HANNES: A Modular Neurointerventional Training Model

Neurovascular diseases are conditions affecting the blood vessels and blood supply to the brain or spinal cord, including strokes, vascular structural anomalies, and pathological vessel narrowing. A stroke is defined by an abrupt onset of neurological symptoms. It can be characterized as an ischemic type if caused by a vessel occlusion or hemorrhagic, in the case of intracranial bleeding [1]. According to the most recent Global Burden of Disease estimates, there were around 12.2 million incident cases of stroke, 143 million disability-adjusted life-years lost, and 6.6 million deaths globally in 2019, making stroke the second leading cause of death and the third leading cause of disability worldwide [2]. Ischemic stroke comprise 85% of stroke casualties [3]. Stenoses, or the pathological narrowing of vessels, can increase the risk of infarction. Brain hemorrhages can be a complication of ischemic strokes or a result of a vessel rupture, e.g., in the case of intracranial aneurysms. Intracranial aneurysms are estimated to be found in up to 3% of the general population [4]. These are abnormal bulges in the blood vessel wall that are filled with blood and cause a weak spot in the vessel wall [5]. Rupture of the aneurysm wall can cause bleeding into the cerebrospinal fluid or even within the brain, which is a serious life-threatening condition [3] with mortality rates of 25–45% within the first week, whereas 12% die immediately after onset [6]. Therefore, the prevention and treatment of neurovascular disease are of great importance. Treatments vary depending on the specific disease, ranging from drug therapy and minimally invasive endovascular procedures to open neurosurgery. Prompt and effective treatment can help reduce the impact of the disease and further complications and improve the quality of life of those affected.

As neuroradiological interventional procedures become increasingly common, medical professionals face several challenges that must be overcome. The minimally invasive treatment of endovascular diseases is a rapidly evolving area that involves a wide variety of neurointerventional instruments, making it necessary for neuroradiologists to have not only anatomic and clinical knowledge but also an understanding of the devices and angiography systems used [3]. To meet these challenges, medical personnel need to undergo simulation training on the different conditions, using the increasingly wide array of instruments and devices available [7].

Currently, animal models such as pigs or rabbits are frequently used for neurointerventional treatment simulation training, despite ethical criticisms and deviations from human anatomy [8]. As a result, artificial in vitro treatment simulations are preferred. However, there is currently no suitable technical solution for a modular simulation environment that can simulate various cerebral vascular diseases, such as aneurysms, stenoses, or strokes, for structured training or product testing [9]. Therefore, a physical simulation model that can simulate different neurovascular disease patterns with original instruments and be adapted to patient-specific conditions is needed.

Nawka et al. have reported that the majority of interventional neuroradiologists have undergone animal-based training [8]. Training in neurointervention has been previously routinely performed using animal models such as dogs, rabbits, rodents, or swine. Different clinical scenarios are being targeted by different animal models. Aneurysms, for instance, may be simulated using a variety of physiologic and pathologic manipulations—surgical constructions, carotid ligation, renovascular hypertension, and lathyrism [10]. Limitations of these methods are the small size of vascular structures and the inherent access anatomical differences, rendering a true-to-life experience in an animal model unlikely. Further challenges rely on the poor training standardization and reproducibility. However, various in vitro or in silico models have been developed to replace animal experiments in training, allowing for the practice of treating pathologies. Although virtual reality (VR) simulators allow animal- and radiation-free training, they are often criticized as they do not provide realistic haptic feedback at the catheter or simulate the behavior of micromaterials realistically [11,12]. In addition, VR simulators cannot be used to test new treatment instruments since the training is performed using generic reusable instruments [13]. Other drawbacks associated with VR models are their costliness [14–16], susceptibility to technical failures, and the need for frequent calibration and maintenance [15]. Therefore, various physical models have been developed [17–20].

