Abstract
Summary
We found good to excellent reproducibility of in vivo hip joint angle measurements during repeated sitting when derived from registering low-resolution Open MRI imagesets with a reference high-resolution conventional MRI scan, despite only moderate similarity of the segmented volumes.
Keywords: hip, kinematics, MRI, femoroacetabular impingement, repeatability
Introduction
Femoroacetabular impingement (FAI) is a mechanical hip disorder caused by an abnormal bony contact between the femur and acetabulum. Open MRIs can enable studies of FAI under weightbearing, but the resolution of such scans is comparatively low, so it is useful to obtain high resolution (HR) reference scans from a conventional MRI and register lower resolution (LR) open MRI images to the HR images. The purpose of this study was to establish the degree of correspondence between the segmented volumes from the two types of scanner and to estimate the repeatability of joint angle measurements.
Patients and Methods
Three healthy subjects were scanned in a lying position to obtain high resolution (HR) MRI images of the pelvis, hip and knee. The same subjects were scanned four times in a sitting position in a 0.5T open MRI scanner to obtain corresponding low resolution (LR) images of the hip joint. Between sittings, subjects rose and sat again, and during each sitting, a block was inserted and removed from underneath their foot. Volumetric models of the femur and acetabulum were manually segmented from the HR and LR MRI images. The LR (sitting) models were registered to the HR (supine) models using an intensity-based rigid registration method and the degree of overlap and the femoropelvic joint angles computed. Analysis is complete for two of the three subjects (results for the third are pending).
Results
The overlap between the LR and HR imagesets is reasonable in most scan slices - per-slice Dice Similarity Coefficients (DSCs) are typically around 85%, although DSCs near the edges of volumes can sometimes drop to about 75%. Nonetheless, the resulting registrations are relatively insensitive to these moderate discrepancies. For the two subjects whose data has been analyzed to date, the femoral angle relative to the scanner is quite repeatable (SD < 0.9° for the flexion angle under each block condition). The mean femoral flexion angle change between block conditions was also comparatively consistent (SD 1.7° & 2.2° for the two subjects), but most of the other hip joint angles (and changes between the up and down conditions) were more variable (SDs up to ∼5.7°).
Discussion/Conclusions
Although there are moderate discrepancies between the LR and HR segmented volumes, the resulting registrations and estimated joint angles are relatively consistent (SDs under 2°). The larger degree of pelvic flexion variability under repositioning indicates that it may be challenging for subjects to reproduce a desired posture on different occasions. Our results could not be directly compared with the only other studies we are aware of using open MRI to investigate FAI because neither combined LR and HR images.
