Isotropic MRI of the Upper Extremity with 3D-FSE-Cube

Introduction Musculoskeletal MRI is commonly done with two-dimensional fast-spin-echo (2D-FSE) and multiple acquisition planes. This results in anisotropic voxels, partial volume artifact and slice gaps. Three-dimensional fast-spin-echo (3D-FSE-Cube), allows for isotropic voxel acquisition with FSE contrast, diminishing the drawbacks of 2D-FSE while reducing exam time and allowing for image reformation in oblique planes [1-3]. This study compares the image quality of 3D-FSE-Cube to 2D-FSE in evaluation of the healthy shoulder and elbow at 3.0T. Methods Eleven shoulders and elbows of healthy subjects were imaged using a GE 750 3.0T MRI scanner (GE Healthcare, Milwaukee, WI). All images of the shoulder were acquired with an 8-channel shoulder coil with TR/TE 2500/35ms, 384 x 288 matrix, 20cm FOV, receiver bandwidth ±31 kHz. 3DFSE-Cube image were acquired in the axial plane with an ETL of 60, acceleration factor of 3.65 and auto-calibrated parallel imaging (ARC), 0.6 mm slice thickness, 250 sections, and an imaging time of 5 min 30 sec [4]. 2D-FSE images of the shoulder were obtained in axial and oblique coronal planes with 3mm slices, 0.5mm gaps, ETL 8, and an imaging time of 3mins 20s. All images of the elbow were acquired using an 8-channel knee coil with TR/TE 3000/35ms, 288 x 256 matrix, 14cm FOV, and receiver bandwidth ±31 kHz. 3D-FSE-Cube was acquired in the coronal plane with an ETL of 60, acceleration factor of 3.65, 0.6 mm slice thickness, 128 sections, and an imaging time of 5 min 55 sec. 2D-FSE images of the elbow were obtained in axial and coronal planes with 2.5mm slices, 0.5mm gaps, ETL 8, and an imaging time of 4mins 54s. To allow noise measurements from identical noise images, both methods were also acquired with the RF pulse off. Noise images were processed through the identical linear reconstruction pipeline as the signal data. Regions of interest (ROIs) were placed in fluid, muscle and cartilage for both methods, as well as in identical noise images. Slice averaging to the same slice thickness as the 2D-FSE data was used to normalize the 3DFSE-Cube SNR and CNR measurements. SNR and CNR were compared with a paired t-test. Two fellowship-trained radiologists compared 3DFSE-Cube with 2D-FSE for image quality, blurring and artifacts using a seven-point scale (-3 for 3D-FSE-Cube much worse than 2D-FSE; 0 for equal; +3 for 3D-FSE-Cube much better than 2D-FSE). Ratings were analyzed with a Wilcoxon signed rank test. Results Normalized muscle and cartilage SNR were similar between 3D-FSE-Cube and 2D-FSE in both the elbow and shoulder. Fluid SNR and fluidcartilage CNR were significantly higher using 3D-FSE-Cube in both the elbow and shoulder (Figure 1; * = p < .05). 3D-FSE-Cube ratings in the shoulder were slightly worse then 2D-FSE in image quality, blurring and artifacts (-0.5+ 0.2; p< .05). 3D-FSE-Cube ratings in the elbow were slightly worse then 2D-FSE in image quality and blurring (-0.6+0.2; p < .05) but not significantly different in artifacts. Due to the reformatting ability of 3D-FSE-Cube images, thin and oblique anatomy can be evaluated in both the shoulder and elbow (Figures 2 and 3). Conclusion The ability to acquire thinner slices using 3D-FSE-Cube greatly decreases partial volume artifact and allows images to be viewed in oblique planes and at arbitrary slice thickness, thereby improving anatomic depiction. Further refinement of 3D-FSE-Cube acquisition is needed for image quality to match that of 2D-FSE. With rapid imaging times, 3D-FSE-Cube is well suited for patients who are in pain, claustrophobic, or are pediatric. 3D-FSECube is a promising high-resolution method that may improve depiction of complex shoulder and elbow anatomy.