For instance, Vascular Simulations (New York) offers a physical replication system that allows the practice of endovascular procedures [21]. The model represents the vascular tree of the upper body, including the head, and a realistic heart model and valves that simulate the pulsatile blood pressure [21]. The head is made of polyurethane and filled with a gel [21]. However, changing the aneurysm model for other patient models is complicated and time-consuming. The EndoVascular Evaluator model of FAIN-Biomedical, Inc. (Nagoya, Japan) [22] enables endovascular surgery training and simulation using patented modeling technology to generate customized human vessel models based on computed tomography (CT) or magnetic resonance imaging data. The model allows the simulation of diverse vascular diseases, such as the treatment of left middle cerebral artery occlusion by mechanical thrombectomy [22]. However, most models do not offer the possibility of being adapted to different anatomies. The presented models provide the possibility of carrying out trainings without animals. Nevertheless, their respective similarity to real treatment varies, such as in the replication of the whole vessel tree without inner edges. Currently, no modular physical model has a complete vascular system and can be adapted to the individual skill level of the participants. Therefore, a new modular neurointerventional training model is needed to improve realistic treatment training and replace animal models in the long term.

While developing the new neurointerventional simulator, a close collaboration was established in an interdisciplinary team of physicians and engineers. With physicians' medical backgrounds and minimally invasive treatment experience, relevant anatomical replications are tried to be included as realistically as possible in the model. The technical knowledge of the involved engineers helps develop modular and technical solutions. The methodical experience resulted in a product family for the neurointerventional training. The developmental steps and their results are described below.

To develop a new simulator, the “integrated PKT-approach for developing modular product families” (see Ref. [23] for further information about the integrated PKT-approach) is used. This approach helps to capture the extensive external variety of the model and provides an acceptable internal variety. This means that several variants of the simulator HANNES (Hamburg ANatomical NEurointerventional Simulator) can be created with a reduced number of components. The training model focuses on the patient-specific and patient-based design of the components. This individualization can be considered at the conceptual design stage and is incorporated into the process of design-methodical model development. For this purpose, the methodological approach for design for mass adaptation by Spallek et al. [9] was used. Allowing a future-proof concept that is suitable for individualization steps in the future.

First, the variant and basic requirements are recorded, and the functions and subfunctions are determined. Subsequently, HANNES is divided into modular structures, considering the patient-specific individualization of various features.

To create a new neurointerventional training model, a close interdisciplinary collaboration between engineers and physicians was established. This collaboration involved interviewing physicians to identify the general and specific requirements for different treatment trainings. Additionally, direct observations of real interventions were conducted to document and analyze treatment workflows, supplemented by questionnaires and free discussions. Insights gained from the development and use of version 1.0 of HANNES [9] were also utilized.

The model should replicate the entire relevant vessel tree so that both radial and femoral interventional treatments can be trained. This includes the femoral and iliac arteries, the aorta, the arm arteries, and cervical and intracranial vessels.

A modular design of the simulator should allow adaptions for different training scenarios. These should include the treatment of pathologies like aneurysms, internal carotid artery (ICA)-stenosis, and thrombus-induced vessel occlusions. For this, the patient-based anatomical models should be easily and quickly exchangeable, facilitating adaption to patient-specific conditions. Furthermore, original treatment instruments should be used to create a highly realistic training environment. Considering the 3R principles (replacement, reduction, and refinement), the training model must be animal-free, ensuring ethical standards are met. The blood-simulating fluid should flow pulsatile and present a physiological temperature. The whole simulator should fit into an angiographic system. It should support the education and training of medical professionals and provide a platform for medical device testing so that they can be developed and optimized further.

Identified requirements were prioritized and categorized into wishes and mandatory requirements. This approach allowed for ranking the importance of these requirements, ensuring that the most critical ones were prioritized in the event of a need to balance or limit the model's scope.

Once the requirements are clear, a conceptual design phase is conducted. A function structure is established to outline the main and subfunctions necessary for the training model. This structure helps identify different working principles for each subfunction and allows for a systematic approach to the design process. Figure 1 shows an excerpt from the hierarchical function structure explicitly developed for the HANNES model. In this structure, the primary goal of “enabling neurointerventional training” is decomposed into several subfunctions, such as “patient mapping,” “replicate angiographic representation,” and “perform endovascular procedures.” Each subfunction is further broken down into more specific tasks, such as “replicate vascular system,” “map landmarks,” and “insert treatment instruments.”

