This repository will be used to track and regroup most/all the relevant methods used for vertebral labeling on MRI and CT scans.
| Method name | Link | Open source | Task | Region | Modality | Deep Learning based | Date | Comments |
|---|---|---|---|---|---|---|---|---|
| Template-based approach | paper | Yes | Discs detection | Disc C2-C3 must be visible | MRI (T1w/ T2w) | No | 2014 | Do not work when the FOV is thoracic or lumbar |
| Spine-GAN | paper | No | Discs, vertebrae and neural foramen segmentation | Lumbar | MRI (T1w/ T2w) | Yes | 2018 | |
| Spine-explorer | paper | No | Vertebrae and discs segmentation | Lumbar | MRI (T2w) | Yes | 2020 | |
| Stacked hourglass | paper | Yes | Discs detection | Cervical FOV | MRI (T1w/ T2w) | Yes (Hourglass) | 2021 | Restricted to a specific FOV |
| SpineNetV2 | paper | Yes | Vertebral bodies detection/ radiological grading | Whole spine | MRI (T1w/ T2w/ STIR/ TIRM) | Yes | 2022 | Strange behaviour with T1w scans |
| Spine-transformers | paper | Yes | Vertebrae segmentation | Whole spine | CT | Yes | 2022 | |
| Total segmentator | paper | Yes | Global segmentation | Whole body | CT | Yes (nnUNet) | 2022 | |
| HCA-NET | paper | Yes | Discs detection | Cervical FOV | MRI (T1w/ T2w) | Yes | 2023 | Restricted to a specific FOV |
| MRI to CT registration then segmentation | paper | Yes | Vertebrae and discs segmentations | Whole spine | MRI (T1w/ T2w) | Yes (Pix2Pix) | 2023 | Relying on VerSe and SpineR |
| Vertebrae segmentation with anatomic consistency cycle | paper | Yes | Vertebrae segmentation | Whole spine | CT | Yes | 2023 | |
| SPINEPS | paper | Yes | Vertebrae and discs segmentations | Whole spine | T2w | Yes (nnUNetV2) | 2024 | No label identification but accurate instance segmentation |
NeuroPoly disc labeling implementations:
- Disc labeling benchmark: https://github.com/spinalcordtoolbox/disc-labeling-benchmark
- Hourglass approach: https://github.com/spinalcordtoolbox/disc-labeling-hourglass
- nnU-Net approach: https://github.com/spinalcordtoolbox/disc-labeling-nnunet
- VerSe: A vertebrae labelling and segmentation benchmark for multi-detector CT images.
- Panoptica: Public package to evaluate instance segmentations methods.
Contributions to this repository are welcome. If you have developed a new method or have improvements to existing methods, please submit a pull request. Additionally, feedback and suggestions for improvement are highly appreciated. Feel free to open an issue to report problems, propose new methods, or ask questions.