Automated segmentation of three dimensional bony structure images
A computer-implemented system: at least one processor communicably coupled to at least one nontransitory processor-readable storage medium storing processor-executable instructions or data receives segmentation learning data comprising a plurality of batches of labeled anatomical image sets, each image set comprising image data representative of a series of slices of a three-dimensional bony structure, and each image set including at least one label which identifies the region of a particular part of the bony structure depicted in each image of the image set, wherein the label indicates one of a plurality of classes indicating parts of the bone anatomy; trains a segmentation CNN, that is a fully convolutional neural network model with layer skip connections, to segment semantically at least one part of the bony structure utilizing the received segmentation learning data; and stores the trained segmentation CNN in at least one nontransitory processor-readable storage medium of the machine learning system.
1. A computer-implemented machine learning system, comprising:
at least one nontransitory processor-readable storage medium that stores at least one of processor-executable instructions or data; and
at least one processor communicably coupled to the at least one nontransitory processor-readable storage medium, wherein the at least one processor:
receives segmentation learning data comprising a plurality of batches of labeled anatomical image sets, each image set of the plurality of batches of labeled anatomical image sets comprising image data representative of a series of slices of a three-dimensional (3D) bony structure, and each image set of the plurality of batches of labeled anatomical image sets including at least one label which identifies a region of a particular part of the 3D bony structure depicted in each image of the image set, wherein each label indicates one of a plurality of classes indicating parts of the 3D bony structure;
trains a segmentation convolutional neural network (CNN), that is a fully convolutional neural network model with layer skip connections, to segment semantically at least one part of the 3D bony structure utilizing the received segmentation learning data; and
stores the trained segmentation CNN in the at least one nontransitory processor-readable storage medium of the machine learning system.
2. The system according to claim 1 , wherein the at least one processor further:
receives denoising learning data comprising a plurality of batches of high quality medical images and low quality medical images, wherein the high quality medical images have a lower noise level than the low quality medical images;
trains a denoising CNN, that is a fully convolutional neural network model with layer skip connections, to denoise an image utilizing the received denoising learning data; and
stores the trained denoising CNN in the at least one nontransitory processor-readable storage medium of the machine learning system.
3. The system according to claim 1 , wherein the at least one processor further operates the trained segmentation CNN to process a set of input anatomical images to generate a set of output segmented anatomical images.
4. The system according to claim 2 , wherein the at least one processor further operates the trained segmentation CNN to process a set of input anatomical images to generate a set of output segmented anatomical images.
5. The system according to claim 4 , wherein the at least one processor further operates the trained denoising CNN to process a set of input anatomical images to generate a set of output denoised anatomical images.
6. The system according to claim 5 , wherein the set of input anatomical images processed by the trained denoising CNN comprises low quality anatomical images.
7. The system according to claim 6 , wherein the set of input anatomical images processed by the trained segmentation CNN comprises the set of output denoised anatomical images of the denoising CNN.
8. The system according claim 2 , wherein the low quality anatomical images are low-dose computer tomography (LDCT) images or low power magnetic resonance images, and wherein the high quality anatomical images are high-dose computer tomography (HDCT) images or high power magnetic resonance images.
9. The system of claim 1 , wherein the segmentation learning data comprises preoperative or intraoperative anatomical images of the 3D bony structure.
10. The system of claim 2 , wherein at least one of the segmentation learning data and the denoising learning data comprise preoperative or intraoperative anatomical images of the 3D bony structure.
11. The system of claim 1 , wherein the segmentation learning data comprises labeled anatomical parts of the 3D bony structure.
12. The system of claim 3 , wherein the at least one processor is further configured to combine the set of output segmented anatomical images with a 3D volume of patient anatomy.
13. A method of operating a machine learning system comprising at least one nontransitory processor-readable storage medium that stores at least one of processor-executable instructions or data, and at least one processor communicably coupled to the at least one nontransitory processor-readable storage medium, the method comprising:
receiving, by the at least one processor, segmentation learning data comprising a plurality of batches of labeled anatomical image sets, each image set of the plurality of batches of labeled anatomical image sets comprising image data representative of a series of slices of a three-dimensional (3D) bony structure, and each image set of the plurality of batches of labeled anatomical image sets including at least one label which identifies the region of a particular part of the 3D bony structure depicted in each image of the image set, wherein each label indicates one of a plurality of classes indicating parts of the 3D bony structure;
training, by the at least one processor, a segmentation convolutional neural network (CNN), that is a fully convolutional neural network model with layer skip connections, to segment semantically at least one part of the 3D bony structure utilizing the received segmentation learning data; and
storing, by the at least one processor, the trained segmentation CNN in the at least one nontransitory processor-readable storage medium of the machine learning system.
14. The method according to claim 13 , further comprising:
receiving denoising learning data comprising a plurality of batches of high quality medical images and low quality medical images, wherein the high quality medical images have a lower noise level than the low quality medical images;
training a denoising CNN, that is a fully convolutional neural network model with layer skip connections, to denoise an image utilizing the received denoising learning data; and
storing the trained denoising CNN in the at least one nontransitory processor-readable storage medium of the machine learning system.
15. The method according to claim 14 , further comprising operating the trained denoising CNN to process a set of input anatomical images to generate a set of output denoised anatomical images, wherein the set of input anatomical images processed by the trained denoising CNN comprises low quality medical images.
16. The method according to claim 15 , further comprising operating the trained segmentation CNN to process a set of input anatomical images to generate a set of output segmented anatomical images, wherein the set of input anatomical images processed by the trained segmentation CNN comprises the set of output denoised anatomical images of the denoising CNN.
17. The method according to claim 16 , further comprising combining set of the output segmented anatomical images with a 3D volume of patient anatomy.
18. The method of claim 13 , wherein the image data is first image data, the method further comprising:
receiving, by the at least one processor, second image data which represents a 3D bony structure;
processing, by the at least one processor, the second image data through the segmentation CNN to generate per-class probabilities for the plurality of classes for each pixel of each image of the second image data;
for each image of the second image data, generating, by the at least one processor, a probability map for each of the plurality of classes using the generated per-class probabilities; and
storing, by the at least one processor, the generated probability map for each image of the second image data in the at least one nontransitory processor-readable storage medium.
19. The method of claim 18 , wherein the segmentation CNN includes a contracting path and an expanding path.
20. The method of claim 19 , wherein the contracting path includes a number of convolutional layers and a number of pooling layers in which each pooling layer is preceded by at least one convolutional layer.