AngioSim: A Novel Augmented-Reality Angiography Simulator for Radiation-Free Neurointerventional Training

Abstract

Cardiovascular diseases require fast and precise treatment, often involving angiography for diagnosis and intervention. However, training in angiographic procedures often entails exposure to ionizing radiation, which carries inherent risks. To reduce this exposure and enhance training realism, we developed AngioSim—a novel augmented-reality angiography simulation system combined with a vascular silicone simulator. This study evaluates the realism, effectiveness, and potential benefits of AngioSim for neurointerventional training. AngioSim was tested during neurointerventional training sessions with 24 physicians at RWTH Aachen University Hospital. Participants completed a questionnaire assessing realism, usefulness, and preferences compared to other simulators using a Likert scale. Responses were converted to binary categories and McNemar tests were applied for paired comparisons. A total of 92% of physicians rated guidewire and catheter visibility during fluoroscopy as sufficient, while 86% found RM and DSA simulations realistic. AngioSim was preferred over camera-based silicone simulators by 93%, and 96% of physicians rated it necessary for training—significantly more than other simulators (p < 0.05). These results demonstrate the high acceptance and perceived realism of the system and suggest that AngioSim offers advantages over existing training methods. AngioSim offers a realistic, cost-effective, and radiation-free training solution while maintaining the benefits of silicone models. It showed high utility for training purposes, making it a promising addition to neurointerventional programs.

1. Introduction

Cardiovascular diseases (CVDs) are the leading cause of death in Europe [1]. Fast and appropriate intervention in various CVDs, such as ischemic strokes, is essential for successful treatment [2]. Angiography is an important diagnostic as well as therapeutic procedure in the endovascular treatment of CVDs. It is essential for the visualization of the vascular system, the identification of pathologies, and the utilization of medical devices using ionizing radiation and contrast agents [3]. The associated radiological techniques, such as digital subtraction angiography (DSA) and roadmap (RM) fluoroscopy, along with medical device control, including the use of stent retrievers, coils, or flow diverters, are complex and require extensive training for neurointerventionalists to acquire the necessary skills [4,5]. While angiography procedures are generally safe, with only 3.3% of patients experiencing minor complications such as transient neurological symptoms [6], their complexity underscores the need for thorough training. Moreover, in cases of ischemic stroke, time to reperfusion is critical, as shorter times have been associated with significantly better clinical outcomes, including reduced mortality and improved functional independence [7].

Despite its benefits in disease management, occupational ionizing radiation exposure poses risks to medical staff and should be minimized as much as possible [8,9,10].

To minimize these risks and reduce costs, particularly during neurointerventional training, training methods such as computer simulators, vascular silicone simulators, or in vivo models are used. These methods allow for the simulation of basic angiographic procedures such as aneurysm coiling, thrombectomy, or diagnostic DSA, as well as the generation and application of RM [5,11]. However, the realism of the simulation remains a major challenge, especially in the visualizing representation of silicone simulators outside an angiography room and the performance of procedures such as DSA and RM.

In this study, we developed and tested the neurointerventional training environment AngioSim (Angiography Simulator) that can simulate angiographic procedures like DSA and RM without the need for ionizing radiation, making handling and visual simulation more realistic. By integrating augmented reality, AngioSim enhances neurointerventional training, reduces the need for radiation exposure for trainees, and provides a cost-effective method while maintaining the characteristics of vascular silicone simulators. Neuroradiologists who participated in an internal neurointerventional training with different simulators were asked to evaluate their experience with AngioSim. Our goal was also to assess the value of AngioSim for neurointerventional training. We hypothesized that AngioSim would offer significant benefits over other simulators, especially over silicone models lacking the simulation of the visual experience through augmented reality.

