models. This includes tailoring the visibility and opacity of the models, as well as translating, rotating, and scaling them with intuitive hand gestures. Moreover, the 3D models can be displayed in the real world following a tracked AR reference marker. Ultimately, clinicians could incorporate those references into the surgical site (i.e. attached to a surgical guide) as a link between virtual and real worlds to facilitate automatic registration between the anatomical structures of the patient and their virtual counterparts [22], [25]. The application could be used for virtual planning and definition of cutting planes or drilling trajectories, to enhance understanding of patient-specific anatomy in three dimensions, and to provide robust and accurate guidance during surgical procedures. Demonstration examples illustrating these use cases are provided in SuppVideo-2_SampleUseCases . The way to download the selected models is through a configuration file generated using the presented 3D Slicer module. The first time the headset recognizes a new configuration file, the models are downloaded from Google Drive and stored in the local memory of the Microsoft HoloLens 2. In subsequent sessions, they can be retrieved directly from the device’s cache. This explains why Table 1 shows loading times consistently longer when models are retrieved from Google Drive compared to loading from local storage. As expected, increasing both the number and total size of the files progressively augments the time required to render all models in the virtual environment. Notably, downloading a single large model from Google Drive resulted in a loading time comparable to eight smaller models. In all cases, however, the loading process was completed within seconds. Various marker-based tracking methods compatible with Microsoft HoloLens 2 have been proposed in the literature for many clinical applications. From the available options, we incorporated two in the AR app: Vuforia, due to its widespread adoption in clinical environments and ease of implementation; and QR tracking, which offers native support in Microsoft HoloLens 2. The decision regarding which tracking method to use is left to the user, providing maximum flexibility based on available resources and individual capabilities. To support this decision, this work presents an objective comparison of both tracking alternatives in terms of detection distance ranges as a function of marker size, as well as AR projection accuracy. The latter was measured using a custom-designed validation platform specifically developed to assess projection accuracy along each anatomical axis, depending on the inclination of the surface onto which the virtual model is projected. The optimal detection distance was analyzed for several AR reference marker sizes, ranging from 3 cm to 10 cm per side in 1 cm increments. The results showed that Vuforia offers a wider detection range and allows users greater freedom to move around the AR reference marker. However, it requires prior definition of the marker’s appearance and size, a valid license for use, and does not guarantee long-term support, as it is proprietary software. In contrast, QR code tracking has a narrower detection range and requires the user to hold the marker perpendicular to their line of sight. Nevertheless, it requires no prior definition of the AR marker, and it provides greater projection stability within its detection range. Moreover, since it is developed and maintained by Microsoft, long-term compatibility within the HoloLens ecosystem is more likely to be ensured. In both cases, reliable detection was achieved within a range of 40 to 55 cm between the headset and the marker when using markers sized between 40 and 55 mm. This range represents a practical compromise: the marker remains small enough to be used in surgical environments without obstructing the procedure, while still allowing for a comfortable working distance within arm’s reach of the headset wearer. The results in terms of AR projection accuracy revealed no statistically significant differences between the two proposed tracking methods. The mean projection error obtained with QR code tracking was 3.3 ± 2.2 mm, while that of Vuforia was 3.9 ± 2.7 mm. These values were recorded under favorable lighting conditions, with the user viewing the markers at a distance of 53 cm, perpendicular to the markers. For Vuforia, additional measurements were also collected with the platform (and therefore, the AR marker) tilted 45°, yielding a mean error of 4.2 ± 3.1 mm. The QR code algorithm failed to detect the marker at oblique angles due to the stereo camera’s limited field of view and lower resolution compared to the RGB camera used by Vuforia. All our results are comparable to the 3.1 mm RMSE reported in [42], which evaluated AR projection accuracy using a patient-based phantom with Vuforia as the tracking method. Regarding axis-specific error, the results obtained using Vuforia and QR tracking were also comparable. As expected, the projection error in surfaces at 90° produced a circular distribution in the density maps, indicating that deviations along the L–R and I–S axes were similar. In contrast, the distributions at 75° and 45° appeared horizontally elongated, reflecting lower vertical error along the I–S axis. This is consistent with the geometry of inclined planes, where projection errors along the vertical axis are smaller than the absolute deviations measured on the sloped surface. At 30°, 15°, and 0°, however, the error in the L–R axis was no longer comparable to that in the P–A axis. This confirms that depth perception remains a primary limitation in AR projection accuracy when using head-mounted displays like the HoloLens 2 [42], [43]. Table 2 compares the axis-specific projection errors obtained in this study with those reported in [43]. In our case, the axis-wise error ranged from 1.4 ± 1.2 mm for Vuforia (perpendicular view, 90°), to 1.9 ± 1.4 mm for Vuforia (tilted view, 45°). In contrast, the results reported in [43] showed much greater variability, ranging from 0.8 ± 0.2 mm with QR tracking (perpendicular view, 90°) and a distance of 300 mm, to as high as 6.3 ± 0.1 mm for Vuforia (tilted view, 30°), also at 300 mm. Despite this variability, our results fall within the same order of magnitude, supporting the reliability and achievable accuracy of both tracking methods. To conclude, it is worth noting that this application may be compatible not only with Microsoft HoloLens 2, but also with other MRTK-based headsets—such as Magic Leap 2—with only minor adjustments to the Unity project [44]. Given the compatibility of this alternative device with both Vuforia and QR code tracking [45], [46], the long-term usability of EZ-AR is ensured, regardless of the future of Microsoft HoloLens 2. Conclusions In this work, we presented EZ-AR, a lightweight, open-access AR framework designed to facilitate the deployment of personalized 3D content on Microsoft HoloLens 2 without requiring programming expertise. By combining a user-friendly module in 3D Slicer with a configurable AR application, the system enables clinical users to visualize custom virtual models anchored to physical reference markers using either Vuforia or QR code tracking. Our evaluation demonstrates that both tracking methods offer comparable AR projection accuracy with errors in the range of 3–4 mm under favorable conditions and from clinically realistic distances. While QR tracking offers greater versatility in marker design and requires no prior configuration, Vuforia provides higher flexibility in user movement around the AR reference marker. Each advantage may be better suited to different clinical scenarios, depending on the available resources, spatial constraints, and interaction requirements of the procedure. Our results validate the system’s usability, stability, and robustness, reinforcing its potential as a versatile tool to support visualization and planning tasks in healthcare environments. Declarations · Funding Research supported by projects TED2021-129392B-I00, TED2021-132200B-I00 and PID2023-149604OB-I00 (MCIU/AEI/10.13039/501100011033 and European Union “NextGenerationEU”/PRTR) and Comunidad de Madrid (MAGERIT-CM TEC-2024/COM-44 and Multiannual Agreement with UC3M SAFEDELIVERY-CM-UC3M). · Conflict of interest The authors declare no conflict of interest. · Ethics statement This study did not involve experimentation on human subjects, nor did it collect or analyze personal or medical data. The experiments were conducted using a 3D-printed validation platform and voluntary participation from individuals who interacted with a non-invasive AR system. As such, no formal ethics approval was required according to the regulations of our institution. All participants were informed of the nature of the study and gave verbal consent to participate. No identifiable personal information was recorded, and all privacy rights were respected throughout the study. · Authors' contributions Conception and design of the work: A. P.-D.-d.-l.-L., J. A. C.-H., R. P.-M., J. P.; Investigation, data curation, analysis, and interpretation: A. P.-D.-d.-l.-L., G. A.-A., J. P.; Software: A. P.-D.-d.-l.-L., G. A.-A.; Funding acquisition, supervision and project administration: J. A. C.-H., R. P.-M., J. P.; Writing - original draft preparation: A. P.-D.-d.-l.-L., G. A.-A; Writing – review and editing: J. P. All authors have read and agreed to the published version of the manuscript. They all have also agreed to be personally accountable for their own contributions and to ensure that questions related to the accuracy or integrity of any part of the work are appropriately investigated, resolved, and the resolution documented in the literature. · Authors' information o Alicia Pose Díez de la Lastra: [email protected] (A. P.-D.-d.-l.-L.) o Gemma Arce Alonso: [email protected] (G. A.-A) o José Antonio Calvo Haro: [email protected] (J. A. C.-H.) o Rubén Pérez Mañanes: [email protected] (R. P.