A variant-specific product architecture with standardized and variant modules is developed to make the model usable for various training situations (described in detail in section “Modular Product Architecture of HANNES”). Modules consist of one or more components that are combined based on technical functional relations or strategic aspects. The combination of modules creates a simulator for multiple training scenarios. During the design of the training setup, aspects and methods of complexity management are implemented using the integrated PKT-approach [8]. This approach helps achieve a manageable internal complexity while providing high external variety, ensuring that the training model can be adapted to different training requirements and scenarios.

The design of the patient-based vessel components occurs in a standardized individualization process introduced by Spallek et al. [24,25] and adapted in Ref. [9]. This process describes the pathway from medical data acquisition through the design of specifications and production of the model until its application, exemplarily for an aneurysm. The standardized process for developing patient-specific models is shown in Fig. 2.

In the first step, the imaging data are acquired through a three-dimensional (3D) angiography scan. Afterwards, the data are segmented, and relevant vessels are retained. Hereafter, the model specifications take place, like the adapters as the interface for integrating the model in the training simulator or the definition of the wall thickness. After finishing this step, the aneurysm model is manufactured by additive manufacturing (“3D-printing”). The significant advantage of these technologies is tremendous geometrical freedom combined with high production flexibility, especially for reduced lot sizes [26]. Many materials and technologies are available for synthetic fabrication, which differ especially in material properties, e.g., elasticity, transparency, and the feasibility of support removal in the inner hollow vascular structure. Therefore, a comprehensive comparison of additive manufacturing technologies for the replication of vessel models was conducted in Ref. [27]. Models fabricated by fused-deposition modeling and stereolithography (SLA) technology already offer a replication with surprisingly high accuracy [26,28,29].

For vessel components, a flexible and translucent material is required. Additionally, a smooth and edge-free inner surface needs to be achieved. This allows treatment training to be performed without obstacles to the catheter as it is advanced into the vessel models.

The simulation model is designed in a modular way and consists of standard and variant components. Thanks to its modular product architecture, see Fig. 3, the model can be easily extended and supplemented. The current version of HANNES (version 2.0) can be seen in Fig. 4. This version can be adapted for the endovascular treatment of ICA-stenosis, ischemic stroke through large vessel occlusion, and intracranial aneurysms.

Accordingly, the blood vessel tree includes all arterial vessel sections catheterized during radial and femoral neurointerventions. The upper body and pelvic vessels are partially embedded within a base frame and are securely held in place by it. The inside of the base frame contains the technical fluid system, which is connected to the blood vessel tree and provides the system with a physiological pulsatile blood flow. The data processing required for the physiological blood flow is handled by a control unit coupled to the base frame. On the side of the base frame facing away from the control box, the posterior and anterior neck vessels protrude beyond the base frame and transition into a head model. A removable skull base serves as the interface between the head, neck, and cerebral vessel components.

The blood vessel tree is divided into several components, each connected to the others via standardized interfaces. Several variants exist for the aortic arch, whole brain model, left and right carotid, and vertebral components. Thus, the simulation model can be adapted to different anatomical and pathologic scenarios. Apart from the variant vessel components, the blood vessel tree also includes standard components. These include the skull base, the arm and pelvic vessels, and the abdominal and thoracic sections of the aorta.

The product platform of the simulation model includes the base frame, the technical fluid system, and the control unit. These are required for every training and thus form the basis of the entire training model.

The base frame is a structure based on the shape of the human torso. Inside the base frame is the technical fluid system, which is central to providing near-physiological circulation. The base frame also acts as a support for the upper body vessels that are integrated in HANNES.

The outer shell of the base frame consists of several polypropylene sheets, which were screwed together by angle brackets, screws, and nuts. This design gives the base frame stability while ensuring cost-effective production. In addition to the polypropylene sheets, the base frame contains several additively manufactured components, such as two arm profiles. These contribute to the humanoid design of the base frame.

