System and methods for image segmentation using convolutional neural network
The present disclosure relates to systems, methods, devices, and non-transitory computer-readable storage medium for segmenting three-dimensional images. In one implementation, a computer-implemented method for segmenting a three-dimensional image is provided. The method may include receiving the three-dimensional image acquired by an imaging device, and creating a first stack of two-dimensional images from a first plane of the three-dimensional image and a second stack of two-dimensional images from a second plane of the three-dimensional image. The method may further include segmenting, by a processor, the first stack and the second stack of two-dimensional images using at least one neural network model. The method may also include determining, by the processor, a label map for the three-dimensional image by aggregating the segmentation results from the first stack and second stack.
1. A computer-implemented method for segmenting a three-dimensional medical image, the method comprising:
receiving the three-dimensional medical image acquired by an imaging device;
creating a first stack of two-dimensional images from a first plane of the three-dimensional medical image and a second stack of two-dimensional images from a second plane of the three-dimensional medical image;
segmenting, by a processor, the first stack and the second stack of two-dimensional images using at least one neural network model; and
determining, by the processor, a label map for the three-dimensional medical image by aggregating the segmentation results from the first stack and second stack.
2. The method of claim 1 , further comprising:
creating a third stack of two-dimensional images from a third plane of the three-dimensional medical image;
segmenting the third stack of two-dimensional images using the at least one neural network model; and
determining the label map for the three-dimensional medical image by aggregating the segmentation results from the first stack, the second stack, and the third stack.
3. The method of claim 1 , wherein segmenting the first stack of two-dimensional images uses a first neural network model and segmenting the second stack of two-dimensional images uses a second neural network model.
4. The method of claim 1 , wherein the first plane or the second plane is selected from an axial plane, a sagittal plane, and a coronal plane of the three-dimensional medical image.
5. The method of claim 1 , wherein segmenting the first stack and the second stack of two-dimensional images includes:
creating a combined stack of two-dimensional images from the first stack and the second stack of two-dimensional images; and
segmenting the combined stack of two-dimensional images using one neural network model.
6. The method of claim 1 , wherein segmenting the first stack and the second stack of two-dimensional images includes determining a plurality of labels for a voxel of the three-dimensional medical image.
7. The method of claim 6 , wherein aggregating the segmentation results includes assigning to the voxel a label with a highest probability value among the plurality of labels.
8. The method of claim 1 , wherein the at least one neural network model is a deep convolutional neural network model.
9. The method of claim 1 , wherein the three-dimensional medical image is a medical image indicative of anatomical structures of a patient, wherein the label map associates an anatomic structure to each voxel of the three-dimensional medical image.
10. The method of claim 1 , further including training the at least one neural network model using at least one three-dimensional medical training image.
11. A device for segmenting a three-dimensional medical image, the device comprising:
an input interface that receives the three-dimensional medical image acquired by an imaging device;
at least one storage device configured to store the three-dimensional medical image; and
an image processor configured to:
create a first stack of two-dimensional images from a first plane of the three-dimensional medical image and a second stack of two-dimensional images from a second plane of the three-dimensional medical image;
segment the first stack and the second stack of two-dimensional images using at least one neural network model; and
determine a label map for the three-dimensional medical image by aggregating the segmentation results from the first stack and second stack.
12. The device of claim 11 , wherein the image processor is further configured to:
create a third stack of two-dimensional images from a third plane of the three-dimensional medical image;
segment the third stack of two-dimensional images using the at least one neural network model; and
determine the label map for the three-dimensional medical image by aggregating the segmentation results from the first stack, the second stack, and the third stack.
13. The device of claim 11 , wherein the image processor is further configured to segment the first stack of two-dimensional images using a first neural network model and segment the second stack of two-dimensional images using a second neural network model.
14. The device of claim 11 , wherein the first plane or the second plane is selected from an axial plane, a sagittal plane, and a coronal plane of the three-dimensional medical image.
15. The device of claim 11 , wherein the image processor is further configured to:
create a combined stack of two-dimensional images from the first stack and the second stack of two-dimensional images; and
segment the combined stack of two-dimensional images using one neural network model.
16. The device of claim 11 , wherein the image processor is further configured to determine a plurality of labels for a voxel of the three-dimensional medical image.
17. The device of claim 16 , wherein the image processor is further configured to assign to the voxel a label with a highest probability value among the plurality of labels.
18. The device of claim 11 , wherein the at least one neural network model is a deep convolutional neural network model.
19. The device of claim 11 , wherein the three-dimensional medical image is a medical image indicative of anatomical structures of a patient, wherein the label map associates an anatomic structure to each voxel of the three-dimensional medical image.
20. The device of claim 11 , wherein the image processor is further configured to train the at least one neural network model using at least one three-dimensional medical training image.
21. A non-transitory computer-readable medium containing instructions that, when executable by at least one processor, cause the at least one processor to perform a method for segmenting a three-dimensional medical image, the method comprising:
receiving the three-dimensional medical image acquired by an imaging device;
creating a first stack of two-dimensional images from a first plane of the three-dimensional medical image and a second stack of two-dimensional images from a second plane of the three-dimensional medical image;
segmenting, by the at least one processor, the first stack and the second stack of two-dimensional images using at least one neural network model; and
determining, by the at least one processor, a label map for the three-dimensional medical image by aggregating the segmentation results from the first stack and second stack.
22. The non-transitory computer-readable medium of claim 21 , wherein the method further comprises:
creating a third stack of two-dimensional images from a third plane of the three-dimensional medical image;
segmenting the third stack of two-dimensional images using the at least one neural network model; and
determining the label map for the three-dimensional medical image by aggregating the segmentation results from the first stack, the second stack, and the third stack.
23. The non-transitory computer-readable medium of claim 21 , wherein segmenting the first stack of two-dimensional images uses a first neural network model and segmenting the second stack of two-dimensional images uses a second neural network model.
24. The non-transitory computer-readable medium of claim 21 , wherein the first plane or the second plane is selected from an axial plane, a sagittal plane, and a coronal plane of the three-dimensional medical image.
25. The non-transitory computer-readable medium of claim 21 , wherein segmenting the first stack and the second stack of two-dimensional images comprises:
creating a combined stack of two-dimensional images from the first stack and the second stack of two-dimensional images; and
segmenting the combined stack of two-dimensional images using one neural network model.
26. The non-transitory computer-readable medium of claim 21 , wherein segmenting the first stack and the second stack of two-dimensional images includes determining a plurality of labels for a voxel of the three-dimensional medical image.
27. The non-transitory computer-readable medium of claim 26 , wherein aggregating the segmentation results includes assigning to the voxel a label with a highest probability value among the plurality of labels.
28. The non-transitory computer-readable medium of claim 21 , wherein the at least one neural network model is a deep convolutional neural network model.
29. The non-transitory computer-readable medium of claim 21 , wherein the three-dimensional medical image is a medical image indicative of anatomical structures of a patient, wherein the label map associates an anatomic structure to each voxel of the three-dimensional medical image.
30. The non-transitory computer-readable medium of claim 21 , wherein the method further comprises training the at least one neural network model using at least one three-dimensional medical training image.