-M.) o Javier Pascau: [email protected] (J. P.) References C. H. Araujo, J. A. Aguirre, and L. Puig, “3D Printing Applications and Extended Realities in Medicine: Systematic Review,” 3D Printing and Additive Manufacturing , Oct. 2024, doi: 10.1089/3dp.2024.0128. Z. M. C. Baum et al. , “Augmented Reality Training Platform for Neurosurgical Burr Hole Localization,” Journal of Medical Robotics Research , vol. 04, no. 03n04, pp. 1942001–1942001, Sep. 2019, doi: 10.1142/S2424905X19420017. D. Amiras et al. , “Augmented reality simulator for CT-guided interventions,” European Radiology , Jun. 2021, doi: 10.1007/s00330-021-08043-0. R. N. Uppot et al. , “Implementing virtual and augmented reality tools for radiology education and training, communication, and clinical care,” Radiology , vol. 291, no. 3, pp. 570–580, 2019, doi: 10.1148/radiol.2019182210. H. El-Hariri, P. Pandey, A. J. Hodgson, and R. Garbi, “Augmented reality visualisation for orthopaedic surgical guidance with pre- and intra-operative multimodal image data fusion,” Healthcare Technology Letters , vol. 5, no. 5, pp. 189–193, 2018, doi: 10.1049/htl.2018.5061. P. Pratt et al. , “Through the HoloLens TM looking glass: augmented reality for extremity reconstruction surgery using 3D vascular models with perforating vessels,” European Radiology Experimental , vol. 2, no. 1, pp. 0–6, 2018, doi: 10.1186/s41747-017-0033-2. J. Yoon et al. , “Augmented reality for the surgeon: Systematic review,” The international journal of medical robotics + computer assisted surgery : MRCAS , vol. 14, no. 4, Aug. 2018, doi: 10.1002/RCS.1914. M. Miller Koop et al. , “The Microsoft HoloLens 2 Provides Accurate Measures of Gait, Turning, and Functional Mobility in Healthy Adults,” Sensors , vol. 22, no. 5, Art. no. 5, Jan. 2022, doi: 10.3390/s22052009. A. Palumbo, “Microsoft HoloLens 2 in Medical and Healthcare Context: State of the Art and Future Prospects,” Sensors , vol. 22, no. 20, Art. no. 20, Jan. 2022, doi: 10.3390/s22207709. M. E. de Oliveira, H. G. Debarba, A. Lädermann, S. Chagué, and C. Charbonnier, “A hand-eye calibration method for augmented reality applied to computer-assisted orthopedic surgery,” The International Journal of Medical Robotics and Computer Assisted Surgery , vol. 15, no. 2, p. e1969, 2019, doi: 10.1002/rcs.1969. J. Hajek et al. , “Closing the Calibration Loop: An Inside-Out-Tracking Paradigm for Augmented Reality in Orthopedic Surgery,” presented at the Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), https://doi.org/10.1007/978-3-030-00937-3_35, 2018, pp. 299–306. doi: 10.1007/978-3-030-00937-3_35. E. Rae, A. Lasso, M. S. Holden, E. Morin, R. Levy, and G. Fichtinger, “Neurosurgical burr hole placement using the Microsoft HoloLens,” in Medical Imaging 2018: Image-Guided Procedures, Robotic Interventions, and Modeling , SPIE, Mar. 2018, pp. 190–197. doi: 10.1117/12.2293680. R. Vassallo, A. Rankin, E. C. S. Chen, and T. M. Peters, “Hologram stability evaluation for Microsoft HoloLens,” in Medical Imaging 2017: Image Perception, Observer Performance, and Technology Assessment , SPIE, Mar. 2017, pp. 295–300. doi: 10.1117/12.2255831. R. Tang et al. , “Augmented reality navigation in open surgery for hilar cholangiocarcinoma resection with hemihepatectomy using video-based in situ three-dimensional anatomical modeling: A case report,” Medicine , vol. 96, no. 37, Sep. 2017, doi: 10.1097/MD.0000000000008083. J. W. Meulstee et al. , “Toward Holographic-Guided Surgery,” Surgical Innovation , vol. 26, no. 1, pp. 86–94, Feb. 2019, doi: 10.1177/1553350618799552. I. Kuhlemann, M. Kleemann, P. Jauer, A. Schweikard, and F. Ernst, “Towards X-ray free endovascular interventions - Using HoloLens for on-line holographic visualisation,” presented at the Healthcare Technology Letters, Institution of Engineering and Technology, 2017, pp. 184–187. doi: 10.1049/htl.2017.0061. M. Akulauskas, K. Butkus, V. Rutkūnas, T. Blažauskas, and D. Jegelevičius, “Implementation of Augmented Reality in Dental Surgery Using HoloLens 2: An In Vitro Study and Accuracy Assessment,” Applied Sciences , vol. 13, no. 14, Art. no. 14, Jan. 2023, doi: 10.3390/app13148315. C. Kunz et al. , “Infrared marker tracking with the HoloLens for neurosurgical interventions,” Current Directions in Biomedical Engineering , vol. 6, no. 1, May 2020, doi: 10.1515/cdbme-2020-0027. D. Ungureanu et al. , “HoloLens 2 Research Mode as a Tool for Computer Vision Research,” Aug. 25, 2020, arXiv : arXiv:2008.11239. doi: 10.48550/arXiv.2008.11239. L. Qian, A. Deguet, and P. Kazanzides, “ARssist: augmented reality on a head-mounted display for the first assistant in robotic surgery,” Healthcare Technology Letters , vol. 5, no. 5, pp. 194–200, 2018, doi: 10.1049/htl.2018.5065. M. Doughty and N. R. Ghugre, “Head-Mounted Display-Based Augmented Reality for Image-Guided Media Delivery to the Heart: A Preliminary Investigation of Perceptual Accuracy,” Journal of Imaging , vol. 8, no. 2, Art. no. 2, Feb. 2022, doi: 10.3390/jimaging8020033. A. Pose-Díez-de-la-Lastra et al. , “Microsoft HoloLens 2 vs. tablet-based augmented reality and 3D printing for fronto-orbital reconstruction of craniosynostosis: a case study,” 3D Printing in Medicine , vol. 11, no. 1, p. 13, Mar. 2025, doi: 10.1186/s41205-025-00251-4. J. N. Costa et al. , “Ultrasound training simulator using augmented reality glasses: an accuracy and precision assessment study,” in 2022 44th Annual International Conference of the IEEE Engineering in Medicine & Biology Society (EMBC) , Jul. 2022, pp. 4461–4464. doi: 10.1109/EMBC48229.2022.9871406. S. Pieper, M. Halle, and R. Kikinis, “3D Slicer; 3D Slicer,” 2004 2nd IEEE International Symposium on Biomedical Imaging: Nano to Macro (IEEE Cat No. 04EX821) , 2004, doi: 10.1109/ISBI.2004.1398617. M. García‐Sevilla et al. , “Augmented reality as a tool to guide psi placement in pelvic tumor resections,” Sensors , vol. 21, no. 23, 2021, doi: 10.3390/s21237824. A. Díez-Montiel, A. Pose-Díez-de-la-Lastra, A. González-Álvarez, J. I. Salmerón, J. Pascau, and S. Ochandiano, “Tablet-based Augmented reality and 3D printed templates in fully guided Microtia Reconstruction: a clinical workflow,” 3D Printing in Medicine , vol. 10, no. 1, p. 17, May 2024, doi: 10.1186/s41205-024-00213-2. P. Liu et al. , “Deep learning to segment pelvic bones: large-scale CT datasets and baseline models,” Int J Comput Assist Radiol Surg , vol. 16, no. 5, pp. 749–756, May 2021, doi: 10.1007/s11548-021-02363-8. “Vuforia Enterprise Augmented Reality (AR) Software | PTC.” Accessed: May 03, 2023. [Online]. Available: https://www.ptc.com/en/products/vuforia X. Cong and T. Li, “Design and Development of Virtual Medical System Interface Based on VR-AR Hybrid Technology,” Computational and Mathematical Methods in Medicine , vol. 2020, p. e7108147, Aug. 2020, doi: 10.1155/2020/7108147. E. M. Bourhim and A. Akhiate, “Augmented Reality SDK’s: A Comparative Study,” in Intelligent Systems Design and Applications , A. Abraham, N. Gandhi, T. Hanne, T.-P. Hong, T. Nogueira Rios, and W. Ding, Eds., Cham: Springer International Publishing, 2022, pp. 559–566. doi: 10.1007/978-3-030-96308-8_52. B. J. Park, S. J. Hunt, G. J. Nadolski, and T. P. Gade, “Augmented reality improves procedural efficiency and reduces radiation dose for CT-guided lesion targeting: a phantom study using HoloLens 2,” Sci Rep , vol. 10, no. 1, p. 18620, Oct. 2020, doi: 10.1038/s41598-020-75676-4. M. Benmahdjoub, W. J. Niessen, E. B. Wolvius, and T. van Walsum, “Virtual extensions improve perception-based instrument alignment using optical see-through devices,” IEEE Transactions on Visualization and Computer Graphics , vol. 27, no. 11, pp. 4332–4341, Nov. 2021, doi: 10.1109/TVCG.2021.3106506. Y. Nagayo, T. Saito, and H. Oyama, “A Novel Suture Training System for Open Surgery Replicating Procedures Performed by Experts Using Augmented Reality,” J Med Syst , vol. 45, no. 5, p. 60, Apr. 2021, doi: 10.1007/s10916-021-01735-6. A. Thabit, M. Benmahdjoub, M.-L. C. van Veelen, W. J. Niessen, E. B. Wolvius, and T. van Walsum, “Augmented reality navigation for minimally invasive craniosynostosis surgery: a phantom study,” Int J CARS , vol. 17, no. 8, pp. 1453–1460, Aug. 2022, doi: 10.1007/s11548-022-02634-y. P. B. Johnson, A. Jackson, M. Saki, E. Feldman, and J. Bradley, “Patient posture correction and alignment using mixed reality visualization and the HoloLens 2,” Medical Physics , vol. 49, no. 1, pp. 15–22, 2022, doi: 10.1002/mp.15349. “Vuforia Package for Unity - Vuforia Engine Library.” Accessed: May 15, 2025. [Online]. Available: https://developer.vuforia.com/library/vuforia-engine/unity-extension/vuforia-engine-package-unity/ lolambean, “QR code tracking overview - Mixed Reality.” Accessed: Apr. 29, 2025. [Online]. Available: https://learn.microsoft.com/en-us/windows/mixed-reality/develop/advanced-concepts/qr-code-tracking-overview Y. Li, A. Reed, N. Kavoussi, and J. Y. Wu, “Eye gaze metrics for skill assessment and feedback in kidney stone surgery,” Int J CARS , vol. 18, no. 6, pp. 1127–1134, Jun. 2023, doi: 10.1007/s11548-023-02901-6. L. Cercenelli, N. Emiliani, C. Gulotta, M. Bevini, G. Badiali, and E. Marcelli, “Augmented Reality to Assist in the Diagnosis of Temporomandibular Joint Alterations,” Inf Syst Front , vol. 27, no. 1, pp. 33–49, Feb. 2025, doi: 10.1007/s10796-024-10545-3. S. Navid, T. Mokhtari, T. Alizamir, A. Arabkheradmand, and G. Hassanzadeh, “Determination of Stature from Upper Arm Length in Medical Students,” Anatomical Sciences Journal , vol. 11, no. 3, pp. 135–140, Aug. 2014. R. Moreta-Martinez, A. Pose-Díez-de-la-Lastra, J. A. Calvo-Haro, L. Mediavilla-Santos, R. Pérez-Mañanes, and J. Pascau, “Combining Augmented Reality and 3D Printing to Improve Surgical Workflows in Orthopedic Oncology: Smartphone Application and Clinical Evaluation,” Sensors 2021, Vol. 21, Page 1370 , vol. 21, no. 4, pp. 1370–1370, Feb. 2021, doi: 10.3390/S21041370. A. Pose-Díez-De-la-Lastra et al. , “HoloLens 1 vs. HoloLens 2: Improvements in the New Model for Orthopedic Oncological Interventions,” Sensors , vol. 22, no. 13, 2022, doi: 10.3390/s22134915. G. M. Costa, M. R. Petry, J. G. Martins, and A. P. G. M. Moreira, “Assessment of Multiple Fiducial Marker Trackers on Hololens 2,” IEEE Access , vol. 12, pp. 14211–14226, 2024, doi: 10.1109/ACCESS.2024.3356722. “HoloLens 2 to Magic Leap 2 Porting Guide | MagicLeap Developer Documentation.” Accessed: Jun. 27, 2025. [Online]. Available: https://developer-docs.magicleap.cloud/docs/guides/third-party/mrtk/hololens-porting/ “Magic Leap - Vuforia Engine Library.” Accessed: Jun. 27, 2025. [Online]. Available: https://developer.vuforia.com/library/vuforia-engine/getting-started/development-environments/getting-started-vuforia-engine-and-magic-leap-2/ “Marker Tracking | MagicLeap Developer Documentation.” Accessed: Jun. 27, 2025. [Online]. Available: https://developer-docs.magicleap.cloud/docs/guides/features/marker-tracking/ Additional Declarations No competing interests reported. Supplementary Files SuppVideo1SetupInstructions.mp4 SuppVideo2SampleUseCases.mov SuppDocumentEZARInstructionsAndDetails.docx Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7545942","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":563448743,"identity":"8a792de2-335d-48ca-8053-34cfde020c37","order_by":0,"name":"Alicia Pose-Díez-de-la-Lastra","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAyklEQVRIiWNgGAWjYNCCCgkwxfiBeC1nIFqYJYjWwdgGZfAQo5p/dvPhDx/nWeTJO/AYfpDMYZCTbyCgReLOsTTJmdskig0P8BhLFG5jMDY4QMiaGzlmzLzbJBI3NvCYMUhuY0jcQEiH/I0c489/50C18G5jqJ9PyGEGN3IMpBkbJBLnM0C0JDAQcpghyC89xyQSNzCzFUtLbpMw3EBIi9xtYIj9qKlLnN/evPHjx2028gRDjAEWewaHUbjEaCFs+CgYBaNgFIxYAABL2zvIzL1AcwAAAABJRU5ErkJggg==","orcid":"","institution":"Carlos III University of Madrid","correspondingAuthor":true,"prefix":"","firstName":"Alicia","middleName":"","lastName":"Pose-Díez-de-la-Lastra","suffix":""},{"id":563448744,"identity":"b317539d-e75d-416a-91c4-72c17a2c1915","order_by":1,"name":"Gemma Arce-Alonso","email":"","orcid":"","institution":"Carlos III University of Madrid","correspondingAuthor":false,"prefix":"","firstName":"Gemma","middleName":"","lastName":"Arce-Alonso","suffix":""},{"id":563448745,"identity":"8ee9cd46-dbf0-4529-9cfc-c6618646353e","order_by":2,"name":"José-Antonio Calvo-Haro","email":"","orcid":"","institution":"Instituto de Investigación Sanitaria Gregorio Marañón","correspondingAuthor":false,"prefix":"","firstName":"José-Antonio","middleName":"","lastName":"Calvo-Haro","suffix":""},{"id":563448747,"identity":"73245464-5195-48ef-ad41-0c577e437387","order_by":3,"name":"Rubén Pérez-Mañanes","email":"","orcid":"","institution":"Instituto de Investigación Sanitaria Gregorio Marañón","correspondingAuthor":false,"prefix":"","firstName":"Rubén","middleName":"","lastName":"Pérez-Mañanes","suffix":""},{"id":563448749,"identity":"8e95e920-82d5-49d7-adab-37bae8c83735","order_by":4,"name":"Javier Pascau","email":"","orcid":"","institution":"Universidad Carlos III de Madrid","correspondingAuthor":false,"prefix":"","firstName":"Javier","middleName":"","lastName":"Pascau","suffix":""}],"badges":[],"createdAt":"2025-09-05 16:23:19","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7545942/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7545942/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":99307652,"identity":"8fabe871-4572-47fb-afd8-132dd3e318d9","added_by":"auto","created_at":"2025-12-31 16:06:28","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":24689,"visible":true,"origin":"","legend":"\u003cp\u003eAR reference markers tracked by the EZ-AR app using (A) Vuforia and (B) QR tracking.