We compared the diagnostic ability of a convolutional neural network (CNN) to that of spine surgeons for differentiating between osteoporotic vertebral fractures and malignant vertebral compression fractures based on MRI. The performance of the CNN was equal or superior to that of spine surgeons.

No abstract

No abstract

Magnetic resonance vessel wall imaging is desirable for evaluating Kawasaki disease (KD)-associated coronary arterial lesions. To evaluate the reproducibility of three-dimensional turbo spin-echo (3D-TSE) and two-dimensional dual inversion-recovery turbo spin-echo (2D-DIR-TSE) for coronary vessel wall imaging in KD. Ten patients were prospectively enrolled. Coronary vessel wall imaging with axial-slice orientation 3D-TSE and 2D-DIR-TSE were acquired for cross-sectional images in aneurysmal and normal regions. Lumen area (LA), wall area (WA), and normalized wall index (NWI) of cross-sectional images were measured in both regions. Reproducibility between 3D-TSE and 2D-DIR-TSE was evaluated via intraclass correlation coefficients (ICCs) and Bland–Altman plots. 48 points (aneurysmal, 27; normal, 21) were evaluated. There were high ICCs between 3D-TSE and 2D-DIR-TSE in LA (0.95) and WA (0.95). In aneurysmal regions, 95% limits of agreement were LA, WA, and NWI of − 29.9 to 30.4 mm2, − 18.8 to 15.0 mm2, and − 0.22 to 0.20, respectively. In normal regions, the 95% limits of agreement were LA, WA, and NWI of − 4.44 to 4.38 mm2, − 3.51 to 4.30 mm2, and − 0.14 to 0.16, respectively. No fixed and proportional biases between 3D-TSE and 2D-DIR-TSE images in aneurysmal and normal regions were noted. 3D-TSE was reproducible with conventional 2D-DIR-TSE for coronary vessel wall assessment on KD.

The purpose of this study is to find the parameters that provide the optimal trade-off in Signal-to-noise-ratio (SNR) and contrast-to-noise ratio (CNR) for 3D ihMT m-FFE (3D-ihMT) and to explore the feasibility of that for the segmentability of gray and white matters in the cervical spinal cord compared with 2D ihMT m-FFE (2D-ihMT). We recommend a Flip angle of 7 and TFE factor of 5 as an optimal trade-off between segmentability and artifact level, at least 16 nr repetitions are required for robust segmentation. The segmentability of 3D-ihMT could be increased to a level close to that of 2D-ihMT.