Neuroblastoma (NB) is a rare embryonic cancer that presents surgical challenges, particularly in high-risk cases. High-risk cases are those stratified according to image-defined risk factors (IDRFs) on medical imaging. In abdominal NB, common IDRFs include encasement of major vessels: the abdominal aorta and inferior vena cava (IVC) amongst others, which poses an increased surgical risk. With a surgical resection goal of over 90% for optimal survival, three-dimensional (3D) models derived from patient datasets have proven beneficial for pre-surgical planning.
This chapter presents the framework used to create a prototype 3D Neuroblastoma model within a virtual reality (VR) application for surgical planning and junior doctor education. The 3D VR model displays a large abdominal neuroblastoma with encasement of the IVC and inferior mesenteric artery (IMA), spatial relationship to the abdominal aorta, tributary vessels and abdominal organs. Key features of the VR development include (i) tumour transparency, (ii) a customisable interactive toggle display, (iii) CT dataset overlay, (iv) 360-degree rotation of the model, (v) medical information relating to neuroblastoma, including tumour volume and (vi) notation.
User testing was conducted at the Royal Hospital for Children, Glasgow (RHCG) with 20 medical professionals participating. Results demonstrate the application had a good usability rating (SUS 79.75) and sense of presence (ITC-SOPI) rating, with resulting scores in each category; Sense of Presence (Mean 3.75 ± 0.55 SD), Engagement (Mean 4.08 ± 0.4 SD), Ecological Validity (Mean 3.72 ± 0.83 SD), and Negative Effect (Mean 1.77 ± 0.78 SD).
A counterbalanced anatomical identification experiment comparing the 2D dataset to the 3D VR model showed a significant difference (p <.05) in errors committed between the control and VR group, demonstrating that participants performed better in VR (MVR = 0.35, σ 0.59) compared to 2D (Mcontrol = 0.85, σ = 0.93). A significant difference (p < 0.05) was also noted in the evolution of error between groups, suggesting that Group 1 participants who have first undertaken the control condition and then VR, have seen the number of errors decreasing indicating that the VR condition promotes greater accuracy for anatomical identification.
Limitations to this study include a small participant number (n=20) and a broad spectrum of knowledge among participants. Future research could address such limitations by increasing participant numbers and narrowing the education-level demographic, focusing on one cohort of participants.