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7545942/v1/cec4c15de71faed3ed813798.png"},{"id":98821187,"identity":"9ff3dbbd-e806-405e-93ba-2b3fd5a40d8f","added_by":"auto","created_at":"2025-12-22 17:20:51","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":36264,"visible":true,"origin":"","legend":"\u003cp\u003eValidation platform designed to assess AR projection accuracy. The 14 red spheres represent the virtual targets projected by the AR application onto the platform surfaces for validation.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-7545942/v1/80a1878ac16e7873c3489707.png"},{"id":98821189,"identity":"5efc5cf4-481c-4e68-9c66-190dcb30194b","added_by":"auto","created_at":"2025-12-22 17:20:51","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":42993,"visible":true,"origin":"","legend":"\u003cp\u003eDetection range by marker size and AR tracking option\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-7545942/v1/a1d926491fc93e063df0d843.png"},{"id":98821191,"identity":"42976c2c-e3d4-4e95-9ecf-feed19e174cc","added_by":"auto","created_at":"2025-12-22 17:20:51","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":316245,"visible":true,"origin":"","legend":"\u003cp\u003eDensity heatmap of the AR projection accuracy error obtained in each inclination plane of the validation platform, shown by anatomical axes for (A) QR tracking (perpendicular view), (B) Vuforia tracking (perpendicular view), and (C) Vuforia tracking (tilted view).\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-7545942/v1/c1272388acc12bfd3a8af66c.png"},{"id":100360169,"identity":"0124e3ff-028b-483a-bdea-a260ce41810f","added_by":"auto","created_at":"2026-01-16 07:37:47","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1174327,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7545942/v1/5da35a8f-a0fc-4365-b32b-322edcb0ed2c.pdf"},{"id":98821193,"identity":"dd166335-59d0-4fe3-b993-a5ae3b8971be","added_by":"auto","created_at":"2025-12-22 17:20:56","extension":"mp4","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":136376261,"visible":true,"origin":"","legend":"","description":"","filename":"SuppVideo1SetupInstructions.mp4","url":"https://assets-eu.researchsquare.com/files/rs-7545942/v1/3d019a1099a2af13d9b3269f.mp4"},{"id":98821194,"identity":"1e60bef3-1558-4f44-b2ee-7c1f39ceb634","added_by":"auto","created_at":"2025-12-22 17:20:59","extension":"mov","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":218053845,"visible":true,"origin":"","legend":"","description":"","filename":"SuppVideo2SampleUseCases.mov","url":"https://assets-eu.researchsquare.com/files/rs-7545942/v1/d51e26864589d63e8b2bcb30.mov"},{"id":98821192,"identity":"23cdb567-fcbc-42ac-b532-aeefa0b4a122","added_by":"auto","created_at":"2025-12-22 17:20:51","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":565857,"visible":true,"origin":"","legend":"","description":"","filename":"SuppDocumentEZARInstructionsAndDetails.docx","url":"https://assets-eu.researchsquare.com/files/rs-7545942/v1/9cc3eb9c87389d7b162b1ce1.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"EZ-AR: A universal AR application for multi-model visualization and real-time tracking in healthcare settings","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eAugmented Reality (AR) is revolutionizing medicine at an unprecedented rate by enabling users to visualize and interact with 3D digital content overlaid on the real world [1]. In clinical settings, AR has been successfully applied to medical education [2], simulation [3], clinical care [4], and intraoperative guidance [5], often demonstrating improved reproducibility, spatial awareness, and ergonomics compared to traditional methods. One of the major advantages of AR in surgery is its ability to display relevant 3D virtual information superimposed onto the patient, allowing the clinician to stay focused on the subject while enhancing spatial understanding [6], [7].\u003c/p\u003e\n\u003cp\u003eHead-mounted displays (HMDs) are the most widely used AR devices in surgical practice due to their hands-free interaction and enhanced visualization capabilities. Despite the rapid technological developments in recent years, the Microsoft HoloLens 2 (Redmond, WA, USA) remains the most commonly adopted platform in the clinical field. Indeed, among all commercially available systems, it is the one that has undergone the most rigorous validation process [8]. Since its first launch in 2019, it has been featured in over 220 clinical publications[1], with a steadily increasing number of studies supporting its use. As a result, HoloLens 2 is now widely regarded as the best commercially available head-mounted display for surgical applications [9].\u003c/p\u003e\n\u003cp\u003eDespite its potential, accurately registering virtual models to the patient's real anatomy remains one of the key technical challenges in AR-assisted surgery. Various approaches have been proposed, including manual alignment [10], [11], [12], SLAM-based anchoring [13], [14], and external tracking systems [15], [16] (commonly referred to as \u003cem\u003eoutside-in\u003c/em\u003e tracking). While these methods can achieve high precision, they are often limited by drift, complexity, the need for additional hardware, or a lack of reproducibility in dynamic clinical environments. Moreover, current implementations often rely on outdated or closed-source code, require external hardware, or offer limited documentation [17], [18], [19].\u003c/p\u003e\n\u003cp\u003eAs a lightweight and accessible alternative, visual markers combined with recognition software and calibrated cameras have been widely used to establish robust spatial registration (marker-based, \u003cem\u003einside-out \u003c/em\u003etracking techniques). Numerous software solutions for AR image marker tracking are already accessible in the market, including ARToolKit, ArUco, Vuforia, and QR code detection. These technologies have been extensively explored in a variety of clinical applications, such as assisting robotic laparoscopic surgeries [20], visualizing virtual models during thoracotomy procedures [21], guiding complex reconstructive surgeries [22], and supporting ultrasound training [23].\u003c/p\u003e\n\u003cp\u003eDespite the efficiency, robustness, and accuracy offered by these ready-to-use tracking alternatives, developing personalized AR applications typically requires advanced programming skills and lacks the support of flexible, user-friendly frameworks. As a result, these barriers hinder the development and broader adoption of this transformative technology, limiting its full potential in clinical and research settings. To address these challenges, this work pursues two main objectives:\u003c/p\u003e\n\u003col\u003e\n\u003cli\u003eTo develop a lightweight, intuitive, and fully customizable AR application for Microsoft HoloLens 2 that enables non-expert users to independently select, configure, and visualize custom 3D models anchored to a physical reference marker. The application should not require programming skills and rely on a cloud-based workflow for model storage and retrieval. \u003c/li\u003e\n\u003cli\u003eTo implement and evaluate two alternative marker-based tracking methods, comparing the AR projection accuracy achieved with each. Among the commercially available options, we selected Vuforia due to its widespread use in clinical applications, and QR code detection, owing to its native integration into the HoloLens 2 system. \u003c/li\u003e\n\u003c/ol\u003e\n\u003cp\u003eBy removing technical barriers, this tool is intended to support the broader adoption of AR in routine clinical workflows and to democratize the use of HoloLens 2 for personalized surgical planning and intraoperative visualization. Moreover, while most previous studies report task-specific tracking errors, we aim to provide a robust, reproducible, and generalizable assessment of projection accuracy using a custom-designed validation platform. To characterize the intrinsic performance of each tracking method, we examine the influence of surface inclination, viewing angle, and distance to the AR marker.\u003c/p\u003e\n\u003cp\u003e[1] This quantification was obtained using the Citation Report tool in Web of Science (https://www.webofscience.com/wos/woscc/basic-search), by performing a keyword search with “HoloLens 2\u003cem\u003e”\u003c/em\u003e and “medicine” in June 2025.\u003c/p\u003e"},{"header":"2.\tMaterials and Methods","content":"\u003cp\u003eThe following subsections provide a detailed explanation of the methodology employed in this work. \u003cstrong\u003e2.1\u003c/strong\u003e offers an overview of the system architecture. \u003cstrong\u003e2.2\u003c/strong\u003e describes the functionalities of the 3D Slicer module created to support user-personalized initialization. \u003cstrong\u003e2.3\u003c/strong\u003e introduces \u003cem\u003eEZ-AR\u003c/em\u003e, the developed universal AR application for HoloLens 2. \u003cstrong\u003e2.4\u003c/strong\u003e outlines the AR reference markers and tracking approaches implemented. Finally, \u003cstrong\u003e2.5\u003c/strong\u003e details the technical and functional evaluation of the system.\u003c/p\u003e\n\u003cp\u003eAll components presented in this work, including the AR application, the 3D Slicer module, and the 3D models used for system validation, are freely available in a public GitHub repository at https://github.com/BSEL-UC3M/EZ-AR.\u003c/p\u003e\n\u003ch2\u003e2.1. System overview\u003c/h2\u003e\n\u003cp\u003eOur proposal comprises two clearly differentiated components. The primary one is the Microsoft HoloLens 2 augmented reality headset, which runs a generic AR application called \u003cem\u003eEZ-AR\u003c/em\u003e, specifically designed to operate natively on this device. Upon launch, the application presents a user interface that enables the download and visualization of up to eight personalized, not predefined, virtual 3D models.\u003c/p\u003e\n\u003cp\u003eThe second key component is a custom-developed module within the 3D Slicer platform [24]. This module provides an intuitive interface for selecting and configuring the desired 3D models for visualization in the headset. The outputs generated by the module must be uploaded to a Google Drive folder, from which they are retrieved by the headset to be displayed in the physical environment.