3.0t scanners

This study was approved by the Ethics Committee of Biomedical Research, West China Hospital of Sichuan University. Written informed consent was waived because of the retrospective nature of data collection (age/gender) and the use of de-identified MRI images. A total of 319 consecutive patients in an age range between 18 and 95, who were prescribed cervical MRI for medical reasons and who were scanned between 2019 and 2021 using either of the Siemens 3.0-T scanners at the West China Hospital of Sichuan University. In all, 19 patients were excluded for the following reasons: 1) incomplete image of the cervical spine (n = 17); 2) insufficient MRI quality (n = 3). Finally, 300 patients were retrospectively collected in this study.
The images were acquired in the T2-weighted (T2W), including the cervical vertebral body, intervertebral disc, spinal canal, and spinal cord. The whole MRI dataset contained 600 images in total, with 16 categories: cervical bodies: C2, C3, C4, C5, C6, C7, and T1; cervical intervertebral disc: C2/3, C3/4, C4/5, C5/6, C6/7, and C7/T1; spinal canal; spinal cord; and background. Since each MRI contained a complete set of 16 categories, the total dataset had a total of 600 of each category.

A total of 452 consecutive patients aged between 18 and 95, for whom cervical MRI was prescribed for medical reasons, were scanned between 2019 and 2021 at the West China Hospital of Sichuan University and the First People’s Hospital of Longquanyi District using either Siemens 3.0T scanners. Overall, 17 patients were excluded for the following reasons: 1) incomplete image of the cervical spine (n = 14) and 2) insufficient MRI quality (n = 3). Finally, 435 patients were retrospectively collected in this study. The inclusion and exclusion flowchart of the study population is shown in Figure 2.
The data were randomly divided into a training set (1,830 study samples of 305 patients) and a test set (780 study samples of 130 patients) according to the ratio of 7:3. The radiomics analysis, feature selection, and model development were implemented in the training set, and then the related radiomics models were validated in the test set.

All 3D T1‐weighted imaging were acquired on Siemens 3.0 T scanners. The scanning parameters were as follows: repetition time (TR) = 2300 ms; echo time (TE) = 2.98 ms; inversion time = 900 ms; flip angle = 90°; slice number = 176; acquisition matrix = 240 × 256 and voxel size = 1 × 1 × 1 mm³.