Metric-Based Virtual Reality Simulation
A Paradigm Shift in Training for Mechanical Thrombectomy in Acute Stroke
Ischemic stroke is the second leading cause of death and the predominant cause of long-term disability in the Western world. Until recently, the standard treatment for ischemic stroke was intravenous administration of r-tPA (recombinant tissue-type plasminogen activator) within the accepted time limit of 4.5 hours from the onset of symptoms. Given later, risks such as intracranial hemorrhage outweigh the potential benefits. In large vessel occlusion (if no contraindication), r-tPA followed by mechanical thrombectomy improves outcomes. In a meta-analysis of large multicenter trials,1 patients receiving usual care (most often r-tPA) followed by mechanical thrombectomy showed significantly higher rates of functional independence at 90 days (46%) than those receiving usual care alone (26.5%). Benefits from mechanical thrombectomy have also been demonstrated in selected patient groups ≤162 and 24 hours3 after they were last known to be well.
Despite the proven effectiveness of mechanical thrombectomy, access is limited. One of the main reasons is because of a shortage of interventional neuroradiologists trained to perform this procedure. Traditionally, doctors acquire their skills on new procedures on patients. However, image-guided procedures impose unique human factor challenges on the operator, which expose patients to potential risk during their learning curve. This current traditional process–driven approach to training4 (ie, procedure numbers done and time in training are assumed to signify skill) does not guarantee that the trained clinician has acquired the ability to effectively and readily execute as independent practitioners by the end of their training.5
A novel approach to enhance the learning experience and training of doctors to competently perform mechanical thrombectomy is metric-based virtual simulation training to proficiency.5 A group of senior interventional neuroradiologists and the ASSERT Center at University College Cork, in conjunction with industry partner Mentice AB (Gothenburg, Sweden), have pioneered this method to help prepare physicians to perform this life-changing procedure.
This vascular intervention system trainer (VIST) utilizes a physics-based high-fidelity endovascular virtual reality (VR) simulator, which enables hands-on procedural training for clinicians. This tool can objectively, consistently, and reliably quantify performance levels of trainees, in a safe environment.6 This technology allows the actual endovascular devices, such as guidewires, catheters, and thrombectomy devices, to be used for simulated endovascular procedures. The VIST VR simulator is a high-powered personal computer–based system that uses a geometric vessel representation together with physics-based calculations to determine the behavior of the endovascular devices and the blood flow in the vessels.7 The simulator (Figure [B1]; VIST; Mentice AB, Gothenburg, Sweden) senses the user’s manipulation of the real endovascular devices via the haptic device (ie, haptics is the science and engineering that deals with the sense of touch8) transfering the real movements to the virtual representation of the devices. It calculates the interaction between the virtual devices and vessels by the use of real physical properties, such as mass, stiffness, and friction. The result of the calculations determines the position and the shape of the devices and the interactions between devices and vessels. It also determines the forces applied on the devices that will be fed back to the user through the endovascular devices via the interface haptic device (aka force feedback device). This emulates the tactile feedback experienced by the user in vivo. The result of the user’s actions is calculated in real-time, fed back instantly to them via the haptic device, and displayed on the virtual fluoroscopy screen (shown in Figure [B1 through B3]). The computer-generated simulation can thus measure precisely and reliably the performance of the user.
A performance characterization of mechanical thrombectomy, derived from experienced interventional neuroradiologists, was utilized to establish procedure metrics. Once validated, these metrics were used to establish a performance benchmark—the level of proficiency—that trainees must demonstrate before progressing to the next level of training or real patients. This approach (proficiency-based progression) ensures a more homogeneous skill set in graduating trainees and can be applied to any level of training. Prospective, randomized, and blinded clinical studies have demonstrated that trainees who acquired their skills to a level of proficiency on a simulator perform significantly better (40%–69% better) in vivo in comparison with their traditionally trained colleagues.9 VR simulation has also been used to retrain experienced physicians in the performance of high-risk procedures novel to them.7 VIST also allows for the implementation of neurointerventional societies’ recommendations on state-of-the-art thrombectomy performance guidelines.