\u003c/p\u003e\n\u003cp\u003eThanks to \u003cem\u003eEZ-\u003c/em\u003eAR, users without prior experience in AR development can display their own 3D models in Microsoft HoloLens 2. A video tutorial showing how to utilize the system step by step is provided in \u003cem\u003eSuppVideo-1_SetupInstructions\u003c/em\u003e.\u003c/p\u003e\n\u003ch2\u003e2.2. 3D Slicer module\u003c/h2\u003e\n\u003cp\u003eThe 3D Slicer module (\u003cem\u003eEZ-AR setup\u003c/em\u003e) can be downloaded from the GitHub repository: https://github.com/BSEL-UC3M/EZ-AR. The module is organized into four main sections: \u003cem\u003eInitialization, Positioning, Save Models, \u003c/em\u003eand\u003cem\u003e Customize Model Info\u003c/em\u003e. Further details regarding each section and user instructions are provided in the \u003cem\u003eSuppDocument_EZ-AR_InstructionsAndDetails\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eThis module guides users through a sequential workflow: first, the tracking method to be used in the AR application (either Vuforia or QR tracking) is selected. Then, the operator chooses up to eight virtual 3D models from the local storage of their PC. This limitation of eight 3D models was established to avoid exceeding the computational capabilities of the Microsoft HoloLens 2 headset. Using a simple slider-based interface, users can define the models\u0026apos; pose relative to the AR reference marker. Finally, they assign a name and display color to each model for visualization in the Microsoft HoloLens 2 application.\u003c/p\u003e\n\u003cp\u003eThe output of the 3D Slicer module is a configuration file containing all these settings. This configuration file, along with the selected 3D models, must be uploaded to a Google Drive folder with access permissions set to \u0026ldquo;Anyone with the link\u0026rdquo;. The Microsoft HoloLens 2 headset subsequently accesses this folder to semi-automatically retrieve, import, and render the virtual content.\u003c/p\u003e\n\u003ch2\u003e2.3. Augmented reality application\u003c/h2\u003e\n\u003cp\u003eThe application was developed using Unity LTS 2022.3.20f1, the C# programming language, and Visual Studio 2022. The application is ready for download and use in the GitHub repository mentioned above. Detailed information about the AR application\u0026apos;s functionality and user interface is provided in \u003cem\u003eSuppDocument_EZ-AR_InstructionsAndDetails\u003c/em\u003e.\u003c/p\u003e\n\u003ch2\u003e2.4. AR reference markers and tracking\u003c/h2\u003e\n\u003cp\u003eThis work proposes the use of physical AR reference markers as a means of spatially aligning real and virtual environments. The Microsoft HoloLens 2 detects the pose of these markers in space, enabling the projection of user-selected 3D virtual models anchored to specific physical locations. Ultimately, a patient-specific surgical guide could be designed to incorporate a physical support for the AR marker, thereby permitting the automatic registration of virtual anatomical structures to their counterparts in the patient [22], [25]. A demonstration video has been recorded to showcase sample use cases of EZ-AR. These are available at \u003cem\u003eSuppVideo-2_SampleUseCases.\u003c/em\u003e The models used in the video recording were extracted from the supplementary data of [26], available at https://zenodo.org/records/10958624; and from patient 1.3.6.1.4.1.9328.50.4.0001 in [27], available at https://github.com/MIRACLE-Center/CTPelvic1K. Two alternative tracking methods were implemented and made available within the application: Vuforia and QR code detection. To facilitate understanding and visual distinction, two clearly differentiated AR reference markers were employed (Figure 1). \u003c/p\u003e\n\u003cp\u003eIn our implementation, users can choose either of the two tracking methods. Here we provide a detailed comparison of both tracking alternatives, focusing on their projection accuracy and ease of use. Regardless of the chosen alternative, the AR app displays a virtual frame surrounding the AR marker to visually verify the alignment between the virtual and real worlds. Since QR codes can vary in dimensions, the frame size is dynamically adjusted to match the detected marker. Additionally, the frame transitions from green to red when tracking is lost. All 3D models employed in the validation experiments, including the AR reference markers, can be found in the GitHub repository at https://github.com/BSEL-UC3M/EZ-AR.\u003c/p\u003e\n\u003ch3\u003e2.4.1. Vuforia SDK\u003c/h3\u003e\n\u003cp\u003eVuforia software development kit (SDK) (Parametric Technology Corporation Inc., Boston, MA, USA) uses computer vision techniques to identify and track predefined image targets and basic 3D objects in real time [28], [29]. Vuforia consistently ranks at the top of most \u0026ldquo;Top AR\u0026rdquo; rankings, mainly due to its ease of implementation and broad compatibility across multiple devices, without requiring manual calibration of camera parameters [30]. In the clinical context, numerous research groups have employed Vuforia-based tracking on Microsoft HoloLens 2 for a variety of purposes, including needle insertion guidance [31], instrument alignment [32], surgical suturing assistance [33], delineation of virtual cutting planes [34], and pre-radiotherapy posture correction [35].\u003c/p\u003e\n\u003cp\u003eIn this work, the Vuforia SDK (version 10.25.4) was integrated into Unity following the official setup guidelines [36]. At runtime, Vuforia compares these extracted features to those captured by the HoloLens 2 front-facing RGB camera and calculates the pose of the tracked object. Vuforia requires prior definition of the marker through the Vuforia developer portal, which generates a Unity-compatible package containing the marker\u0026rsquo;s features and size. The AR marker employed in this work consisted of a custom black-and-white square (55 \u0026times; 55 mm) rated with the highest quality score by the Vuforia platform (Figure 1A).\u003c/p\u003e\n\u003cp\u003eDespite its widespread adoption, Vuforia has its limitations. It functions as a closed-source \u0026quot;black box,\u0026quot; meaning that users have no access to its internal processes or image processing pipelines and thus cannot customize or fine-tune the tracking algorithms. Furthermore, as a proprietary platform developed and maintained by an external company, long-term compatibility and continued support across devices cannot be guaranteed, posing a potential risk for critical clinical applications.\u003c/p\u003e\n\u003ch3\u003e2.4.2. QR tracking\u003c/h3\u003e\n\u003cp\u003eMicrosoft HoloLens 2 includes native software that utilizes the device\u0026apos;s stereo cameras to detect QR codes in the scene. This open-source QR code detection system can be integrated and accessed in Unity with a plugin [37]. Developers can freely access the detection parameters and extract the 3D world coordinates associated with the QR code, allowing it to be an AR reference marker to which virtual 3D models can be anchored. Examples of QR code tracking in clinical environments include skill and clinical assessment [38], [39].\u003c/p\u003e\n\u003cp\u003eThe QR code tracking method overcomes the limitations previously described for Vuforia. Since it is developed and maintained by Microsoft, long-term compatibility within the HoloLens ecosystem is more likely to be ensured. Moreover, unlike Vuforia, this method does not require prior definition of the marker\u0026rsquo;s visual content or size. Instead, the system detects any valid QR code in real-time and establishes a coordinate system at its location. This can be particularly convenient in a universal AR application designed for deployment across varied healthcare environments.\u003c/p\u003e\n\u003cp\u003eNevertheless, QR code detection presents its own constraints: according to the official documentation [37], only versions 1\u0026ndash;10 are supported, a minimum size of 50 \u0026times; 50 mm is required, and markers must be flat, logo-free, and viewed within an in-plane orientation of \u0026plusmn;45\u0026deg;. Detection performance is highly sensitive to lighting conditions and contrast, and a quiet zone of at least four modules surrounding the code is mandatory. These requirements, along with the chosen QR version, directly influence the physical size of the marker and its detection range. Although lower versions allow for longer working distances due to larger quiet zones, they may be suboptimal in surgical environments where minimizing the physical footprint of the marker is critical. To balance usability and performance, a version 3 QR code measuring 55 mm per side was selected (Figure 1B). This size enabled a reliable working distance of over 40 cm between the headset and the reference marker, appropriate for use in operating rooms, considering the average length of a human forearm [40].\u003c/p\u003e\n\u003ch2\u003e2.5. System evaluation\u003c/h2\u003e\n\u003cp\u003eThe system was evaluated based on three main factors: EZ-AR performance, detection range, and AR projection accuracy.\u003c/p\u003e\n\u003ch3\u003e2.5.1. EZ-AR performance\u003c/h3\u003e\n\u003cp\u003eTo analyze the performance of the proposed application, we measured the time required to load and display virtual 3D models on the HoloLens 2 when varying numbers of models and model file sizes are requested. These measurements were carried out both during the initial download from Google Drive and during subsequent loads from the headset\u0026rsquo;s local storage. Each trial was performed three times.\u003c/p\u003e\n\u003ch3\u003e2.5.2. Detection range\u003c/h3\u003e\n\u003cp\u003eThe versatility of the two proposed tracking methods, Vuforia and QR code detection, was also evaluated. Specifically, we analyzed the minimum and maximum distances between the AR marker and the headset required for reliable tracking as a function of the marker\u0026rsquo;s physical size. These metrics are essential to ensure compatibility with clinical environments, where workspace is limited and marker dimensions must remain unobtrusive while still providing a comfortable and reliable detection range for surgical use.\u003c/p\u003e\n\u003cp\u003eTo evaluate the detection range of each tracking method, eight versions of each marker type (Vuforia and QR code), ranging from 3 to 10 cm per side in 1 cm increments, were printed on paper. For each situation, the AR application was launched, and the marker was first placed beyond the detection range. Then, the marker was gradually moved closer until it was recognized, annotating the minimum detection distance. In a separate trial, the marker was positioned within detectable range and slowly moved away until tracking was lost or became unstable, establishing the maximum detection distance. Each measurement was repeated three times for every marker size and tracking method.