Automatically segmenting vertebral bones in 3D medical images

1. A method comprising:

receiving a three-dimensional medical imaging dataset that includes a plurality of vertebral bones of a subject; and

segmenting the plurality of vertebral bones, wherein the segmenting includes:

obtaining a plurality of axial images from the three-dimensional medical imaging dataset;

extracting a spinal canal of the subject using a plurality of axial images from the three-dimensional medical imaging dataset;

after the extracting, generating an anterior line of the plurality of vertebral bones using one or more sagittal images from the medical imaging dataset;

after the extracting, obtaining a posterior line based on the extracted spinal canal; and generating a vertebral centerline using the anterior line and the posterior line;

wherein detecting one or more intervertebral discs; and determining a separating plane between adjacent vertebral bones from the detected intervertebral discs;

comprising: finding a bound box for each of the vertebral bones;

further comprising: segmenting intervertebral discs using a neural network; and

generating separating planes from the segmented intervertebral discs.

2. The method of claim 1 , wherein extracting the spinal canal includes:

identifying the spinal canal using an outer contour of the spinal canal.

3. The method of claim 1 , wherein extracting the spinal canal includes:

stacking the plurality of axial images; and

finding a path through the plurality of the axial images; and

using the path for canal segmentation.

4. The method of claim 1 , further comprising:

using the segmented plurality of vertebral bones to generate integer labels for voxels of the three-dimensional medical imaging dataset.

5. The method of claim 1 , wherein the medical imaging dataset is generated from a CT scanner.