High-fidelity simulation affords the opportunity in replicated real-world scenarios, to provide step-by-step guidance and metric-based feedback to the trainee throughout the procedure. Figure (A2) demonstrates a 75-year-old previously healthy man who experienced an ischemic stroke at 8:00 PM and arrived at the primary stroke unit 1 hour later. A computed tomography angiography was performed and intravenous r-tPA administered at 10 PM.
A right-sided occlusion of the internal carotid artery at its cervical origin and a left-sided middle cerebral artery occlusion without early signs of infarction was interpreted from the computed tomography (CT) data. The patient was transferred to a tertiary center for thrombectomy, arriving at 11:15 PM, and procedure start at 11:30 PM. Thrombectomy was successfully performed with 2 different types of stentrievers within a total procedure time of 45 minutes with near-total reperfusion (Thrombolysis in Cerebral Infarction 2b; Figure [A3]). However, the first control CT showed contrast enhancement and some swelling of the left hemisphere that corresponded to the frontal middle cerebral artery territory.
Figure (B2 and B3) shows virtual representation of the identical same patient anatomy as described above (ie, Figure [A2 and A3]). A CT digital imaging and communications in medicine angiography scan from the real patient—a high-resolution 2-dimensional gray scale images were used to extract the important anatomic structures for the virtual simulation and to create a VIST-specific 3D representation of the patient (Figure [B2 and B3]). The concordance between the physical behavior of the devices and visual appearance of the anatomy, in vivo and in the VR simulator, is high.
The developments (outlined above) in physics-based simulation provide/offer the use of real patient cases to augment the training of high-skill and high-risk procedures, such as mechanical thrombectomy. Furthermore, metric-based characterization of optimal and suboptimal performance, derived from proficient/master interventionists means that trainees can be given objective, transparent, and fair quantitative feedback on their performance, and this training can be repeated over and over and is standardized.8 The simulator and the validated metrics can in addition be used to quantitatively define a proficiency benchmark based on the performance of experienced practitioners.5,10 Trainees would not progress to performing the procedure on real patients until they had demonstrated this performance level, consistently. Evidence from prospective and randomized clinical trials has demonstrated that this approach to training produces a much more homogeneous skill set that translates into significantly improved intraoperative performance.7,9 It has, however, also been demonstrated that the exact same simulation training without the validated performance metrics achieved a training outcome only marginally better than traditional training (ie, standard training, 29%, versus simulation training, 36%, versus proficiency-based progression simulation training, 75%, demonstrated the proficiency benchmark at the end of training).9 This approach to training will not replace clinical in vivo training, but it has the potential to supplant a significant part of the learning curve on real patients. The trainee will have learned to perform the procedure on the simulator to a quantitatively defined performance level (defined on the performance of experienced and proficient practitioners) but not performed on a real patient. This means that the consultant/attending will in future be teaching a pretrained novice in the intervention suite.5,8,10 Training in this way is more than an interesting educational experience.8 Trainers will know precisely the performance level of their trainee if they have successfully navigated a proficiency-based progression simulation training curriculum.4
High-fidelity, veridical VR simulations afford operators the opportunity to acquire and maintain new skills. It also permits expansion of their procedure experience because they can engage in deliberate practice rather than simply repeated practice. Crucially, this capability presumes the truthful nature of the simulation training cases. The trainee can accurately experience the span of appropriate sensory responses and physical actions that are consistent with what would be experienced in real life. This includes the opportunity to enact both appropriate and inappropriate actions and receive performance feedback that accurately depicts their performance level. VR simulation training programs allow professionals of all levels to increase proficiency in the detailed steps for a given procedure, as well as an awareness of the potential pitfalls and crucial moments in a safe environment.
