Poster R8 - Evaluating the Impact of Virtual Reality (VR) on Surgeons' Mental Model During Pre-Surgical Planning

Introduction: Surgical planning is a complex task which requires prior experience, general knowledge of anatomy and physiology, integration of patient specific information, and psychomotor skills to build a proactive mental model. Virtual reality (VR) has been evaluated for many applications in healthcare, including surgical practice and pre-surgical planning. However, the specific opportunities and benefits of VR for pre-surgical planning are often reported as speculation based on self-reported perceptions from users, and many efforts seek to validate specific VR systems separate from any clinical workflow. We see an opportunity to better understand the benefits VR may confer to surgical planning in real clinical workflows, and describe both qualitatively as well as measure quantitatively the impact VR has on constructing mental models for complex surgical cases.

Materials and

Methods: To study the impact of VR on pre-surgical planning, surgeons were invited to volunteer real patient imaging for review in VR. A modified NASA-TLX cognitive demand survey was deployed as a pre and post anchor for each session. Surgeons were recorded as they freely explored a 3D model of specific patient anatomy prepared from conventional medical imaging (CT, MRI). Surgeons were encouraged to talk through their thought process as they explored, and audio and video were captured from the first person view of the VR headset. Three congenital heart proceduralists contributed 10 session recordings which were qualitatively coded using a grounded theory approach with 5 independent coders, and themes were identified across cases.

Results, Conclusions, and Discussions: Results
The qualitative analysis of VR sessions revealed several key strategies physicians would employ to build and refine their mental models for surgery. We found surgeons would trace paths in 3D as they explored, and would often use the model to confirm their interpretation of 2D imaging. We found that these strategies often intersected with emotions of excitement, uncertainty, and frustration, and ultimately led to a significant increase in self-reported confidence in understanding of the patient anatomy after the VR session (Welch’s t-test, p=0.012), despite finding no significant change in self-reported confidence in surgical approach (p=0.310).

Conclusion
Our findings suggest that confidence may be a poor predictor of preparedness given the uncertainty portrayed in VR and the significant increase in understanding of the patient anatomy after the VR session, especially given quotes such as: “I'm confident that I can do this. The real question is, can I close one hole and be done and this patient will be fine?”. The strategies observed during exploration in 3D indicated VR helped the surgeons refine their mental model of the patient anatomy and develop a better sense of how their surgical approach would be physically carried out in the operating theater.

Discussion
The cases submitted were volunteered, and constituted challenging surgical cases, so our findings are subject to selection bias. We also observed interesting behavioral differences between the most experienced surgeon and other participants, which we hope to explore in further work by expanding our pool of subjects. Our findings may have broader implications for how VR tools are assessed for impact, and may be useful to programs looking to better understand how to prepare surgeons for complex surgical cases. Finally, our finding that surgeons reported an increase in understanding of patient anatomy from exploring in 3D after reviewing in 2D may offer further opportunities for technologies to enhance clinical workflows for providers and patients in the future.

Acknowledgements (Optional): A portion of funding for this work came from the Jump-ARCHES foundation. Connor Davey provided engineering support for preparing the models for VR. Alexa Waltz assisted in administering the VR equipment. UIUC Students AH, AT, NS, and BO provided coding.