\u003c/p\u003e\n\u003ch3\u003e2.5.3. AR projection accuracy\u003c/h3\u003e\n\u003cp\u003eFinally, we analyzed how the AR projection accuracy is affected by the surface orientation on which the virtual model is projected. To this end, we designed a custom evaluation platform consisting of multiple inclined planes at known angles: 90\u0026deg;, 75\u0026deg;, 45\u0026deg;, 30\u0026deg;, 15\u0026deg;, and 0\u0026deg;, with varying orientations. In addition to the inclined planes, the platform features 12 conical holes (\u0026Oslash; 4 mm \u0026times; 3 mm depth) distributed across its surface for point-based registration (Figure 2). The platform was 3D printed in basic PLA using an Ender 3 3D printer (Creality 3D, Shenzhen, China).\u003c/p\u003e\n\u003cp\u003eIn this procedure, an OptiTrack V230 Duo motion capture system (OptiTrack, Oregon, USA), connected to 3D Slicer via PLUS and OpenIGTLink [37], served as the gold standard for all measurements. This system simultaneously tracked the retroreflective spheres attached to a handheld pointer and those mounted on the validation platform through a rigid body. The rigid body in the validation platform was secured using a custom 3D-printed PLA adapter manufactured with the same Ender 3 printer. Additionally, one physical copy of each type of AR reference marker was 3D printed in black and white using a Bambu Lab P1S printer with the AMS unit. Each marker was produced as a 55 mm square.\u003c/p\u003e\n\u003cp\u003eEight users with varying levels of AR experience (three experts, two intermediates, and three novices) followed the same evaluation procedure. Each user launched the AR application, selected one of the two tracking methods (depending on the AR reference marker being evaluated), and downloaded two virtual 3D models to the HoloLens 2 local storage. One model represented the physical validation platform, and the other consisted of fourteen spheres (\u0026Oslash; 3 mm) virtually placed on its surfaces\u0026mdash;three on the 90\u0026deg; and 0\u0026deg; planes, and two on each inclined surface (red spheres in Figure 2). Once the marker was detected, users stood at a fixed distance of 53 cm, aligned at eye level. This distance was selected based on the optimal detection range determined in the previous experiment, taking into account the size of the AR markers and the average length of a human forearm [40]. Without changing head position or viewing angle, each user sequentially collected the perceived positions of all virtual spheres as projected on the validation platform using the tracked pointer\u003c/p\u003e\n\u003cp\u003eThe recorded positions were later compared with their known reference coordinates to compute the projection deviation (AR projection accuracy), following the same procedure as in previous works [25], [41], [42]. Each volunteer repeated the procedure twice for both markers, with the order randomized across participants. Although Microsoft officially states that QR codes remain detectable even when tilted up to \u0026plusmn;45\u0026deg; [37], we observed that tracking was lost at relatively small angles. Consequently, QR-based measurements were only recorded from a perpendicular (90\u0026deg;) perspective to the marker. In contrast, Vuforia tracking remained stable under inclined viewpoints, allowing two additional trials per user, tilting the platform 45\u0026deg;.\u003c/p\u003e\n\u003cp\u003eDuring the analysis, Mann-Whitney U tests were conducted to identify statistically significant differences between AR tracking methods and between repetitions. In addition, Kruskal-Wallis tests were used to assess whether user variability or plane inclination resulted in significantly different outcomes. When significant differences were detected, post hoc Conover tests were performed to determine which specific groups were most dissimilar.\u003c/p\u003e"},{"header":"3.\tResults","content":"\u003ch2\u003e3.1.\u0026nbsp;EZ-AR performance\u003c/h2\u003e\n\u003cp\u003eUpon clicking the \u003cem\u003eLoad Models\u003c/em\u003e button in the AR application\u0026apos;s control panel, all models specified in the configuration file were sequentially loaded into the virtual environment.\u003c/p\u003e\n\u003cp\u003eTable 1 summarizes the average time required to load and display the virtual models, whether retrieved from Google Drive or the headset\u0026rsquo;s local storage. Models were downloaded from Google Drive using a stable Wi-Fi connection provided by a standard home router with 1 Gb fiber-optic internet service. The measurements cover the entire process, starting when the \u003cem\u003eLoad Models\u003c/em\u003e button was pressed, and finishing when the last model was fully rendered with its designated color and position.\u003c/p\u003e\n\u003cp\u003eTable 1. Time required to load and display virtual models from either Google Drive or local storage, based on the number of models and the total file size.\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003eNumber of models\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003eTotal size of all models\u003c/p\u003e\n \u003cp\u003e(KB)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 198px;\"\u003e\n \u003cp\u003eLoading time from Google Drive, in seconds\u003c/p\u003e\n \u003cp\u003e(mean \u0026plusmn; std)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 214px;\"\u003e\n \u003cp\u003eLoading time from local storage, in seconds\u003c/p\u003e\n \u003cp\u003e(mean \u0026plusmn; std)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e59,161\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 198px;\"\u003e\n \u003cp\u003e24.7 \u0026plusmn; 2.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 214px;\"\u003e\n \u003cp\u003e10.5 \u0026plusmn; 0.3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e2,724\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 198px;\"\u003e\n \u003cp\u003e4.3 \u0026plusmn; 0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 214px;\"\u003e\n \u003cp\u003e0.9 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e395\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 198px;\"\u003e\n \u003cp\u003e3.3 \u0026plusmn; 0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 214px;\"\u003e\n \u003cp\u003e0.5 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e7,633\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 198px;\"\u003e\n \u003cp\u003e10.9 \u0026plusmn; 0.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 214px;\"\u003e\n \u003cp\u003e1.7 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e24,720\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 198px;\"\u003e\n \u003cp\u003e14.1 \u0026plusmn; 1.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 214px;\"\u003e\n \u003cp\u003e2.9 \u0026plusmn; 0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e63,826\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 198px;\"\u003e\n \u003cp\u003e31.9 \u0026plusmn; 1.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 214px;\"\u003e\n \u003cp\u003e8.9 \u0026plusmn; 0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 85px;\"\u003e\n \u003cp\u003e8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 104px;\"\u003e\n \u003cp\u003e11,835\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 198px;\"\u003e\n \u003cp\u003e23.0 \u0026plusmn; 1.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 214px;\"\u003e\n \u003cp\u003e2.6 \u0026plusmn; 0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003ch2\u003e3.2.\u0026nbsp;Detection range\u003c/h2\u003e\n\u003cp\u003eFigure 3 illustrates the detection range for each AR tracking method. In both cases, the distance to the marker was considered insufficient or excessive when tracking was completely lost or became too unstable to be considered reliable. All measurements were taken with the user looking at the marker perpendicularly.\u003c/p\u003e\n\u003cp\u003eVuforia exhibited a wider detection range, and the system could track the marker even when the platform was tilted up to 45\u0026deg;, as long as it remained within the valid detection distance. On the other hand, QR code tracking was only reliable when the marker was viewed perpendicularly, and any deviation from this angle resulted in immediate loss of tracking. Nevertheless, under ideal conditions (i.e., within the detection range and from a perpendicular perspective), QR code tracking provided a greater sense of stability in projecting virtual 3D models, whereas minor jitter was consistently observed with Vuforia.\u003c/p\u003e\n\u003ch2\u003e3.3.\u0026nbsp;AR projection accuracy\u003c/h2\u003e\n\u003cp\u003eAll eight users completed the experiments without difficulty, regardless of their prior experience with AR technology. A Kruskal-Wallis test revealed no statistically significant differences among users. Similarly, a Mann-Whitney U test showed no statistically significant differences between repeated trials by the same user. However, statistically significant differences were found based on the orientation of the surfaces onto which the spheres were projected. Interestingly, a Conover test with Holm correction revealed no statistically significant overall differences in projection accuracy across the three experimental conditions: QR tracking (perpendicular view), Vuforia tracking (perpendicular view), and Vuforia tracking (tilted view).\u003c/p\u003e\n\u003cp\u003eA Conover test was performed to identify which platform inclination planes exhibited the most significant differences between them. The results revealed two main clusters. No statistically significant differences were found among the surfaces within each cluster, but clear differences emerged when comparing surfaces across the two groups. The first group included the planes inclined at 90\u0026deg;, 75\u0026deg;, and 45\u0026deg;, while the second cluster comprised the 30\u0026deg;, 15\u0026deg;, and 0\u0026deg; planes. Figure 4 presents a density heatmap showing the distribution of all projection errors grouped by surface orientation and tracking method used from each perspective. Data collected using Vuforia (tilted view) over the 30\u0026deg; inclined surface was systematically excluded, as that portion of the platform was not visible from the user\u0026rsquo;s point of view due to its steep angle. Additionally, no data could be collected from a tilted view using QR code tracking, as the system became unstable whenever the marker was not viewed perpendicularly.\u003c/p\u003e\n\u003cp\u003eIn [43], the authors conducted a similar experiment, analyzing axis-specific errors across different tracking methods, distances, and viewing angles. Table 2 compares their reported error values with those obtained in this study under the angle and distance conditions that most closely match our setup. In their case, rather than measuring the projection error of virtual points, the authors assessed the tracking error associated with marker detection, which was always performed with the marker oriented perpendicularly to the headset. Accordingly, only the results from this work corresponding to projections on the vertical surface of the platform (90\u0026deg;) are included in the comparison.\u003c/p\u003e\n\u003cp\u003eTable 2. Mean Euclidean distance and absolute error considering targets oriented at 90\u0026deg;, per axis. Data is structured according to the AR tracking method employed and the tracking perspective, and compared to the values provided in [43].