A quantitative comparison of edges in 3D intraoperative ultrasound and preoperative MR images of the brain

Abstract

The displacement of brain tissue during neurosurgery is a significant source of error for image-guidance systems. We are investigating the use of a computational model to predict brain shift and warp preoperative MR images accordingly. 3D ultrasound appears to be a valid means of acquiring intraoperative images to use as sparse data for guiding the model calculations. We present here a study of edge detection in MR and ultrasound images to investigate the accuracy of our ultrasound system and explore methods of extracting sparse data. Ultrasound images are acquired prior to deformation and compared to their corresponding oblique MR/CT images. Results from phantom images show misalignment by an average of 2.2 ± 1.67 degrees, 0.90 ± 0.82 mm in x, and 0.87 ± 0.87 mm in y. Patient images are misaligned by an average of 2.2 ± 1.4 degrees, 1.24 ± 0.89 mm inx, and 0.97 ± 0.87 mm in y.