It remains controversial whether some lesions of femoral head osteonecrosis regress during the natural course of the disease. With image registration, accurately matched image sets of the same subject can be acquired at different times. We applied image registration to evaluate lesion size change and assessed accuracy and usefulness compared to volume measurements and a conventional method. We also investigated whether lesions regress with this technique and with volume measurements. Baseline and 1 year minimum follow-up scans were conducted on 25 patients (31 hips) without radiological evidence of collapse. A three-dimensional (3D) spoiled gradient recalled echo sequence was used in the coronal direction (slice thickness = 2 mm; slice pitch = 1 mm). Size change was evaluated on all contiguous pairs of matched images after image registration. As a conventional method, coronal images (slice thickness = 5 mm) were reconstructed, and size change was evaluated on the five representative coronal slices. Evaluation with the conventional method identified eight lesions with apparent reduction; assessments using image registration and volume measurements identified three lesions, all within a year of initial steroid treatment and remaining at ARCO stage I at follow up. Evaluation of lesion size change using image registration was comparable to volume measurements. Inaccurate estimation of lesion size change due to mismatching of slice planes can be excluded. We demonstrated that some early lesions detected less than a year after initial steroid treatment can show size reduction with image registration as well as with volume measurements.
(c) 2006 Orthopaedic Research Society. Published by Wiley Periodicals, Inc.