\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eTracking method\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eSource\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eAngle to AR marker, in \u0026deg;\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eDistance to AR marker, in mm\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eEuclidean distance, in mm\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(mean \u0026plusmn; std)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eAbs error in R, in mm\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(mean \u0026plusmn; std)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eAbs error in S, in mm\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003e(mean \u0026plusmn; std)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eQR\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e[43]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e300\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3.1 \u0026plusmn; 0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.8 \u0026plusmn; 0.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eQR\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eThis work\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e530\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2.5 \u0026plusmn; 1.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.7 \u0026plusmn; 1.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.5 \u0026plusmn; 1.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eQR\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e[43]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e600\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.9 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2.5 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eVuforia\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e[43]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e300\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e3.1 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.3 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eVuforia\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eThis work\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e530\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2.6 \u0026plusmn; 1.3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.9 \u0026plusmn; 1.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.4 \u0026plusmn; 1.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eVuforia\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e[43]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e90\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e600\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2.1 \u0026plusmn; 0.0\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e0.9 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eVuforia\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e[43]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e300\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e6.3 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.8 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eVuforia\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003eThis work\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e45\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e530\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e2.7 \u0026plusmn; 1.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.9 \u0026plusmn; 1.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.6 \u0026plusmn; 1.3\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd\u003e\n \u003cp\u003e\u003cstrong\u003eVuforia\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e[43]\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e30\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e600\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e4.7 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd\u003e\n \u003cp\u003e1.8 \u0026plusmn; 0.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eIn this work, we have developed \u003cem\u003eEZ-AR\u003c/em\u003e, a user-friendly, fully customizable AR application for Microsoft HoloLens 2, readily available for download and use at https://github.com/BSEL-UC3M/EZ-AR. Our primary objective was to equip non-expert users, such as clinicians, with the ability to display their personalized 3D virtual models in the headset without requiring any software development skills. In this case, they simply need to follow the instructions provided in the 3D Slicer module to adjust and upload the desired structures to Google Drive. The AR application will semi-automatically access them and display them integrated into the real world. Setup instructions are provided in the form of a video tutorial \u003cem\u003eSuppVideo-1_SetupInstructions\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eThe application offers a high level of versatility with respect to the virtual 3D models. This includes tailoring the visibility and opacity of the models, as well as translating, rotating, and scaling them with intuitive hand gestures. Moreover, the 3D models can be displayed in the real world following a tracked AR reference marker. Ultimately, clinicians could incorporate those references into the surgical site (i.e. attached to a surgical guide) as a link between virtual and real worlds to facilitate automatic registration between the anatomical structures of the patient and their virtual counterparts [22], [25]. The application could be used for virtual planning and definition of cutting planes or drilling trajectories, to enhance understanding of patient-specific anatomy in three dimensions, and to provide robust and accurate guidance during surgical procedures. Demonstration examples illustrating these use cases are provided in \u003cem\u003eSuppVideo-2_SampleUseCases\u003c/em\u003e.\u003c/p\u003e\n\u003cp\u003eThe way to download the selected models is through a configuration file generated using the presented 3D Slicer module. The first time the headset recognizes a new configuration file, the models are downloaded from Google Drive and stored in the local memory of the Microsoft HoloLens 2. In subsequent sessions, they can be retrieved directly from the device’s cache. This explains why Table 1 shows loading times consistently longer when models are retrieved from Google Drive compared to loading from local storage. As expected, increasing both the number and total size of the files progressively augments the time required to render all models in the virtual environment. Notably, downloading a single large model from Google Drive resulted in a loading time comparable to eight smaller models. In all cases, however, the loading process was completed within seconds.\u003c/p\u003e\n\u003cp\u003eVarious marker-based tracking methods compatible with Microsoft HoloLens 2 have been proposed in the literature for many clinical applications. From the available options, we incorporated two in the AR app: Vuforia, due to its widespread adoption in clinical environments and ease of implementation; and QR tracking, which offers native support in Microsoft HoloLens 2. The decision regarding which tracking method to use is left to the user, providing maximum flexibility based on available resources and individual capabilities. To support this decision, this work presents an objective comparison of both tracking alternatives in terms of detection distance ranges as a function of marker size, as well as AR projection accuracy. The latter was measured using a custom-designed validation platform specifically developed to assess projection accuracy along each anatomical axis, depending on the inclination of the surface onto which the virtual model is projected.\u003c/p\u003e\n\u003cp\u003eThe optimal detection distance was analyzed for several AR reference marker sizes, ranging from 3 cm to 10 cm per side in 1 cm increments. The results showed that Vuforia offers a wider detection range and allows users greater freedom to move around the AR reference marker. However, it requires prior definition of the marker’s appearance and size, a valid license for use, and does not guarantee long-term support, as it is proprietary software. In contrast, QR code tracking has a narrower detection range and requires the user to hold the marker perpendicular to their line of sight. Nevertheless, it requires no prior definition of the AR marker, and it provides greater projection stability within its detection range. Moreover, since it is developed and maintained by Microsoft, long-term compatibility within the HoloLens ecosystem is more likely to be ensured.\u003c/p\u003e\n\u003cp\u003eIn both cases, reliable detection was achieved within a range of 40 to 55 cm between the headset and the marker when using markers sized between 40 and 55 mm. This range represents a practical compromise: the marker remains small enough to be used in surgical environments without obstructing the procedure, while still allowing for a comfortable working distance within arm’s reach of the headset wearer.\u003c/p\u003e\n\u003cp\u003eThe results in terms of AR projection accuracy revealed no statistically significant differences between the two proposed tracking methods. The mean projection error obtained with QR code tracking was 3.3 ± 2.2 mm, while that of Vuforia was 3.9 ± 2.7 mm. These values were recorded under favorable lighting conditions, with the user viewing the markers at a distance of 53 cm, perpendicular to the markers. For Vuforia, additional measurements were also collected with the platform (and therefore, the AR marker) tilted 45°, yielding a mean error of 4.2 ± 3.1 mm. The QR code algorithm failed to detect the marker at oblique angles due to the stereo camera’s limited field of view and lower resolution compared to the RGB camera used by Vuforia. All our results are comparable to the 3.1 mm RMSE reported in [42], which evaluated AR projection accuracy using a patient-based phantom with Vuforia as the tracking method.\u003c/p\u003e\n\u003cp\u003eRegarding axis-specific error, the results obtained using Vuforia and QR tracking were also comparable. As expected, the projection error in surfaces at 90° produced a circular distribution in the density maps, indicating that deviations along the L–R and I–S axes were similar. In contrast, the distributions at 75° and 45° appeared horizontally elongated, reflecting lower vertical error along the I–S axis. This is consistent with the geometry of inclined planes, where projection errors along the vertical axis are smaller than the absolute deviations measured on the sloped surface. At 30°, 15°, and 0°, however, the error in the L–R axis was no longer comparable to that in the P–A axis. This confirms that depth perception remains a primary limitation in AR projection accuracy when using head-mounted displays like the HoloLens 2 [42], [43].\u003c/p\u003e\n\u003cp\u003eTable 2 compares the axis-specific projection errors obtained in this study with those reported in [43]. In our case, the axis-wise error ranged from 1.4 ± 1.2 mm for Vuforia (perpendicular view, 90°), to 1.9 ± 1.4 mm for Vuforia (tilted view, 45°). In contrast, the results reported in [43] showed much greater variability, ranging from 0.8 ± 0.2 mm with QR tracking (perpendicular view, 90°) and a distance of 300 mm, to as high as 6.3 ± 0.1 mm for Vuforia (tilted view, 30°), also at 300 mm. Despite this variability, our results fall within the same order of magnitude, supporting the reliability and achievable accuracy of both tracking methods.