For maintenance purposes, the base frame has four maintenance flaps through which access to the technical fluid system is possible. This allows necessary maintenance work, such as filter replacement, to be carried out. A removable tank cover provides access to the technical fluid system's main tank.

A plug-in connector connects the base frame to the control unit. This connection provides a mechanical coupling and enables data and energy transmission. The transmitted data include readings from flow, pressure, and temperature sensors. To prevent artifacts in the X-ray imaging, the base frame contains a metal-free section that starts at the height of the aortic arch and extends distally.

In HANNES, the technical fluid system is responsible for providing near-physiological flow. It is located inside the base frame and has several main components: a main tank, an adjustable pump, a proportional valve, and various sensors to monitor the flow parameters. Figure 5 shows the basic structure of the technical fluid system in a simplified form.

During training, the integrated stainless steel main tank is filled with approximately 6.5 l of water and 25 ml of shampoo (PENATEN Shampoo by Johnson & Johnson GmbH, Neuss, Germany). The water level is continuously monitored by a capacitive level sensor and an additional float switch (RSF73H100RN by Cynergy3 Components Ltd., Wimborne, UK). The heating rod installed in the tank heats the fluid to the predefined body temperature of 37 °C. A temperature sensor continuously measures the current temperature. An adjustable centrifugal pump connected to the main tank delivers the tank water to the ascending section of the aortic arch. The pump speed can be adjusted as needed but usually remains constant. A near-physiological pulsation is achieved by periodically opening and closing a proportional valve, whereby a defined volume is returned to the main tank. In parallel with the proportional valve is a safety valve which opens automatically in the event of overpressure.

The pump delivers water to the ascending branch of the aortic arch, which supplies water to all the vessels of the upper body, including the head and neck. A portion of the water leaves the torso vessels at the brachial and renal arteries and returns to the main tank (see Fig. 5). If necessary, this water flow can be throttled by a servo-valve to influence the intracranial blood pressure.

A pressure sensor measures the pressure of the water flowing into the aortic arch. The water flows from the torso vessels into the head and neck vessels and continues, depending on the connected cerebral vessels, through up to six tubes back to the main tank. In each case, the fluid flow passes through a ball valve, a filter, and a flow sensor. Any leakage occurring within the skull is captured and collected in an external head tank. An integrated float switch detects a filled head tank. If required, the head tank can be drained via a ball valve.

The total intracranial return flow typically ranges from about 1.2 to 1.8 l/min, corresponding to an average flow rate of approximately 0.2 to 0.3 l/min per vessel branch. The specific distribution can vary significantly, particularly during stroke training, due to the introduction of stenoses and thrombi (see section “Blood Vessel Tree and Training Scenarios” for more details). A complete vessel occlusion results in a drop in flow rate to 0 l/min due to the absence of a venous system and collateral vessels.

The pump output, along with the servo and proportional valve settings, is configured to ensure that the pressure sensor proximal to the aortic arch measures a pressure between 120 and 70 mm Hg at a pulse rate of 60 beats per minute. Once set, the parameters of the technical fluid system are not further adjusted intra-operatively. Experience from training sessions with HANNES has shown that specific pressure values are not necessarily required for training purposes, as exercises could be successfully performed even with nonphysiological pressure values without participants, even those with significant experience, noticing the difference. In contrast, the pulsatile pressure and flow behavior appear to be highly relevant, as they directly influence the behavior of the contrast medium during digital subtraction angiography.

