Can virtual reality improve traditional anatomical education?: A randomized controlled trial on use of 3D skull model

Background Anatomy teaching is trending towards a mixture of lectures, cadaveric models, 2D atlas and computer simulations. This paper presents a study which compare the educational effectiveness of virtual reality (VR) skull model with that of cadaveric skulls and atlas. Methods A randomized controlled study with 73 medical students was carried out with three different groups: VR skull (N = 25), cadaveric skull (N = 25) and atlas (N = 23). Anatomical structures were taught through an introductory lecture and a model-based learning. All students completed the pre- and post-intervention test, which is composed of a theory test and an identification test. Results et al. How spatial abilities and dynamic

computer simulations 3 . The miniaturization and improvement of the processing capacity of computers has caused VR to become a feasible technology in our society 4,5 . VR provides participants with an interactive experience within an immersive environment in which faithful representation of almost any objects. In addition, its ability to make the skin, organs and muscles semitransparent as well as to repeat the operative process will facilitate in the understanding of complex anatomical structure 6 .
Theoretical lecture and cadaveric dissection are often conducted in a centralized setting due to time and space constraints. Students can use the selected 3D model for self-learning through VR technology at their free time, which will be a strong supplementary in medical education. Previous studies evaluating VR teaching models in anatomical education have yielded mixed results. 3D models of carpal bones 7 − 9 , shoulder 10 and brain 11 didn't show superiority over textbook in theoretical test. While studies in temporal bone 12 , biliary system 13 and bronchus 14 showed a clear benefit of VR teaching. As education is moving towards evidence-based practice 15 , randomized controlled trials are required to provide evidence from different aspects.
Structure of skull is always one of the most complicated areas of anatomy. In order to take advantage of the VR, we constructed a colored and detachable skull model in a simulated classroom. In this randomized controlled trial, we combined the VR simulation learning with theoretical lecture, and compare the educational effectiveness of virtual reality (VR) skull model with that of cadaveric skulls and atlas. Efficacy in terms of anatomical knowledge memorization, structure identification and subjective evaluation by the participants were recorded.

Material And Method Skull model based on VR technology
We reconstructed a VR model of skull anatomy based on the previous cadaver skull information 16 that contained normal skull structures (Fig. 1). The STL (STereoLithography) file were converted into a MAX file and several defective structures (ethmoid plate, crista galli, anterior clinoid process and inferior orbital fissure) were modified, using the 3D Studio Max 2016. In addition, each bone was isolated from the whole skull and painted in different color ( Fig. 2c & d). The model was then placed into a simulated classroom (Fig. 2a)) using the HTC VIVE Software Development Kit and Unreal Engine 4.15, which is compatible with the HTC VIVE (resolution: 2160⋅1200), a VR head-mounted display (HMD) developed by HTC (Taiwan, China, 2016). Users can zoom in/out the isolated skull bone and the remaining part at the same time through the handheld controllers. When the isolated structure is placed back to its original position, it will reset. To make up for the inability to view the printout in simulated classroom, a projector was created to project the print-out teaching materials on the screen in the front of the simulation classroom (Fig. 2b).

Participants
We recruited 74 third-year medical students of the PUMC, who finished their pre-medical program in Tsinghua University but had not yet began anatomy course. All participants sighed the informed consents. Each student got a random number generated by SPSS and were ranked number size. No.1-25 were assigned to VR skull group (VR group), No.26-50 were assigned to cadaveric skull group (cadaver group), No.51-74 were assigned to the 2D atlas group (atlas group). 73 the participants completed the trial, 1 student in the atlas group quitted for personal reason before the preintervention test.

Design
The study protocol and design were approved by the Institutional Review Board of the Institute of Peking Union Medical College Hospital (PUMCH) (Project No: ZS-1724). The flowchart of the study is displayed in Fig. 3. All participants finished pre-intervention tests. Then they attended a 30-min PowerPoint-based introductory lecture on cranial anatomy by a teacher from PUMC. During the lecture, each participant received a single printout of teaching materials for note-taking. After that, three groups were leaded to three separate rooms for a 30-min self-directed learning session using VR skulls, cadaveric skulls, 2D atlas. There was a 2-min instruction about the manipulation of VR equipment for VR group before learning. Study mentors were assigned to each room to prevent intragroup communication, and they were forbidden to answer questions related to anatomy. Each The subjective questionnaire aimed to assess participant's motivation and engagement ( Table 1). The questionnaire consisted of five parts, including enjoyment, learning efficiency, authenticity, attitude and intention to use, and used standard five-point Likert-scale to quantify responses (1-strongly disagree, 5-strongly agree with the statement). Comparison test scores between groups Table 1 displays the results on the pre-and post-intervention tests. The total score is the sum of the theoretical test score and the identification test score. Within-subject analysis showed overall improvement in pre-and post-intervention test scores, which was statistically significant different for participants in all the three groups (p < 0.001).
There were no statistically significant differences between the three groups on the scores of preintervention test (p = 0.634, 0.667, 0.176 in total score, theory test, and identification test, respectively), the post-intervention test (p = 0.571, 0.824, 0.511 in total score, theory test, and identification test, respectively), and score changes between preintervention and postintervention tests (p = 0.317, 0.524, 0.278 in total score, theory test, and identification test, respectively), as shown in Fig. 4. Since the theory score of each participant wasn't directly associated with the identification score, the median and quartiles in the total score are not equal to the sum of the theory score and the identification score.

