Quantification of the accuracy of MRI generated 3D models of long bones

Abstract

Orthopaedic implants are typically designed using accurate computer tomography (CT) based 3D models of long bones. Unlike CT, magnetic resonance imaging (MRI) does not involve ionising radiation and would therefore be a suitable alternative to CT. This study aims to quantify the accuracy of MRI based 3D models of long bones.
The femora of three intact cadaver ovine limbs were scanned using 1.5T MR and CT scanners. Image segmentation was performed using adaptive thresholding. Reference models were generated by digitising the dissected bone surfaces. The MR derived models were validated against reference and CT derived models.-----
The MR derived models generated mean errors of 0.23, 0.25, 0.22 mm, and 0.20, 0.21, 0.20 mm compared to the reference and CT-based models, respectively. For all models, the shaft was more accurate than the distal and proximal ends.
The results indicate that the accuracy of MRI-based models is comparable to that of CT-based models.