Comparative accuracy of lower limb bone geometry determined using MRI, CT, and direct bone 3D models.

ABSTRACT

Advancements in imaging and segmentation techniques mean that three dimensional (3D) modeling of bones is now increasingly used for preoperative planning and registration purposes. Computer tomography (CT) scans are commonly used due to their high bone-soft tissue contrast, however they expose subjects to radiation. Alternatively, magnetic resonance imaging (MRI) is radiation-free however, geometric field distortion and poor bone contrast have been reported to degrade bone model validity compared to CT. The present study assessed the accuracy of 3D femur and tibia models created from "Black Bone" 3T MRI and high resolution CT scans taken from 12 intact cadaveric lower limbs by comparing them with scans of the de-fleshed and cleaned bones carried out using a high-resolution portable compact desktop 3D scanner (Model HDI COMPACT C210; Polyga). This scanner used structured light (SL) to capture 3D scans with an accuracy of up to 35 µm. Image segmentation created 3D models and for each bone the corresponding CT and MRI models were aligned with the SL model using the iterative closest point (ICP) algorithm and the differences between models calculated. Hausdorff distance was also determined. Compared to SL scans, the CT models had an ICP error of 0.82 ± 0.2 and 0.85 ± 0.2 mm for the tibia and femur respectively, whilst the MRI models had an error of 0.97 ± 0.2 and 0.98 ± 0.18 mm. A one-way analysis of variance found no significant difference in the Hausdorff distances or ICP values between the three scanning methods (p > .05). The black bone MRI method can provide accurate geometric measures of the femur and tibia that are comparable to those achieved with CT. Given the lack of ionizing radiation this has significant benefits for clinical populations and also potential for application in research settings.