This study makes several key contributions. First, we introduce AngioSim, a novel angiography simulation system that integrates augmented reality with a vascular silicone model to realistically simulate fluoroscopy, RM, and DSA without the use of ionizing radiation. We also present a cost-efficient and modular hardware setup based on widely available components, which allows for straightforward implementation in training environments. Furthermore, we evaluate AngioSim’s realism, usefulness, and acceptance using feedback from 24 physicians and apply McNemar tests to identify statistically significant differences compared to other training systems. Lastly, we provide open access to the simulator’s design and a detailed component list to support transparency, replication, and scalability in medical education.

2. Materials and Methods

The novel angiography simulation system for vascular silicone simulators was used by twenty-four physicians at the Diagnostic and Interventional Neuroradiology department, University Hospital, RWTH Aachen during internal neurointerventional device training sessions between 2 September 2024 and 13 September 2024.

To evaluate the prototype, each physician was asked to complete a questionnaire after the training session. This included questions about neuroendovascular intervention and training experience, their assessment of the realism of the AngioSim simulator, their preferences regarding the use of AngioSim compared to other training methods, and their evaluation of the need for different training options for neuroendovascular interventions.

This study was approved by the Ethics Committee of the University Hospital RWTH Aachen. (EK 24-306). Individual consent was waived since participation was voluntary and the questionnaires did not contain any personal information and were thus fully anonymous.

The aim of the training sessions was to practice basic angiographic skills with different materials using a computer simulator and a vascular silicone simulator in combination with AngioSim. Scenarios comprised navigating to segments of the anterior, middle, and posterior cerebral artery under RM guidance and performing a DSA using different kinds and combinations of catheters and wires.

2.1. Computer Simulator

The computer simulator used in this study, Vascular Intervention System Trainer (VIST G5, Mentice, Gothenburg, Sweden), includes modules for interventions on the femoral, iliac, aortic, renal, carotid, neural, and coronary arteries. The simulator features two monitors and two joysticks for controlling the table, along with various settings accessible through buttons for tasks like adjusting the zoom. Additionally, foot pedals are provided for operating the fluoroscopy and cine loop functions. Real interventional instruments, such as wires and catheters, can be used after their tips have been cut off. The system digitally simulates the characteristics of these instruments as well as X-ray imaging and cine loops. Contrast agent behavior is mimicked by injecting air via a syringe. The simulator provides virtual haptic feedback to imitate the tactile sensations experienced during actual interventions.

2.2. Vascular Silicone Simulator

The vascular silicone simulator (Neuro Testing Model Plus, United Biologics, Santa Ana, CA, USA) (Figure 1a) used in combination with the AngioSim is made of soft transparent silicone, enabling the direct visualization of endovascular devices. The silicone model includes the aorta, the supraaortic branches, the Circle of Willis, and the proximal segments of the anterior, middle, and posterior cerebral artery. The vascular silicone simulator is filled with distilled water at room temperature, and soap is added to reduce friction. To simulate blood flow, the inlet of the vascular silicone simulator at the aortic arch is connected to a pulsatile pump (Figure 1i), which is connected to a water reservoir (Figure 1j). When not used with AngioSim, the simulator is typically paired with a camera (Figure 1b), with the video feed displayed on a screen (Figure 1c) in front of the physician.

2.3. AngioSim

The AngioSim consists of two main compartments: the traditional vascular silicone simulator (as described above) with components for flow control and basic visualization (Figure 1a–c,i,j), and the platform that enables the simulation of angiographic procedures (Figure 1d–h,k,l).

To perform a DSA or RM, a ready-made red dye solution (Supreme Blood Internal, Kryolan, Berlin, Germany) is used as a contrast agent substitute. To prevent contamination of the entire water volume when the contrast agent substitute is injected into the system, the outlet of the vascular silicone simulator located at the cranial vessels is connected to a 6V solenoid 2-position 3-way fluid valve (Figure 1k), which can redirect the fluid flow from the water reservoir to a wastewater reservoir (Figure 1l). This solenoid valve is controlled by a relay (KF-301 5V 1-relay module, AZ-Delivery, Deggendorf, Germany) (Figure 1g), powered by an external 9V battery. The valve is operated by the AngioSim software v1.0 on a Windows notebook (Figure 1j) via a USB-to-GPIO adapter (Adafruit MCP2221A Breakout—General Purpose USB to GPIO ADC I2C, Adafruit, New York, NY, USA) (Figure 1g).