The control unit controls and reads out the electrical components installed in HANNES. These include the pressure and flow sensors as well as the pump, heating rod, and servo and proportional valve. To read data or make changes, the user can access a control panel permanently attached to the control unit. Within the standard menu of the control panel, the user can toggle the pump and pulse functions on and off and access the current sensor data. An expert menu allows for adjusting additional settings, such as the proportional–integral–derivative control parameters for the heating rod and the pressure control. For external access to the control unit, HANNES is equipped with its own Android tablet (Samsung Galaxy Tab A7, Samsung Electronics Co., Ltd., Suwon, South Korea) that is connected to the control unit via WIFI. The application software, designed and programed for this purpose, allows the visualization of data from the flow and pressure sensors and the adjustment of key circuit parameters (see Fig. 6(a)). The sensor data can be saved and exported as a comma-separated values file via WIFI. In addition, a detailed multimedia user manual (see Fig. 6(b)) is available on the tablet as a supplement to the written user manual for the simulation model. The main objective of the multimedia user manual is to provide the user with interactive support in the form of videos, photographs, and illustrations for the technical activities relevant to the training.

The head module is essentially an additively manufactured replica of the human skull and the upper three vertebrae of the cervical spine [30]. It connects and fixes the anterior cervical vessels to the intracranial vascular models of the blood vessel tree. In addition, the head module supports the trainings performed on HANNES by making the replicated bones visible in the X-ray imaging, thus serving as an orientation aid. To ensure that the outlines of the bones are visible in the X-ray imaging, the bone components were manufactured in two steps. First, the components were fabricated using the SLA process on a Form 3 (Formlabs, Inc., Somerville, MA) 3D printer from the material Clear V4 (Formlabs, Inc.) and then manually wrapped in plaster fabric (Cellona, Lohmann & Rauscher International GmbH & Co. KG, Neuwied, Germany). Due to the high radiopaque density of the plaster fabric, the bone contours are visible in the X-ray imaging.

The head module was designed in catia (Dassault Systèmes, Vélizy-Villacoublay, France) using a Standard Tesselation Language model obtained from a CT scan sourced from the embodi3D.com online database [30]. The anatomical features of the head module are based on a specific patient CT scan and, therefore, do not represent an average anatomy. For ease of use, it was decided to reduce the number of components of the head module by connecting individual bones, adding new subdivisions, and defining new components accordingly. The head module consists of five bone components and a skull holder. The bones have been divided into an anterior and posterior skull half, a skull base, a cervical spine, and a mandible. Each component is equipped with interfaces and can be connected to each other. The skull base component connects the anterior and posterior skull halves. In addition, the skull base has interfaces on both sides through which the anterior neck vessels can be connected to the intracranial vessel models. To guide the reflux tubes attached to the cerebral vessel models out of the skull's interior, the posterior skull's lower half has several recesses. During assembly and disassembly of the vessel models, it is often unavoidable that a small amount of water will enter the posterior half of the skull. To prevent the water from interfering with the imaging, the skull's posterior half has a drain that allows any leakage to flow directly into the external head tank.

The blood vessel tree of the simulation model has a modular structure and, depending on the application, consists of up to 13 individual components. These components can be interconnected through standardized interfaces. Two standardized vessel interfaces have been developed, each enabling edge- and shoulder-free connection of vessel components. A two-part bayonet connector (see Fig. 7 right) allows the connection of vessel components with small inner diameters. To connect components with large vessel diameters, a clamp connector (see Fig. 7 left) has been developed [31].

The vessels of the upper body include all blood vessel sections explored during transradial and transfemoral neurointerventions. This includes, in particular, the ascending and descending aorta down to the femoral artery, the arm arteries, as well as the left and right carotid and vertebral arteries.

The development of the models was based on the systematic approach described by Spallek et al. [9] for the development of additively manufactured vascular models. First, multiple CT scans were acquired to obtain original patient data. Subsequently, the data were analyzed, and the vessels were segmented. The segmented vessel geometry was then transferred to catia. In catia, the vessel models were designed based on the previously segmented vessel geometries. Subsequently, the segmented vessel was reconstructed in catia, provided with a wall thickness of 2 mm and enhanced with standardized interfaces. This wall thickness was iteratively developed to compromise mechanical stability and a material-saving design. The additive manufacturing was prepared in the software preform (Formlabs, Inc.). Since internal support structures are difficult or impossible to remove while keeping lumen patency, special care was taken to ensure that the models were only provided with support structures from the outside. For the manufacturing, SLA 3D printers from Formlabs were used. The material used was Flexible 80a (Formlabs, Inc.). This material was chosen based on a previous study that evaluated the haptic quality, particularly the friction characteristics, of additively manufactured blood vessel models [32]. Following production, the models were washed, cured, and postprocessed.

