Objective: To create 3DMR osseous models of the shoulder similar to 3DCT models using a gradient-echo-based two-point/Dixon sequence.
Materials and methods: CT and 3TMR examinations of 7 cadaveric shoulders were obtained. Glenoid defects were created in 4 of the cadaveric shoulders. Each MR study included an axial Dixon 3D-dual-echo-time T1W-FLASH (acquisition time of 3 min/30 s). The water-only image data from the Dixon sequence and CT data were post-processed using 3D software. The following measurements were obtained on the shoulders: surface area (SA), height/width of the glenoid and humeral head, and width of the biceps groove. The glenoid defects were measured on imaging and compared with measurements made on en face digital photographs of the glenoid fossae (reference standard). Paired t tests/ANOVA were used to assess the differences between the imaging modalities.
Results: The differences between the glenoid and humeral measurements were not statistically significant (cm): glenoid SA 0.12 ± 0.04 (p = 0.45) and glenoid width 0.13 ± 0.06 (p = 0.06) with no difference in glenoid height measurement; humeral head SA 0.07 ± 0.12 (p = 0.42), humeral head height 0.03 ± 0.06 (p = 0.42), humeral head width 0.07 ± 0.06(p = 0.18), and biceps groove width 0.02 ± 0.01 (p = 0.07). The mean/standard deviation difference between the reference standard and 3DMR measurements was 0.25 ± 0.96 %/0.30 ± 0.14 mm; 3DCT 0.25 ± 0.96 /0.75 ± 0.39 mm. There was no statistical difference between the measurements obtained on 3DMR and 3DCT (percentage, p = 0.45; mm, p = 0.20).
Conclusion: Accurate 3D osseous models of the shoulder can be produced using a 3D two-point/Dixon sequence and can be added to MR examinations with a minor increase in imaging time, used to quantify glenoid loss, and may eliminate the need for pre-surgical CT examinations.