For image acquisition, a document camera (Epson ELPDC21 Document Camera, Epson, Suwa, Japan) (Figure 1b) is connected to the notebook via USB. The edited visual input is displayed on a 55-inch screen (PH55F-P, Samsung, Seoul, Republic of Korea) (Figure 1c) via HDMI. Two pedals of a PC-compatible programmable three-pedal footswitch (PCsensor USB Programmable 3-Pedal Foot Switch, PCsensor, Shenzhen, China) (Figure 1f) are used to control the X-ray and cine loop functions. If the cine pedal is pressed, the solenoid valve redirects the liquid flow into the wastewater reservoir. The effect of the pedal sustains for additional 10 s after it is released to prevent contrast agent from entering the main compartment. A numeric keypad (LogiLink ID0120, LogiLink, Schalksmühle, Germany) (Figure 1e) is used to control the different modules of the simulator. The total cost of the AngioSim extension amounted to EUR 68.12, which covered the newly purchased components. A detailed price list of these materials is available at https://zenodo.org/records/15148405 (accessed on 4 April 2025). The remaining equipment, including the silicone simulator, pulsatile pump, laptop, document camera, and display, was already available and did not incur additional costs.

2.4. Software

To simulate angiographic procedures in real time, AngioSim utilizes image-processing techniques that mimic the visual output of X-ray-based imaging. The system captures live video from the vascular silicone simulator and applies grayscale conversion, background subtraction, and intensity-based masking to generate realistic fluoroscopy, RM, and DSA visualizations. These processes are triggered via footswitch input and processed frame by frame using Python-based algorithms. The resulting images are displayed in real time or recorded for playback.

The AngioSim software was developed using Python 3.10 (Python Software Foundation, Wilmington, DE, USA, 2021) and incorporates the PyGame 2.4.0 interface library (PyGame, London, UK, 2023) to facilitate the simulation process. Additionally, the OpenCV 4.8.0 library (OpenCV, Palo Alto, CA, USA, 2023) and NumPy 1.26.0 library (NumPy, Austin, TX, USA, 2023) are employed for image-processing tasks. The CircuitPython library (CircuitPython, New York, NY, USA, 2023) is used as the interface to the USB-to-GPIO adapter.

AngioSim features a modular design, allowing for easy interchangeability between different imaging modules to replicate angiography techniques and processes more closely. These modules, including fluoroscopy, RM, DSA, and DSA Viewer, are accessible through the numeric keypad and can be controlled using the footswitch to mimic a real angiographic unit. For example, if the footswitch is not engaged, the program and video feed will pause. The numeric keypad also enables the adjustment of zoom, the use of a collimator, and a display for instructions. When initiated, the program displays a start screen, which gives an overview of the functions and modules. The zoom, aperture, and help functions can be controlled via the numpad.

2.5. Fluoroscopy Mode

To enhance vessel and device visibility, as well as the overall visual experience, the fluoroscopy mode converts the camera signal to grayscale, rotates it by 90 degrees to replicate the orientation of an angiography unit, and displays the result in real time. The model’s clear silicone vessels naturally allow for the visualization of vessels and endovascular devices. Figure 2a illustrates the fluoroscopy simulation with a balloon catheter, guiding catheter, and guide wire, while Figure 2b shows a balloon catheter, microcatheter, and microwire.