Discussion
Previous studies focused on the comparison between VR models and 2D atlas or lacked the assessment of spatial ability 11 − 13,17−21 , while little attempt has been made to compare the VR models with cadaver models and 2D atlas in the same trial. Our study is the first randomized controlled trial that compare these three methods at the same time through both objective assessments and subjective assessments, which is designated as "various question types" 16 . The objective assessments comprised a theory test and an identification test. The latter is a crucial part to assess participants' spatial ability since structure identification outweighs theoretical knowledge in the study of anatomy, and it is a predictor of improved learning outcomes following 3D learning 16,22 .
The results of the objective assessment demonstrated that VR simulation learning had equivalent efficiency in anatomical learning as cadaver skull and atlas, despite the relative simplicity of the 3D VR model used in this study, lacking texture and haptic feedback. The stereoscopic three-dimensionality of 3D VR model directly incorporates the intrinsic spatial relationships of the anatomical sites studied, and thus may confer a spatial knowledge advantage 21 , which is consistent with the insignificant higher post-identification score of VR group than the other two groups.
Subjective evaluation of VR group and cadaver group also showed a more positive attitude towards learning model than 2D atlas group. Responses indicated that the two groups unanimously agreed with the enjoyment, teaching efficacy and authenticity of their skull models. Novel interventions usually arouse participants' curiosity and lead to better results and all participants were willing to promote the use of VR model in anatomy education. Similarly, previous studies comparing 3D VR model with traditional 2D teaching method also reported that VR was evaluated as a more enjoyable and useful educational tool 11,21,23 .
Cadavers offer high realism, haptic feedback, and the opportunity to use real instruments and tools, which is thought to be the gold standard in anatomical learning 24 − 26 , particularly in surgery 27 .
However, they are expensive, limited resource and offer no objective feedback of skill. 3D VR models, which are able for rapid and feedback-based modification, offer an opportunity for repetitive practice 23 . Another advantage of VR is that students are able to observe and receive instant visual feedback based on pre-defined practical tasks 28 . Critiques argued that this approach lacked the expert guidance during the learning process which play an important role in forming the basic framework 29 . In fact, teachers could also assess students' learning progress and mistakes through digital reports to further strengthen students' skills. What's more, participants in our trial conducted a self-learning in the absence of guidance and gain substantial progress in anatomical knowledge, which is consistent with previous study 30 . It suggested that individuals' self-learning at their private places is a feasible way to implement VR simulation learning without constrains of places or time 30 . In addition, 3D models are likely to enhance rather than replace lecture-based teaching by experts 23 .
We could move the initial part of the learning curve from practice on cadavers to VR simulation, allowing the participants to practice the procedure and acquire basic skills before using expensive lab facilities 30 .
In our trial, the scores of cadaver skull group showed no statistically significant differences from those of VR group and atlas group in this trial. This discrepancy might partially result from structural variation and damaged structures in the cadaveric skulls, and the negative psychological reaction in participants triggered by the cadaveric skulls 31,32 . Besides, we combined lecture and model learning together to simulate the real learning process. The lecture allowed participants to get hints about the correct answers and narrow the differences between the three groups. What's more, all these participants practiced and took the examination together introducing competitive element.
Our study incorporated a room-scale HMD unit, which is available for individuals. Many more manipulations can be achieved easily, such as rotating to a suitable view, isolating a single cranial bone and zooming in/out the model, making it a better tool in understanding of difficult anatomical structures. Participants in the VR group only received a 2-min instruction about the VR equipment, which suggested that the operation of our equipment and software can be quickly adapted to. This HMD unit provides a completely immersive experience with a high display resolution, a high refresh rate, and a more precise, low-latency constellation head tracking capability. Previous study reported a high adverse rate in VR group (headaches 25%; blurred vision 35%) 20 , while the adverse rate was lower in our trial (headaches 20%, blurred vision 4%). It can be inferred that the discomforts caused by the activities in the virtual environment was relieved with increased resolution and lower latency.
Large amount of knowledge to memorize, compact exam arrangement and unfamiliarity with learning material might bring too much burden to the brain, causing headache, blurred vision and other discomforts in all groups.
The results suggest two minor findings. First, male participants scored higher than female participants in post-intervention identification test. Previous studies have shown that individual spatial ability is associated with improved learning outcomes following small-group 3D learning 22,33 , which is consistent with our finding. One possible explanation is that males have stronger mental rotation, which is closely associated spatial imagination 34 . Similar results between genders were observed in VR group, indicating VR model learning relies on good spatial imagination partially. Second, 2D atlas tended to improve participants' understanding of structures at superior view. A possible explanation may be that the structures at superior seems to be at the same plane when observing from the top of the skull and they can be well painted on atlas.    Flowchart of study design.

Figure 4
Comparison between genders of three groups in post-intervention test score and change in score. (a) Group A: significant differences were found in post-intervention total score, identification test score and change in identification test score. (b) Group B: no significant differences were found. (c) Group C: significant differences were found in change in total score and change in identification test score.

Supplementary Files
This is a list of supplementary files associated with this preprint. Click to download.