Abstract
Background
Unreliable neuronavigation owing to inaccurate patient-to-image registration and brain shift is a major problem in conventional magnetic resonance imaging–guided neurosurgery. We performed a prospective intraoperative validation of a system for fully automatic correction of this inaccuracy based on intraoperative three-dimensional ultrasound and magnetic resonance imaging-to-ultrasound registration.
Methods
The system was tested intraoperatively in 13 tumor resection cases, and performance was evaluated intraoperatively and postoperatively.
Results
Intraoperatively, the system was accurate enough for tumor resection guidance in 9 of 13 cases. Manually placed anatomic landmarks showed improvement of alignment from 5.12 mm to 2.72 mm (median) after intraoperative correction. Postoperatively, the limitations of the current system were identified and modified for the system to be sufficiently accurate in all cases.
Conclusions
Automatic and accurate correction of spatially unreliable neuronavigation is feasible within the constraints of surgery. The current limitations of the system were also identified and addressed.
Unreliable neuronavigation owing to inaccurate patient-to-image registration and brain shift is a major problem in conventional magnetic resonance imaging–guided neurosurgery. We performed a prospective intraoperative validation of a system for fully automatic correction of this inaccuracy based on intraoperative three-dimensional ultrasound and magnetic resonance imaging-to-ultrasound registration.
Methods
The system was tested intraoperatively in 13 tumor resection cases, and performance was evaluated intraoperatively and postoperatively.
Results
Intraoperatively, the system was accurate enough for tumor resection guidance in 9 of 13 cases. Manually placed anatomic landmarks showed improvement of alignment from 5.12 mm to 2.72 mm (median) after intraoperative correction. Postoperatively, the limitations of the current system were identified and modified for the system to be sufficiently accurate in all cases.
Conclusions
Automatic and accurate correction of spatially unreliable neuronavigation is feasible within the constraints of surgery. The current limitations of the system were also identified and addressed.