Vertebra segmentation and labeling
About
Image Version:
69b0764d-dbe2-4b7d-bcc8-d0343c30ebf1 — Aug. 11, 2022
Associated publication:
Lessmann N, van Ginneken B, de Jong PA, Išgum I. Iterative fully convolutional neural networks for automatic vertebra segmentation and identification. Medical Image Analysis. 2019;53:142-155.
Summary
Left empty by the Algorithm Editors
Mechanism
This algorithm segments the vertebrae in a CT scan, assigns anatomical labels to each detected vertebra and classifies vertebrae as completely or partially visible in the scan. The underlying method is an iterative fully convolutional neural network, details can be found in the article published in Medical Image Analysis.
The algorithm is trained with a wide variety of CT scans of adults, including cases with compression fractures and pedicle screws.
Interfaces
This algorithm implements all of the following input-output combinations:
|
Inputs |
Outputs |
| 1 |
-
- Description
- Any CT image
- Kind
- Image
- Read from
-
/input/images/ct/<uuid>.mha or
/input/images/ct/<uuid>.tif
CT Image
|
-
- Description
- A collection of results of unknown type. Legacy, if possible please use alternative interfaces.
- Kind
- Anything
- Write to
-
/output/results.json
- Example
-
{
"key": "value",
"None": null
}
Results JSON File
-
- Description
- 0: background, 1-7: C1-C7, 8-19: T1-T12, 20-25: L1-L6, and 101-125 for partially visible vertebrae.
- Kind
- Segmentation
- Write to
-
/output/images/vertebra/<uuid>.mha or
/output/images/vertebra/<uuid>.tif
Vertebra
-
- Description
- A thumbnail depicting a segmentation of the spine
- Kind
- Thumbnail jpg
- Write to
-
/output/spine-segmentation-thumbnail.jpeg
Spine Segmentation Thumbnail
|
Validation and Performance
Left empty by the Algorithm Editors
Uses and Directions
This algorithm was developed for research purposes only.
If the processing fails or no vertebra are detected, there is often a problem with the input data, e.g., unsupported format, wrong data type, incomplete header, etc. Check that the image header contains the correct voxel spacing and direction cosine vectors and that the image values are Hounsfield Units (Air = -1000, Water = 0).
Please feel free to contact Nikolas Lessmann for help with trying out the algorithm, or if you would like to use it on a larger set of images.
Warnings
Left empty by the Algorithm Editors
Common Error Messages
Left empty by the Algorithm Editors
Information on this algorithm has been provided by the Algorithm Editors,
following the Model Facts labels guidelines from
Sendak, M.P., Gao, M., Brajer, N. et al.
Presenting machine learning model information to clinical end users with model facts labels.
npj Digit. Med. 3, 41 (2020).
10.1038/s41746-020-0253-3