Augmented reality and artificial intelligence-based navigation during percutaneous vertebroplasty: a pilot randomised clinical trial

• Published in: European spine journal 2020-07, Vol.29 (7), p.1580-1589
• Main Authors: Auloge, Pierre, Cazzato, Roberto Luigi, Ramamurthy, Nitin, de Marini, Pierre, Rousseau, Chloé, Garnon, Julien, Charles, Yan Philippe, Steib, Jean-Paul, Gangi, Afshin
• Format: Article
• Language: English
• Subjects: Accuracy; Artificial intelligence; Augmented reality; Bone surgery; Clinical trials; Compression; Fluoroscopy; Life Sciences; Medicine; Medicine & Public Health; Neurosurgery; Original Article; Skin; Surgical Orthopedics; Vertebrae
• ISSN: 0940-6719; 1432-0932
• DOI: 10.1007/s00586-019-06054-6

Purpose To assess technical feasibility, accuracy, safety and patient radiation exposure of a novel navigational tool integrating augmented reality (AR) and artificial intelligence (AI), during percutaneous vertebroplasty of patients with vertebral compression fractures (VCFs). Material and methods This prospective parallel randomised open trial compared the trans-pedicular access phase of percutaneous vertebroplasty across two groups of 10 patients, electronically randomised, with symptomatic single-level VCFs. Trocar insertion was performed using AR/AI-guidance with motion compensation in Group A, and standard fluoroscopy in Group B. The primary endpoint was technical feasibility in Group A. Secondary outcomes included the comparison of Groups A and B in terms of accuracy of trocar placement (distance between planned/actual trajectory on sagittal/coronal fluoroscopic images); complications; time for trocar deployment; and radiation dose/fluoroscopy time. Results Technical feasibility in Group A was 100%. Accuracy in Group A was 1.68 ± 0.25 mm (skin entry point), and 1.02 ± 0.26 mm (trocar tip) in the sagittal plane, and 1.88 ± 0.28 mm (skin entry point) and 0.86 ± 0.17 mm (trocar tip) in the coronal plane, without any significant difference compared to Group B ( p  > 0.05). No complications were observed in the entire population. Time for trocar deployment was significantly longer in Group A (642 ± 210 s) than in Group B (336 ± 60 s; p  = 0.001). Dose–area product and fluoroscopy time were significantly lower in Group A (182.6 ± 106.7 mGy cm 2 and 5.2 ± 2.6 s) than in Group B (367.8 ± 184.7 mGy cm 2 and 10.4 ± 4.1 s; p  = 0.025 and 0.005), respectively. Conclusion AR/AI-guided percutaneous vertebroplasty appears feasible, accurate and safe, and facilitates lower patient radiation exposure compared to standard fluoroscopic guidance. Graphic abstract These slides can be retrieved under Electronic Supplementary Material.