\u003c/p\u003e\n\u003cp\u003eTo conclude, it is worth noting that this application may be compatible not only with Microsoft HoloLens 2, but also with other MRTK-based headsets—such as Magic Leap 2—with only minor adjustments to the Unity project [44]. Given the compatibility of this alternative device with both Vuforia and QR code tracking [45], [46], the long-term usability of EZ-AR is ensured, regardless of the future of Microsoft HoloLens 2.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eIn this work, we presented EZ-AR, a lightweight, open-access AR framework designed to facilitate the deployment of personalized 3D content on Microsoft HoloLens 2 without requiring programming expertise. By combining a user-friendly module in 3D Slicer with a configurable AR application, the system enables clinical users to visualize custom virtual models anchored to physical reference markers using either Vuforia or QR code tracking. Our evaluation demonstrates that both tracking methods offer comparable AR projection accuracy with errors in the range of 3–4 mm under favorable conditions and from clinically realistic distances. While QR tracking offers greater versatility in marker design and requires no prior configuration, Vuforia provides higher flexibility in user movement around the AR reference marker. Each advantage may be better suited to different clinical scenarios, depending on the available resources, spatial constraints, and interaction requirements of the procedure. Our results validate the system’s usability, stability, and robustness, reinforcing its potential as a versatile tool to support visualization and planning tasks in healthcare environments.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e· \u003cstrong\u003eFunding\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eResearch supported by projects TED2021-129392B-I00, TED2021-132200B-I00 and PID2023-149604OB-I00 (MCIU/AEI/10.13039/501100011033 and European Union “NextGenerationEU”/PRTR) and Comunidad de Madrid (MAGERIT-CM TEC-2024/COM-44 and Multiannual Agreement with UC3M SAFEDELIVERY-CM-UC3M).\u003c/p\u003e\n\u003cp\u003e· \u003cstrong\u003eConflict of interest\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no conflict of interest.\u003c/p\u003e\n\u003cp\u003e· \u003cstrong\u003eEthics statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study did not involve experimentation on human subjects, nor did it collect or analyze personal or medical data. The experiments were conducted using a 3D-printed validation platform and voluntary participation from individuals who interacted with a non-invasive AR system. As such, no formal ethics approval was required according to the regulations of our institution. All participants were informed of the nature of the study and gave verbal consent to participate. No identifiable personal information was recorded, and all privacy rights were respected throughout the study.\u003c/p\u003e\n\u003cp\u003e· \u003cstrong\u003eAuthors' contributions\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConception and design of the work: A. P.-D.-d.-l.-L., J. A. C.-H., R. P.-M., J. P.; Investigation, data curation, analysis, and interpretation: A. P.-D.-d.-l.-L., G. A.-A., J. P.; Software: A. P.-D.-d.-l.-L., G. A.-A.; Funding acquisition, supervision and project administration: J. A. C.-H., R. P.-M., J. P.; Writing - original draft preparation: A. P.-D.-d.-l.-L., G. A.-A; Writing – review and editing: J. P. All authors have read and agreed to the published version of the manuscript. They all have also agreed to be personally accountable for their own contributions and to ensure that questions related to the accuracy or integrity of any part of the work are appropriately investigated, resolved, and the resolution documented in the literature.\u003c/p\u003e\n\u003cp\u003e· \u003cstrong\u003eAuthors' information\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eo Alicia Pose Díez de la Lastra: [email protected] (A. P.-D.-d.-l.-L.)\u003c/p\u003e\n\u003cp\u003eo Gemma Arce Alonso: [email protected] (G. A.-A)\u003c/p\u003e\n\u003cp\u003eo José Antonio Calvo Haro: [email protected] (J. A. C.-H.)\u003c/p\u003e\n\u003cp\u003eo Rubén Pérez Mañanes: [email protected] (R. P.-M.)\u003c/p\u003e\n\u003cp\u003eo Javier Pascau: [email protected] (J. P.)\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eC. H. Araujo, J. A. Aguirre, and L. Puig, \u0026ldquo;3D Printing Applications and Extended Realities in Medicine: Systematic Review,\u0026rdquo; \u003cem\u003e3D Printing and Additive Manufacturing\u003c/em\u003e, Oct. 2024, doi: 10.1089/3dp.2024.0128.\u003c/li\u003e\n\u003cli\u003eZ. M. C. Baum \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Augmented Reality Training Platform for Neurosurgical Burr Hole Localization,\u0026rdquo; \u003cem\u003eJournal of Medical Robotics Research\u003c/em\u003e, vol. 04, no. 03n04, pp. 1942001\u0026ndash;1942001, Sep. 2019, doi: 10.1142/S2424905X19420017.\u003c/li\u003e\n\u003cli\u003eD. Amiras \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Augmented reality simulator for CT-guided interventions,\u0026rdquo; \u003cem\u003eEuropean Radiology\u003c/em\u003e, Jun. 2021, doi: 10.1007/s00330-021-08043-0.\u003c/li\u003e\n\u003cli\u003eR. N. Uppot \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Implementing virtual and augmented reality tools for radiology education and training, communication, and clinical care,\u0026rdquo; \u003cem\u003eRadiology\u003c/em\u003e, vol. 291, no. 3, pp. 570\u0026ndash;580, 2019, doi: 10.1148/radiol.2019182210.\u003c/li\u003e\n\u003cli\u003eH. El-Hariri, P. Pandey, A. J. Hodgson, and R. Garbi, \u0026ldquo;Augmented reality visualisation for orthopaedic surgical guidance with pre- and intra-operative multimodal image data fusion,\u0026rdquo; \u003cem\u003eHealthcare Technology Letters\u003c/em\u003e, vol. 5, no. 5, pp. 189\u0026ndash;193, 2018, doi: 10.1049/htl.2018.5061.\u003c/li\u003e\n\u003cli\u003eP. Pratt \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Through the HoloLens\u003csup\u003eTM\u003c/sup\u003e looking glass: augmented reality for extremity reconstruction surgery using 3D vascular models with perforating vessels,\u0026rdquo; \u003cem\u003eEuropean Radiology Experimental\u003c/em\u003e, vol. 2, no. 1, pp. 0\u0026ndash;6, 2018, doi: 10.1186/s41747-017-0033-2.\u003c/li\u003e\n\u003cli\u003eJ. Yoon \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Augmented reality for the surgeon: Systematic review,\u0026rdquo; \u003cem\u003eThe international journal of medical robotics + computer assisted surgery\u003c/em\u003e\u003cem\u003e : MRCAS\u003c/em\u003e, vol. 14, no. 4, Aug. 2018, doi: 10.1002/RCS.1914.\u003c/li\u003e\n\u003cli\u003eM. Miller Koop \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;The Microsoft HoloLens 2 Provides Accurate Measures of Gait, Turning, and Functional Mobility in Healthy Adults,\u0026rdquo; \u003cem\u003eSensors\u003c/em\u003e, vol. 22, no. 5, Art. no. 5, Jan. 2022, doi: 10.3390/s22052009.\u003c/li\u003e\n\u003cli\u003eA. Palumbo, \u0026ldquo;Microsoft HoloLens 2 in Medical and Healthcare Context: State of the Art and Future Prospects,\u0026rdquo; \u003cem\u003eSensors\u003c/em\u003e, vol. 22, no. 20, Art. no. 20, Jan. 2022, doi: 10.3390/s22207709.\u003c/li\u003e\n\u003cli\u003eM. E. de Oliveira, H. G. Debarba, A. L\u0026auml;dermann, S. Chagu\u0026eacute;, and C. Charbonnier, \u0026ldquo;A hand-eye calibration method for augmented reality applied to computer-assisted orthopedic surgery,\u0026rdquo; \u003cem\u003eThe International Journal of Medical Robotics and Computer Assisted Surgery\u003c/em\u003e, vol. 15, no. 2, p. e1969, 2019, doi: 10.1002/rcs.1969.\u003c/li\u003e\n\u003cli\u003eJ. Hajek \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Closing the Calibration Loop: An Inside-Out-Tracking Paradigm for Augmented Reality in Orthopedic Surgery,\u0026rdquo; presented at the Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), https://doi.org/10.1007/978-3-030-00937-3_35, 2018, pp. 299\u0026ndash;306. doi: 10.1007/978-3-030-00937-3_35.\u003c/li\u003e\n\u003cli\u003eE. Rae, A. Lasso, M. S. Holden, E. Morin, R. Levy, and G. Fichtinger, \u0026ldquo;Neurosurgical burr hole placement using the Microsoft HoloLens,\u0026rdquo; in \u003cem\u003eMedical Imaging 2018: Image-Guided Procedures, Robotic Interventions, and Modeling\u003c/em\u003e, SPIE, Mar. 2018, pp. 190\u0026ndash;197. doi: 10.1117/12.2293680.\u003c/li\u003e\n\u003cli\u003eR. Vassallo, A. Rankin, E. C. S. Chen, and T. M. Peters, \u0026ldquo;Hologram stability evaluation for Microsoft HoloLens,\u0026rdquo; in \u003cem\u003eMedical Imaging 2017: Image Perception, Observer Performance, and Technology Assessment\u003c/em\u003e, SPIE, Mar. 2017, pp. 295\u0026ndash;300. doi: 10.1117/12.2255831.\u003c/li\u003e\n\u003cli\u003eR. Tang \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Augmented reality navigation in open surgery for hilar cholangiocarcinoma resection with hemihepatectomy using video-based in situ three-dimensional anatomical modeling: A case report,\u0026rdquo; \u003cem\u003eMedicine\u003c/em\u003e, vol. 96, no. 37, Sep. 2017, doi: 10.1097/MD.0000000000008083.\u003c/li\u003e\n\u003cli\u003eJ. W. Meulstee \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Toward Holographic-Guided Surgery,\u0026rdquo; \u003cem\u003eSurgical Innovation\u003c/em\u003e, vol. 26, no. 1, pp. 86\u0026ndash;94, Feb. 2019, doi: 10.1177/1553350618799552.\u003c/li\u003e\n\u003cli\u003eI. Kuhlemann, M. Kleemann, P. Jauer, A. Schweikard, and F. Ernst, \u0026ldquo;Towards X-ray free endovascular interventions - Using HoloLens for on-line holographic visualisation,\u0026rdquo; presented at the Healthcare Technology Letters, Institution of Engineering and Technology, 2017, pp. 184\u0026ndash;187. doi: 10.1049/htl.2017.0061.\u003c/li\u003e\n\u003cli\u003eM. Akulauskas, K. Butkus, V. Rutkūnas, T. Blažauskas, and D. Jegelevičius, \u0026ldquo;Implementation of Augmented Reality in Dental Surgery Using HoloLens 2: An In Vitro Study and Accuracy Assessment,\u0026rdquo; \u003cem\u003eApplied Sciences\u003c/em\u003e, vol. 13, no. 14, Art. no. 14, Jan. 2023, doi: 10.3390/app13148315.\u003c/li\u003e\n\u003cli\u003eC. Kunz \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Infrared marker tracking with the HoloLens for neurosurgical interventions,\u0026rdquo; \u003cem\u003eCurrent Directions in Biomedical Engineering\u003c/em\u003e, vol. 6, no. 1, May 2020, doi: 10.1515/cdbme-2020-0027.\u003c/li\u003e\n\u003cli\u003eD. Ungureanu \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;HoloLens 2 Research Mode as a Tool for Computer Vision Research,\u0026rdquo; Aug. 25, 2020, \u003cem\u003earXiv\u003c/em\u003e: arXiv:2008.11239. doi: 10.48550/arXiv.2008.11239.\u003c/li\u003e\n\u003cli\u003eL. Qian, A. Deguet, and P. Kazanzides, \u0026ldquo;ARssist: augmented reality on a head-mounted display for the first assistant in robotic surgery,\u0026rdquo; \u003cem\u003eHealthcare Technology Letters\u003c/em\u003e, vol. 5, no. 5, pp. 194\u0026ndash;200, 2018, doi: 10.1049/htl.2018.5065.