Assessment of registration accuracy during computer-aided oncologic limb-salvage surgery

• Published in: International journal for computer assisted radiology and surgery 2015-09, Vol.10 (9), p.1469-1475
• Main Authors: Stoll, Kurt E., Miles, Joan D., White, Jedediah K., Punt, Stephanie E. W., Conrad, Ernest U., Ching, Randal P.
• Format: Article
• Language: English
• Subjects: Adolescent; Algorithms; Bone Neoplasms - diagnostic imaging; Bone Neoplasms - surgery; Child; Child, Preschool; Computer Imaging; Computer Science; Health Informatics; Humans; Imaging; Limb Salvage - methods; Magnetic Resonance Imaging - methods; Male; Medicine; Medicine & Public Health; Models, Statistical; Original Article; Osteosarcoma - diagnostic imaging; Osteosarcoma - surgery; Pattern Recognition and Graphics; Radiographic Image Interpretation, Computer-Assisted; Radiology; Reproducibility of Results; Sarcoma, Ewing - diagnostic imaging; Sarcoma, Ewing - surgery; Surgery; Surgery, Computer-Assisted - methods; Tomography, X-Ray Computed - methods; Vision
• ISSN: 1861-6410; 1861-6429
• DOI: 10.1007/s11548-014-1146-1

Purpose Computer-aided surgery is used in musculoskeletal tumor procedures to improve the surgeon’s orientation to local anatomy during tumor resection. For the navigation system to function correctly, preoperative imaging (e.g., CT, MR) must be registered to the patient in the operating room. The goals of this study were (1) to directly quantify registration accuracy in computer-aided tumor surgery and (2) to validate the “system reported error” (SRE) of the navigation system. Methods Registration accuracy was evaluated in eight bone sarcoma cases by determining the location of the anatomical paired-points used for registration following surface matching. Coordinates of specific intraoperative post-registration points were compared with the corresponding coordinates in preoperative CT scans to determine the measurement error (ME). Results The mean difference between post-registration points and planned registration points was 12.21 ± 6.52 mm , significantly higher than the mean SRE ( 0.68 ± 0.15 mm ; p = 0.002 ; 95 % CI 6.11–16.96 mm). The SRE poorly correlated with the calculated ME ( R 2 = 0.040 ). Anatomical paired-point registration with surface matching results in a substantial shift in the post-registration coordinates of the same paired-points used for registration, and this shift is not represented by the SRE. Conclusion The SRE of a surgical navigation system was poorly correlated with direct measurements obtained in musculoskeletal tumor surgery. Improvement in registration accuracy is needed to better navigate tumor boundaries and ensure clear margins while maximally preserving the unaffected tissues and reducing operative morbidity.