Automatic 3D medical image orientation determination
A system, method and non-transitory computer readable storage medium for aligning a set of medical images. The operations for aligning the medical images include receiving a set of medical images, selecting anchor organs from the medical images, training a segmentation model to identify the anchor organs in the medical images based on a training dataset and generating, based on the segmentation model, a segmentation mask for the anchor organs. The operations also include computing image coordinates for the anchor organs in each of the medical images based on a center of mass of each of the anchor organs, determining a correlation between the image coordinates for the anchor organs in each of the medical images and corresponding anatomical coordinates for the anchor organs in the training dataset and aligning each of the set of medical images based on the correlation between the image coordinates and the anatomical coordinates.
1 . A method, comprising:
receiving a set of medical images from a medical imager;
selecting one or more anchor organs from the medical images;
generating, based on applying a segmentation model to each of the set of medical images, a segmentation mask for the selected one or more anchor organs, wherein the segmentation model is trained to identify the selected one or more anchor organs in the set of medical images based on a training dataset received from a database;
computing image coordinates for the selected one or more anchor organs in each of the set of medical images based on a center of mass of each of the selected one or more anchor organs and the generated segmentation mask;
determining a correlation between the image coordinates for the selected one or more anchor organs in each of the set of medical images and corresponding anatomical coordinates for the selected one or more anchor organs in the training dataset; and
aligning each of the set of medical images based on the correlation between the image coordinates and the anatomical coordinates.
2 . The method of claim 1 , wherein the aligning is further based on a decision function comprising one of a flipping case, a shift case or a rotation case.
3 . The method of claim 2 , wherein the aligning is further based on a set of rules based on anatomical features of a person.
4 . The method of claim 1 , wherein the segmentation model is trained to consume the set of medical images in a low resolution.
5 . The method of claim 1 , wherein the segmentation model is trained to be scale and rotation invariant based on the training dataset being augmented with random scaling and rotations.
6 . The method of claim 1 , wherein the segmentation model is trained to segment each of the set of medical images by giving each voxel of each medical image a label, wherein the label comprises one of a voxel belonging to the selected one or more anchor organs, a voxel belonging to a surrounding of the selected one or more anchor organs or a voxel not belonging to the selected one or more anchor organs.
7 . A system, comprising:
a memory storing a set of medical images received from a medical imager and a training dataset;
a processor configured to:
select one or more anchor organs from the set of medical images,
generate, based on application of a segmentation model to each of the set of medical images, a segmentation mask for the selected one or more anchor organs, wherein the segmentation model is trained to identify the selected one or more anchor organs in the set of medical images based on a training dataset received from a database,
compute image coordinates for the selected one or more anchor organs in each of the set of medical images based on a center of mass of each of the selected one or more anchor organs and the generated segmentation mask,
determine a correlation between the image coordinates for the selected one or more anchor organs in each of the set of medical images and corresponding anatomical coordinates for the selected one or more anchor organs in the training dataset, and
align each of the set of medical images based on the correlation between the image coordinates and the anatomical coordinates.
8 . The system of claim 7 , wherein the aligning is further based on a decision function comprising one of a flipping case, a shift case or a rotation case.
9 . The system of claim 7 , wherein the segmentation model is trained to consume the set of medical images in a low resolution.
10 . The system of claim 7 , wherein the segmentation model is trained to be scale and rotation invariant based on the training dataset being augmented with random scaling and rotations.
11 . The system of claim 7 , wherein the segmentation model is trained to segment each of the set of medical images by giving each voxel of each medical image a label, wherein the label comprises one of a voxel belonging to the selected one or more anchor organs, a voxel belonging to a surrounding of the selected one or more anchor organs or a voxel not belonging to the selected one or more anchor organs.
12 . A non-transitory computer readable storage medium comprising a set of instructions which, when executed by a processor, cause the processor to:
receive a set of medical images from a medical imager;
select one or more anchor organs from the medical images;
generate, based on application of a segmentation model to each of the set of medical images, a segmentation mask for the selected one or more anchor organs, wherein the segmentation model is trained to identify the selected one or more anchor organs in the set of medical images based on a training dataset received from a database;
compute image coordinates for the selected one or more anchor organs in each of the set of medical images based on a center of mass of each of the selected one or more anchor organs and the generated segmentation mask;
determine a correlation between the image coordinates for the selected one or more anchor organs in each of the set of medical images and corresponding anatomical coordinates for the selected one or more anchor organs in the training dataset; and
align each of the set of medical images based on the correlation between the image coordinates and the anatomical coordinates.
13 . The non-transitory computer readable storage medium of claim 12 , wherein the aligning is further based on a decision function comprising one of a flipping case, a shift case or a rotation case.
14 . The non-transitory computer readable storage medium of claim 12 , wherein the segmentation model is trained to one of (i) consume the set of medical images in a low resolution, (ii) be scale and rotation invariant based on the training dataset being augmented with random scaling and rotations or (iii) segment each of the set of medical images by giving each voxel of each medical image a label.