\u003c/li\u003e\n\u003cli\u003eM. Doughty and N. R. Ghugre, \u0026ldquo;Head-Mounted Display-Based Augmented Reality for Image-Guided Media Delivery to the Heart: A Preliminary Investigation of Perceptual Accuracy,\u0026rdquo; \u003cem\u003eJournal of Imaging\u003c/em\u003e, vol. 8, no. 2, Art. no. 2, Feb. 2022, doi: 10.3390/jimaging8020033.\u003c/li\u003e\n\u003cli\u003eA. Pose-D\u0026iacute;ez-de-la-Lastra \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Microsoft HoloLens 2 vs. tablet-based augmented reality and 3D printing for fronto-orbital reconstruction of craniosynostosis: a case study,\u0026rdquo; \u003cem\u003e3D Printing in Medicine\u003c/em\u003e, vol. 11, no. 1, p. 13, Mar. 2025, doi: 10.1186/s41205-025-00251-4.\u003c/li\u003e\n\u003cli\u003eJ. N. Costa \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Ultrasound training simulator using augmented reality glasses: an accuracy and precision assessment study,\u0026rdquo; in \u003cem\u003e2022 44th Annual International Conference of the IEEE Engineering in Medicine \u0026amp; Biology Society (EMBC)\u003c/em\u003e, Jul. 2022, pp. 4461\u0026ndash;4464. doi: 10.1109/EMBC48229.2022.9871406.\u003c/li\u003e\n\u003cli\u003eS. Pieper, M. Halle, and R. Kikinis, \u0026ldquo;3D Slicer; 3D Slicer,\u0026rdquo; \u003cem\u003e2004 2nd IEEE International Symposium on Biomedical Imaging: Nano to Macro (IEEE Cat No. 04EX821)\u003c/em\u003e, 2004, doi: 10.1109/ISBI.2004.1398617.\u003c/li\u003e\n\u003cli\u003eM. Garc\u0026iacute;a‐Sevilla \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Augmented reality as a tool to guide psi placement in pelvic tumor resections,\u0026rdquo; \u003cem\u003eSensors\u003c/em\u003e, vol. 21, no. 23, 2021, doi: 10.3390/s21237824.\u003c/li\u003e\n\u003cli\u003eA. D\u0026iacute;ez-Montiel, A. Pose-D\u0026iacute;ez-de-la-Lastra, A. Gonz\u0026aacute;lez-\u0026Aacute;lvarez, J. I. Salmer\u0026oacute;n, J. Pascau, and S. Ochandiano, \u0026ldquo;Tablet-based Augmented reality and 3D printed templates in fully guided Microtia Reconstruction: a clinical workflow,\u0026rdquo; \u003cem\u003e3D Printing in Medicine\u003c/em\u003e, vol. 10, no. 1, p. 17, May 2024, doi: 10.1186/s41205-024-00213-2.\u003c/li\u003e\n\u003cli\u003eP. Liu \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;Deep learning to segment pelvic bones: large-scale CT datasets and baseline models,\u0026rdquo; \u003cem\u003eInt J Comput Assist Radiol Surg\u003c/em\u003e, vol. 16, no. 5, pp. 749\u0026ndash;756, May 2021, doi: 10.1007/s11548-021-02363-8.\u003c/li\u003e\n\u003cli\u003e\u0026ldquo;Vuforia Enterprise Augmented Reality (AR) Software | PTC.\u0026rdquo; Accessed: May 03, 2023. [Online]. Available: https://www.ptc.com/en/products/vuforia\u003c/li\u003e\n\u003cli\u003eX. Cong and T. Li, \u0026ldquo;Design and Development of Virtual Medical System Interface Based on VR-AR Hybrid Technology,\u0026rdquo; \u003cem\u003eComputational and Mathematical Methods in Medicine\u003c/em\u003e, vol. 2020, p. e7108147, Aug. 2020, doi: 10.1155/2020/7108147.\u003c/li\u003e\n\u003cli\u003eE. M. Bourhim and A. Akhiate, \u0026ldquo;Augmented Reality SDK\u0026rsquo;s: A Comparative Study,\u0026rdquo; in \u003cem\u003eIntelligent Systems Design and Applications\u003c/em\u003e, A. Abraham, N. Gandhi, T. Hanne, T.-P. Hong, T. Nogueira Rios, and W. Ding, Eds., Cham: Springer International Publishing, 2022, pp. 559\u0026ndash;566. doi: 10.1007/978-3-030-96308-8_52.\u003c/li\u003e\n\u003cli\u003eB. J. Park, S. J. Hunt, G. J. Nadolski, and T. P. Gade, \u0026ldquo;Augmented reality improves procedural efficiency and reduces radiation dose for CT-guided lesion targeting: a phantom study using HoloLens 2,\u0026rdquo; \u003cem\u003eSci Rep\u003c/em\u003e, vol. 10, no. 1, p. 18620, Oct. 2020, doi: 10.1038/s41598-020-75676-4.\u003c/li\u003e\n\u003cli\u003eM. Benmahdjoub, W. J. Niessen, E. B. Wolvius, and T. van Walsum, \u0026ldquo;Virtual extensions improve perception-based instrument alignment using optical see-through devices,\u0026rdquo; \u003cem\u003eIEEE Transactions on Visualization and Computer Graphics\u003c/em\u003e, vol. 27, no. 11, pp. 4332\u0026ndash;4341, Nov. 2021, doi: 10.1109/TVCG.2021.3106506.\u003c/li\u003e\n\u003cli\u003eY. Nagayo, T. Saito, and H. Oyama, \u0026ldquo;A Novel Suture Training System for Open Surgery Replicating Procedures Performed by Experts Using Augmented Reality,\u0026rdquo; \u003cem\u003eJ Med Syst\u003c/em\u003e, vol. 45, no. 5, p. 60, Apr. 2021, doi: 10.1007/s10916-021-01735-6.\u003c/li\u003e\n\u003cli\u003eA. Thabit, M. Benmahdjoub, M.-L. C. van Veelen, W. J. Niessen, E. B. Wolvius, and T. van Walsum, \u0026ldquo;Augmented reality navigation for minimally invasive craniosynostosis surgery: a phantom study,\u0026rdquo; \u003cem\u003eInt J CARS\u003c/em\u003e, vol. 17, no. 8, pp. 1453\u0026ndash;1460, Aug. 2022, doi: 10.1007/s11548-022-02634-y.\u003c/li\u003e\n\u003cli\u003eP. B. Johnson, A. Jackson, M. Saki, E. Feldman, and J. Bradley, \u0026ldquo;Patient posture correction and alignment using mixed reality visualization and the HoloLens 2,\u0026rdquo; \u003cem\u003eMedical Physics\u003c/em\u003e, vol. 49, no. 1, pp. 15\u0026ndash;22, 2022, doi: 10.1002/mp.15349.\u003c/li\u003e\n\u003cli\u003e\u0026ldquo;Vuforia Package for Unity - Vuforia Engine Library.\u0026rdquo; Accessed: May 15, 2025. [Online]. Available: https://developer.vuforia.com/library/vuforia-engine/unity-extension/vuforia-engine-package-unity/\u003c/li\u003e\n\u003cli\u003elolambean, \u0026ldquo;QR code tracking overview - Mixed Reality.\u0026rdquo; Accessed: Apr. 29, 2025. [Online]. Available: https://learn.microsoft.com/en-us/windows/mixed-reality/develop/advanced-concepts/qr-code-tracking-overview\u003c/li\u003e\n\u003cli\u003eY. Li, A. Reed, N. Kavoussi, and J. Y. Wu, \u0026ldquo;Eye gaze metrics for skill assessment and feedback in kidney stone surgery,\u0026rdquo; \u003cem\u003eInt J CARS\u003c/em\u003e, vol. 18, no. 6, pp. 1127\u0026ndash;1134, Jun. 2023, doi: 10.1007/s11548-023-02901-6.\u003c/li\u003e\n\u003cli\u003eL. Cercenelli, N. Emiliani, C. Gulotta, M. Bevini, G. Badiali, and E. Marcelli, \u0026ldquo;Augmented Reality to Assist in the Diagnosis of Temporomandibular Joint Alterations,\u0026rdquo; \u003cem\u003eInf Syst Front\u003c/em\u003e, vol. 27, no. 1, pp. 33\u0026ndash;49, Feb. 2025, doi: 10.1007/s10796-024-10545-3.\u003c/li\u003e\n\u003cli\u003eS. Navid, T. Mokhtari, T. Alizamir, A. Arabkheradmand, and G. Hassanzadeh, \u0026ldquo;Determination of Stature from Upper Arm Length in Medical Students,\u0026rdquo; \u003cem\u003eAnatomical Sciences Journal\u003c/em\u003e, vol. 11, no. 3, pp. 135\u0026ndash;140, Aug. 2014.\u003c/li\u003e\n\u003cli\u003eR. Moreta-Martinez, A. Pose-D\u0026iacute;ez-de-la-Lastra, J. A. Calvo-Haro, L. Mediavilla-Santos, R. P\u0026eacute;rez-Ma\u0026ntilde;anes, and J. Pascau, \u0026ldquo;Combining Augmented Reality and 3D Printing to Improve Surgical Workflows in Orthopedic Oncology: Smartphone Application and Clinical Evaluation,\u0026rdquo; \u003cem\u003eSensors 2021, Vol. 21, Page 1370\u003c/em\u003e, vol. 21, no. 4, pp. 1370\u0026ndash;1370, Feb. 2021, doi: 10.3390/S21041370.\u003c/li\u003e\n\u003cli\u003eA. Pose-D\u0026iacute;ez-De-la-Lastra \u003cem\u003eet al.\u003c/em\u003e, \u0026ldquo;HoloLens 1 vs. HoloLens 2: Improvements in the New Model for Orthopedic Oncological Interventions,\u0026rdquo; \u003cem\u003eSensors\u003c/em\u003e, vol. 22, no. 13, 2022, doi: 10.3390/s22134915.\u003c/li\u003e\n\u003cli\u003eG. M. Costa, M. R. Petry, J. G. Martins, and A. P. G. M. Moreira, \u0026ldquo;Assessment of Multiple Fiducial Marker Trackers on Hololens 2,\u0026rdquo; \u003cem\u003eIEEE Access\u003c/em\u003e, vol. 12, pp. 14211\u0026ndash;14226, 2024, doi: 10.1109/ACCESS.2024.3356722.\u003c/li\u003e\n\u003cli\u003e\u0026ldquo;HoloLens 2 to Magic Leap 2 Porting Guide | MagicLeap Developer Documentation.\u0026rdquo; Accessed: Jun. 27, 2025. [Online]. Available: https://developer-docs.magicleap.cloud/docs/guides/third-party/mrtk/hololens-porting/\u003c/li\u003e\n\u003cli\u003e\u0026ldquo;Magic Leap - Vuforia Engine Library.\u0026rdquo; Accessed: Jun. 27, 2025. [Online]. Available: https://developer.vuforia.com/library/vuforia-engine/getting-started/development-environments/getting-started-vuforia-engine-and-magic-leap-2/\u003c/li\u003e\n\u003cli\u003e\u0026ldquo;Marker Tracking | MagicLeap Developer Documentation.\u0026rdquo; Accessed: Jun. 27, 2025. [Online]. Available: https://developer-docs.magicleap.cloud/docs/guides/features/marker-tracking/\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Augmented reality, Microsoft HoloLens 2, marker-based tracking, surgical planning, computer-assisted interventions.","lastPublishedDoi":"10.21203/rs.3.rs-7545942/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7545942/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Augmented Reality (AR) offers significant potential to enhance surgical precision, improve clinical training, and support intraoperative decision-making. However, developing personalized AR applications for head-mounted displays such as Microsoft HoloLens 2 typically requires advanced technical expertise. Additionally, identifying a robust and accessible method for registering virtual models to the real world remains an open challenge. These limitations hinder the widespread adoption of AR in healthcare, particularly among non-technical clinical users.\nThis work presents EZ-AR, a user-friendly AR application for Microsoft HoloLens 2 that automates complex tasks such as configuration and deployment, while ensuring low latency, high reliability, and smooth interaction. Ultimately, the system enables non-expert users to independently load and visualize personalized 3D models anchored to physical reference markers. The application accesses files from Google Drive in seconds and displays them in the physical environment using one of two integrated tracking methods: Vuforia or QR code detection. Once configured, the application no longer requires Internet access.\nThe system's usability was assessed in terms of detection range and AR projection accuracy under varying surface inclinations and viewing angles. Results show a mean projection error of 3.3 ± 2.2 mm with QR code tracking and 3.9 ± 2.7 mm with Vuforia, both measured at a working distance compatible with typical surgical ergonomics. Further analysis revealed that accuracy is more influenced by viewing angle and surface inclination than by the tracking method itself. Overall, EZ-AR provides a robust and accessible AR framework to democratize AR adoption in clinical environments.","manuscriptTitle":"EZ-AR: A universal AR application for multi-model visualization and real-time tracking in healthcare settings","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-12-22 17:20:46","doi":"10.21203/rs.3.rs-7545942/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"ac4ba54f-ad18-4768-9457-c7dfc662eb59","owner":[],"postedDate":"December 22nd, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-01-21T10:50:09+00:00","versionOfRecord":[],"versionCreatedAt":"2025-12-22 17:20:46","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7545942","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7545942","identity":"rs-7545942","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}