Reliability of 3D localisation of ACL attachments on MRI: comparison using multi-planar 2D versus high-resolution 3D base sequences

Abstract

Anatomic placement of anterior cruciate ligament (ACL) grafts at arthroscopic reconstruction can be challenging. Localising ACL attachments on magnetic resonance imaging (MRI) sequences pre-operatively could aid with planning for anatomic graft placement. Though ACL attachments can be identified on two-dimensional (2D) MRI, slice thickness theoretically limits out-of-plane accuracy and a 3D MRI base sequence with smaller isotropic voxels may improve observer reliability in localising ACL attachment locations. The purpose of this study was to test whether a high-resolution 3D sequence improved inter- and intra-observer reliability of ACL attachment localisation compared with conventional 2D MRI for this application. Twenty paediatric knees were retrospectively scanned at 1.5 Tesla with multi-planar 2D proton density (slice thickness 3-4mm) and T2-weighted 3D multiple-echo data image combination gradient echo (isotropic 0.8mm voxels) sequences. Two observers blinded to each others' findings identified ACL attachments on MRI slices, and 3D reconstructions showing ACL attachments were produced. ACL attachment centre locations and areas were calculated, and reliability assessed. Inter-observer variation of centre locations of ACL attachments identified on 3D versus 2D sequences was not significantly different (mean±SD): 1.8±0.6 versus 1.5±0.7mm at femoral attachments, 1.7±0.7 versus 1.5±0.8mm at tibial attachments (p>0.05). The 95% confidence interval for centre locations was <4.0mm in all cases. Inter-observer reliability of attachment areas was not higher for 3D sequences. ACL attachment centres were localised with high and similar inter- and intra-observer reliability on a high-resolution 3D and multi-planar conventional 2D sequences. Using this technique, MRI could potentially be used for planning and intra-operative guidance of anatomic ACL reconstruction, whether from 2D or 3D base sequences. Surgeons in clinical practice need not order a lengthy dedicated 3D MRI to localise ligament attachments, but can confidently use a standard 2D MRI for this application. III.