The aorta was divided into three components: the femoral aorta, abdominal aorta, and aortic arch. Custom-made clamp shell adapters are used to link the individual components of the aorta together. In order to make the training on HANNES extensive and diverse, several variants have been developed for individual vessel components. Three variants were developed for the aortic arch segment, each representing one of the basic three known aortic arch types (I–III). Additionally, another variant was developed to include a bovine arch, which means that the brachiocephalic trunk and left-side common carotid artery share the same branch from the aortic arch [33]. Thus, there are a total of four different vascular variants for the aortic arch segment.

The left and right vertebral and carotid arteries were defined as individual components, branching off from the aortic arch component. Each component is coupled to the aortic arch via standardized bayonet connectors. Bayonet connectors connect the left and right carotid arteries to the skull base. The vertebral arteries can be connected to intracranial models using bayonet connectors. Several variants are available for these neck vessels, such as elongated and nonelongated vessels, allowing for the simulation of different anatomical challenges. For instance, elongated vessels can simulate scenarios requiring more advanced navigation techniques.

Intracranially, two whole brain models and more than 30 different aneurysm models are available, which can be connected to the skull base component. In contrast to the patient-based components, the intracranial vessel components are patient-specific models with patient-original geometries. The treatment complexity can be adjusted through various factors, such as vessel curvature, lesion size, and position. For example, complex scenarios might involve more tortuous vessels, larger or more distally located lesions, or severe occlusions. These variations help to tailor the training to match the individual abilities of the trainees, allowing them to progress from less complex to more challenging scenarios.

Thanks to its modular design, the simulation model is versatile and can be used for different training scenarios. The treatment of cerebral aneurysms can be trained on over 30 different aneurysm models (see Fig. 8(a)). Training participants can practice endovascular treatment on HANNES using original coils, flow diverters, and intrasaccular devices such as woven endobridge devices (WEB, Microvention, Inc., Aliso Viejo, CA). The success of the treatment can then be realistically validated using fluoroscopy (see Figs. 8(b) and 8(c)).

In addition to aneurysm treatment, HANNES can also be used to train the treatment of ischemic strokes. This includes the treatment of stenosis and cerebral vascular occlusions caused by thrombi. Carotid artery stenosis can be simulated using clamps attached to the outside of the carotid arteries (see Fig. 9(a)) [34]. The constriction of the vessel can then be treated by percutaneous transluminal angioplasty, which causes the clamping shells to fall loose from the vessel. The success of the treatment can again be validated using fluoroscopy (see Fig. 9(a)). To simulate vessel occlusions, synthetic thrombi have been developed, which can be inserted into the blood vessels [35–38]. This makes it possible to train mechanical thrombectomy techniques. Clot removal can be confirmed using fluoroscopy (see Fig. 9(b)).

The transradial training includes the entire treatment procedure, including the application of percutaneous arterial access using the Seldinger technique [39]. In contrast, permanently installed accesses are available for transfemoral training. This allows an immediate insertion of treatment instruments. Furthermore, different variants of the aortic arch can be used to train the neck vessel catheterization procedures, including the necessary configurations and turning maneuvers in the aortic arch.

To facilitate the ongoing evaluation and iterative improvement of HANNES, regular training sessions have been conducted at the University Medical Center Hamburg-Eppendorf since 2016, see Fig. 10. Since 2022, these sessions have been conducted using version 2.0 of HANNES. Over time, the scope of training offerings has expanded, currently encompassing courses tailored for physicians in the transfemoral treatment of aneurysms and ischemic strokes, as well as recently introduced training sessions focusing on transradial catheterization (cf. Fig. 10). Additionally, courses are conducted to train radiologic technologists in stroke transfemoral treatment. Each training course involves up to six participants. The global COVID-19 pandemic necessitated significant temporary restrictions on the training program, which can be seen in a significant decrease in training courses (cf. Fig. 10).

