Accuracy Analysis of Line-Based Registration for Image Guided Neurosurgery at Different Operating Areas-A Phantom Study

Dafeng Ji, Yuan Dong, Manning Wang, Zhijian Song

(2017)Computer Assisted Surgery

Abstract: Space registration is the primary function of neuronavigation systems. According to the stage of the operation, the registration could be classified as rigid and non-rigid methods. Scientists have proposed three types of rigid registration methods: point-based registration (PBR), line-based registration (LBR), and surface-based registration (SBR). PBR has been widely used in clinical applications. Recently, LBR was proposed as a new spacing registration method. However, the range and accuracy of LBR are still not defined for clinical applications. In this paper, LBR has been evaluated directly with target registration error (TRE) in different operating areas: sphenoid-frontal, parietal-temporal, and occipital areas. We used two head phantoms: elastic and rigid phantoms. After scanning with computerized tomography (CT), the difference between the TRE of the elastic and rigid phantom had been evaluated based on LBR method. Then, LBR had been employed at the rigid phantom using different line patterns: single (100 points), double (200 points), triple (300 points), and quartic lines (400 points). TRE were directly measured on the phantom. Then, t-tests were applied to evaluate the difference between the TRE of LBR of both the phantoms and line patterns. Results indicate that there is no statistical difference in TRE between the phantoms. TRE were reduced to less than 3 mm after the use of double lines which was significantly less than those after the use of single lines. Except for sphenoid tumor, the other operating areas showed statistical differences in TRE between double and triple lines. Except for temporal tumor, the differences between the TRE of triple and quartic lines are not significant.