In this paper, we describe a system for noninvasively determining bone surface points using an optically tracked A-mode ultrasound transducer. We develop and validate a calibration method; acquire cranial surface points for a skull phantom, three volunteers, and one patient; and register these points to surfaces extracted from CT images of the phantom and patient. Our results suggest that the bone surface point localization error of this system is less than 0.5 mm. The target registration error (TRE) of the cranial surface-based registration for the skull phantom was computed by using as a reference gold standard the point-based registration obtained with eight bone-implanted markers. The mean TRE for a 150-surface-point registration is 1.0 mm, and ranges between 1.0 and 1.7 mm for six 25-surface-point registrations. Our preliminary results suggest that accurate, noninvasive, image-to-physical registration of head images may be possible using an A-mode ultrasound-based system.
|LECTURE NOTES IN COMPUTER SCIENCE
|2nd International Conference on Medical Image Computing and Computer-Assisted Intervention (MICCAI 99)
|1/01/1999 → …