In a survey involving a total of 13 participants who took part in the transfemoral aneurysm training, the realism of the anatomy in the simulation model was rated with a median of 2 (mean 2.2, standard deviation 1.0) on a scale from 1 (very realistic) to 5 (completely unrealistic). Participants had an average of 7 years of professional experience in neurointervention. Three participants had previous experience with animal models. Additionally, participants rated the design of the simulation model with a median of 1 (mean 1.7, standard deviation 1.3), where 1 represents the best and 5 the worst rating. Regarding the question of whether the HANNES setup is suitable for improving technical skills before treating patients, respondents rated this with a median score of 1 (mean 1.9, standard deviation 1.4) on a scale from 1 (completely suitable) to 5 (not suitable).

The testing was conducted using a transfemoral approach, wherein participants catheterized and treated various intracranial aneurysm models. Each training session incorporated a type II aortic arch setup. X-ray and fluoroscopic imaging were utilized throughout the procedures to ensure realistic visualization. Additionally, contrast agents were employed to enhance the imaging and accurately simulate real-life clinical conditions. Multiple aneurysm models were used during the training, offering a broad range of treatment scenarios. This allowed participants to experience and manage different levels of complexity, providing them with a comprehensive training experience across a wide spectrum of potential cases.

The HANNES simulation model enables a comprehensive approach for the simulation of neurointerventional, ultimately aiming to increase patient safety. HANNES addresses key factors influencing the success of neurointerventional treatment, including individual patient characteristics, medical training and expertise, and the technical aspects of medical devices used during procedures.

By replicating various disease patterns, pathologies, and anatomies, HANNES reflects the diverse patient characteristics that clinicians may encounter. A survey of 13 participants in the aneurysm training reveals positive evaluations of the HANNES model. Anatomy realism received a median rating of 2 (mean 2.2), design a median of 1 (mean 1.7), and suitability for improving technical skills a median of 1 (mean 1.9). These results suggest that HANNES effectively enhances technical skills in neurointervention and is well-received by participants. Investigations of HANNES version 1.0 have already shown that objective procedural metrics correlate with operator experience, suggesting that the system could help assess operator proficiency [40]. Similar studies should also be conducted on HANNES version 2.0 to further validate and improve the system.

Currently, HANNES provides regular education and training to medical staff, enhancing their skills and expertise. In addition, the model allows for product testing and evaluation of established or newly developed instruments, contributing to the technical improvement of medical devices and their safety in the neurointerventional setting.

While HANNES is an innovative approach, it is not yet used for all clinician training. Local clinical centers have a high demand for neurointerventional training due to the increasing prevalence of neurovascular diseases. The potential of HANNES to fulfill this demand is significant. If widely adopted, HANNES could provide comprehensive training for all clinicians specializing in neurointerventional procedures. Further studies are needed to quantify the exact number of clinicians requiring this type of training and to evaluate the scalability of HANNES in different clinical settings.

By providing a realistic and controlled environment, HANNES allows trainees to practice and refine their skills without the pressures and risks of patient care. The model will be iterated and refined to enhance its realism and usability. Future work will expand the range of training scenarios and further validate the model's effectiveness in improving clinical outcomes. Incorporating feedback from training participants will continuously enhance the realism and educational value of HANNES. An important aspect of future research will be the quantification of technical skills acquired through HANNES training. This will involve developing metrics and methodologies to objectively measure skill improvement and training effectiveness, providing valuable data for further refinement and validation of the model.

We extend our sincere appreciation to the Central Research Workshop Electrical Engineering at Hamburg University of Technology for their invaluable support.

• COSY-SMILE-2 — Completely Synthetic Stroke Model for Interventional Development 2, funded by the German Federal Ministry of Education and Research—BMBF (Grant No. 16LW0165K; Funder ID: 10.13039/501100002347).

The datasets generated and supporting the findings of this article are obtainable from the corresponding author upon reasonable request.