2.6. Roadmap (RM)

AngioSim’s RM mode mimics the roadmap generation process of a real angiography unit, enhancing vascular visualization. When ‘Cine’ is pressed, a subtraction mask is generated by accumulating all recorded frames until ‘Cine’ is released. The accumulation is achieved by taking the minimum intensity of each pixel across frames to enhance vessel visibility under contrast agent injection, providing a clear image of the contrasted vessels. The mask generation is displayed in real time to guide the procedure, as illustrated in Figure 3a, which shows an accumulated RM mask. When a mask is successfully captured, pressing ‘X-ray’ subtracts the mask from the live image to reveal the contrasted vascular structures, as shown in Figure 3b. Figure 3c,d show how the RM mode highlights changes in the image, such as the movement of catheters and wires. Additional image processing, including brightness adjustments, is applied to further refine the RM image.

2.7. Digital Subtraction Angiography (DSA)

The DSA mode removes background structures to visualize contrasted vessels and fluid flow. First, the mask of the vascular structures without contrast agent is acquired when the “Cine” pedal is pressed. The program calculates the absolute difference between the grayscale live feed and the mask, highlighting changes where contrast agent is present. This image is then inverted to simulate real angiographic images, where vessels appear lighter. To further resemble original DSA images, a black background is blended with the inverted difference image, followed by additional image processing, such as brightness and adjustments, to refine the DSA. As shown in Figure 4, the images are processed and displayed at a reduced frame rate of 3 frames per second (fps), replicating the low fps used in actual DSA. All DSA images are stored and can be viewed as a loop forward and backward with the DSA Viewer, using designated keys on the numeric keypad.

2.8. Questionnaire

Physicians were given a questionnaire to complete immediately after the training. The questionnaire included a section on the physicians’ previous experience, such as their years of experience, the number of neuroendovascular interventions they had assisted with or performed independently, and their experience with various neuroendovascular training options, including computer simulators, vascular silicone simulators, or in vivo models. Following this, physicians were asked to rate the image quality and realism of the simulator, their preferences for simulators, their overall experience, and how necessary they rate the simulator for training in neuroendovascular procedures, using a Likert scale for their ratings. The questionnaire and the participants’ responses are available at https://zenodo.org/records/15148405 (accessed on 4 April 2025).

2.9. Analysis

We evaluated the responses from all twenty-four physicians. To simplify the results and facilitate clear conclusions, Likert-scale responses were converted into a binary format: ‘strongly disagree,’ ‘disagree,’ and ‘neutral’ were grouped as ‘no,’ while ‘agree’ and ‘strongly agree’ were categorized as ‘yes.’ Similarly, responses on the scale of ‘not necessary,’ ‘slightly necessary,’ and ‘somewhat necessary’ were grouped as ‘not necessary,’ while ‘very necessary’ and ‘absolutely necessary’ were categorized as ‘necessary.’ Additionally, we reported the mean response values along with the standard deviation.

Responses from physicians without prior experience with in vivo training options and vascular silicon simulators were excluded from the analysis of preference and necessity for such training options. Similarly, responses from physicians who had never assisted during a neuroendovascular intervention were excluded regarding the realism of simulations and the necessity rating of clinical training on humans. We used the McNemar test, given the paired nature of the data, to investigate if there was a significant difference between the necessity ratings of the various training methods. p-values with an α-level ≤ 0.05 were considered statistically significant. Statistical analysis was calculated using JASP 0.18.3 [12].

2.10. Positioning AngioSim Among Existing Training Tools

Previous work on neurointerventional training simulators has explored various approaches, including animal models, computer-based virtual reality platforms, and physical silicone models with live camera systems. Virtual simulators such as Mentice offer a high level of interactivity and performance tracking but lack haptic realism. Camera-based silicone models reproduce anatomical geometry well but fail to simulate fluoroscopic vision. In contrast, AngioSim combines real catheter manipulation with augmented-reality visualization, bridging the gap between realism and radiation-free digital overlay. A comparison of key features is summarized in

Table 1. Comparison of different neurointerventional training simulators.