Physics-based simulations do have difficulties; they are expensive because they run on high-end personal computer platforms and have associated interface platforms, which are also not inexpensive (Figure [B1]). With the development of sensor technologies, the simulations run considerably more stably than they did a decade ago but not if used incorrectly or inappropriately (ie, the user should not do anything with thrombectomy devices they would not do in a real patient). Real patient cases derived from the CT or magnetic resonance imaging digital imaging and communications in Medicine data take 2 to 5 days work by a computer engineer/scientist to produce a simulation of sufficient fidelity that would be acceptable for a proficiency-based progression simulation training curriculum. These issues aside, physics-based VR simulation for mechanical thrombectomy and other high-risk and neurovascular procedures is a game changer. It affords the trainee (no matter how senior) the opportunity to acquire and hone their skills outside the intervention suite to a quality-assured performance level using real-time metric-based performance feedback operating with the exact same devices, in the same order as they would on a patient.
It is feasible and practical to take CT digital imaging and communications in Medicine data from a real stroke case and recreate it in VR. Metric-based simulation training can supplant parts of the learning curve for endovascular procedures, with consequently safer operations, particularly for high-skill, high-risk procedures, such as mechanical thrombectomy for ischemic stroke. It should not be viewed as a replacement for traditional fellowship training but as a powerful tool to reproducibly augment the training experience when fully integrated into the current in vivo education curriculum. This approach to training is conceptually and intellectually appealing, and it represents a paradigm shift in how doctors are educated and trained. Training for these procedures must be more than an interesting educational experience.
All of the authors contributed to the writing of the article. Drs Liebig and Holtmannspötter and R. Crossley are experienced mechanical thrombectomy interventional neuroradiologists. Dr Lönn and P. Henn are experienced clinicians and medical education specialists. J. Lindkvist is a senior computer engineer. Dr Gallagher developed and helped validate proficiency-based progression simulation training. Dr Liebig, Dr Holtmannspötter, R. Crossley, Dr Lönn, J. Lindkvist, and Dr Gallagher have characterized a reference approach to mechanical thrombectomy. J. Lindkvist developed and formatted the virtual reality simulation from the computed tomographic angiography of the stroke patients reported in this article. Dr Gallagher and P. Henn produced the first draft of this article. All authors contributed to editing the paper post-editorial review.
The research and researchers on this article were supported by a grant from the Swedish government agency for innovation (Vinnova) to Mentice AB (Gothenburg, Sweden) to characterize, develop, and then validate the metrics for a reference approach to performance of mechanical thrombectomy. Dr Holtmannspötter has received honoraria from Microvention, Medtronic Neurovascular, Mentice AB, and Stryker Neurovascular for consulting and proctoring. R. Crossley has received honorarium for speaking (Stryker Neurovascular, United Kingdom) and educational sponsorship to attend meetings/conferences from Microvention, Stryker, Medtronic, Penumbra, Johnson & Johnson. J. Lindkvist works as an engineer at Mentice and developed the virtual reality model of the real patient data. Dr Lönn has served as a clinical advisor and then as medical director for Mentice.
- Received February 6, 2018.
- Revision received April 29, 2018.
- Accepted May 9, 2018.
- © 2018 American Heart Association, Inc.
- Goyal M,
- Menon BK,
- van Zwam WH,
- Dippel DW,
- Mitchell PJ,
- Demchuk AM,
- et al
- Nogueira RG,
- Jadhav AP,
- Haussen DC,
- Bonafe A,
- Budzik RF,
- Bhuva P,
- et al
- Patel AD,
- Gallagher AG,
- Nicholson WJ,
- Cates CU
- Cates CU,
- Lönn L,
- Gallagher AG
- Gallagher AG,
- O’Sullivan GC
- Angelo RL,
- Ryu RK,
- Pedowitz RA,
- Beach W,
- Burns J,
- Dodds J,
- et al