Simulator Type	Real Catheter Use	Fluoroscopy- Like Imaging	Radiation-Free	Replicable Hardware	Anatomy and Vessel Material	Cost- Efficient
Animal Model	✔	✔	✖	✖	non-human; real vessel	✖
Mentice (VR Platform)	✖	✔ virtual only	✔	✖	human anatomy; virtual	✖
Camera + Silicone Model	✔	✖ camera view	✔	✔	human anatomy; silicone	✔
AngioSim	✔	✔ AR overlays	✔	✔	human anatomy; silicone	✔

.

3. Results

Twenty-four physicians, with a median experience of 1.5 years and an interquartile range from 1.0 to 4.25 years, participated in the device training. Among them, 14 (58%) physicians rated their abilities in neuroendovascular interventions as inexperienced, five (21%) as moderate and five (21%) as advanced. Twelve (50%) physicians reported performing neuroendovascular interventions, while the other twelve had not performed any interventions. Except for two physicians, the remaining 22 (92%) physicians had assisted during neuroendovascular interventions. Fourteen (58%) of the physicians had previously used a vascular silicone or computer simulator, while 11 (46%) of the 24 had practiced on in vivo models.

Among the twenty-four physicians, 92% considered the visibility of catheters during fluoroscopy mode to be sufficient, with a mean visibility rating of 4.29 ± 0.86. Similarly, 96% found the visibility of guidewires sufficient, with a mean of 4.41 ± 0.72. During RM mode, catheter visibility was rated as sufficient by 96% of physicians, with a mean of 4.29 ± 0.55, while 88% considered guidewire visibility sufficient, with a mean of 4.20 ± 0.78.

For the 22 physicians who had at least assisted in one neuroendovascular intervention, 86% rated the visibility of catheters and guidewires during the RM mode as realistic, with a mean of 4.13 ± 0.64 and 3.90 ± 0.87, respectively. Additionally, 86% found the RM simulation itself realistic, with a mean of 4.18 ± 0.8. The simulation of DSA was deemed realistic by 81% with a mean of 4.00 ± 0.87, and 76% rated the entire simulation as realistic, with a mean of 3.81 ± 0.81. Figure 5 provides a detailed breakdown of the physician’s responses concerning the perceived realism of the simulation.

The majority (92%) of physicians, with a mean of 4.50 ± 0.66, indicated they would use the AngioSim again, while all respondents (100%) found it useful for neuroendovascular intervention training, with a mean of 4.67 ± 0.48.

Among the 14 physicians with prior experience using vascular silicone simulators combined with a camera, 93% preferred the vascular silicone simulator combined with AngioSim over purely camera-based simulators, with a mean preference rating of 4.50 ± 0.65. In contrast, only 28% of the 11 physicians with experience using in vivo training preferred the AngioSim over in vivo training, with a mean preference rating of 3.09 ± 1.30. When compared to computer simulators like Mentice, 48% of physicians preferred the AngioSim, with a mean preference rating of 3.52 ± 1.04. Figure 6 provides a detailed breakdown of physicians’ preferences for different simulators, as well as their overall satisfaction with AngioSim.

Among the respondents who had previously used a vascular silicone simulator, 36% deemed the camera-based vascular silicone simulator model necessary for training in neuroendovascular procedures, with a mean rating of 3.14 ± 0.77, while 96% considered the AngioSim necessary, with a mean rating of 4.13 ± 0.46. Additionally, 55% of those with experience using animal models found them necessary, with a mean rating of 3.27 ± 1.35. 63% deemed computer models necessary, with a mean rating of 3.83 ± 0.87, and 95% of those who had assisted in a neuroendovascular intervention considered clinical training with humans essential for neuroendovascular procedure training, with a mean rating of 4.68 ± 0.57. Detailed feedback from physicians on the perceived necessity of the different simulators is presented in Figure 7.

The McNemar test revealed significant differences in the necessity ratings between the AngioSim compared to camera-based vascular silicone simulators (p = 0.003), computer models (p = 0.046), or animal models (p = 0.011), indicating that the respondents considered the AngioSim to be necessary for angiographic training. However, there was no significant difference between the necessity ratings for AngioSim and clinical training with human subjects (p = 1).

4. Discussion

This study evaluated the performance of AngioSim, a novel augmented-reality angiography simulation system combined with a vascular silicone simulator, in enhancing neurointerventional training. The results demonstrate that AngioSim provides realistic and valuable training experiences. All physicians who participated in the study (100%) rated AngioSim as useful for neurointerventional training.

Our findings showed that 93% of physicians who had prior experience with camera-based silicone simulators preferred AngioSim, suggesting that the augmented-reality features significantly improve the training experience by providing more realistic imaging without the need for ionizing radiation.

Although only 48% preferred AngioSim over computer simulators, this is still a promising result, especially considering that AngioSim is a prototype and lacks features to simulate the movement of the C-arm or table and multi-plane imaging. Notably, AngioSim was rated significantly more essential for neuroendovascular training than computer simulators (p < 0.05). This perceived necessity may stem from AngioSim’s ability to facilitate hands-on training with real devices within a realistic anatomical model, providing more authentic haptic feedback, a critical component for achieving realistic simulation experiences [13].

Our finding that 72% of the physicians favor in vivo models over AngioSim aligns with previous studies, where 77% of the physicians indicated a desire for more training time with in vivo models, compared to just 7% for the silicone model [14]. This might be due to the usage of real angiography rooms in animal experiments. Nevertheless, the necessity of the AngioSim was rated significantly higher (p < 0.05) than that of in vivo models. This could be due to the ethical concerns of medical in vivo training [15].

The lack of a significant difference in necessity ratings between AngioSim and clinical training with human subjects (p = 1) highlights its potential as a complementary tool that can prepare trainees before they perform procedures in angio-suits on patients, thereby improving patient safety and trainee confidence [16].

While AngioSim adequately displays catheters and guidewires sufficiently, 24% of participants still considered the simulation unrealistic, underscoring the need for further improvements. Future enhancements should focus on controlling the camera via a joystick, enabling camera rotation around the silicone model and incorporating a second camera for biplanar imaging. However, these upgrades may not fully resolve the issue of overlapping vascular structures, as the system relies on light-based imaging instead of the penetrating capabilities of radiation. Additionally, the grayscale display makes it challenging to distinguish between microcatheters and microwires without markers for ionizing radiation as depicted in Figure 2b and Figure 3d and to identify the exact position and orientation of other medical devices such as stent retrievers.

Due to its low cost, modular architecture and reliance on commonly available hardware, AngioSim is highly scalable and easily adaptable to various training environments. Its straightforward setup allows for implementation not only in institutional settings but also in workshops. As a radiation-free solution, it aligns well with modern simulation-based training curricula [16] and could serve as a valuable component in structured neurointerventional education programs for residents, fellows, and continuing education initiatives.

This study represents a proof-of-concept evaluation of the AngioSim system, aimed at assessing its feasibility, realism, and acceptance among physicians in a training context. Rather than focusing on technical performance or skill acquisition, we prioritized initial user impressions and perceived value. These early insights provide a foundation for future, more extensive evaluations involving objective performance metrics and multi-center participation.

The main limitation of the questionnaire-based part of the study was the relatively small sample size and the single-center design, which may have affected the generalizability of the findings. Moreover, the physicians had varying levels of experience, which could influence their perceptions and evaluations of the simulator. A multicenter study with larger cohorts and standardized assessment tools is warranted to validate these findings.

5. Conclusions

We introduced AngioSim, an augmented-reality extension for conventional silicone models simulating a wide range of X-ray-based imaging techniques used in angiographical suites for interventional procedures. AngioSim offers a valuable and affordable upgrade for any silicone vessel simulator in neurointerventional training programs, costing less than EUR 70 and offering realistic visualization for a practical training experience without the risks associated with ionizing radiation.