Anatomical segmentation identifying modes and viewpoints with deep learning across modalities
A mechanism is provided in a data processing system comprising a processor and a memory, the memory comprising instructions that are executed by the processor to specifically configure the processor to implement a multi-modal classification and segmentation engine for anatomical segmentation identifying modes and viewpoints in biomedical images. The mechanism trains a neural network perform simultaneous classification and segmentation using a set of training images. The neural network provides a classification output that identifies a class label and a second output that identifies a segmentation label. The multi-modal classification and segmentation engine provides a biomedical image as the input image to the neural network. The neural network outputs a plurality of class label probabilities for a plurality of class labels and a plurality of segmentation label probabilities for each of a plurality of segmentation labels. A post-processing component executing within the multi-modal classification and segmentation engine classifies the biomedical image as an identified modality and an identified viewpoint based on the plurality of class label probabilities. The multi-modal classification and segmentation engine segments the biomedical image based on the plurality of segmentation label probabilities. The multi-modal classification and segmentation engine outputs the classified and segmented biomedical image.
1. A method, in a data processing system comprising a processor and a memory, the memory comprising instructions that are executed by the processor to specifically configure the processor to implement a multi-modal classification and segmentation engine for anatomical segmentation identifying modes and viewpoints in biomedical images, the method comprising:
training a single neural network to perform simultaneous classification and segmentation using a set of training images, wherein the neural network comprises a classification head and a segmentation head, wherein the classification head generates a plurality of class label probabilities for a plurality of class labels, and wherein the segmentation head generates a plurality of segmentation label probabilities for a plurality of segmentation labels;
providing, by the multi-modal classification and segmentation engine, a biomedical image as the input image to the neural network;
outputting, by the neural network, the plurality of class label probabilities and the plurality of segmentation label probabilities based on the input image;
classifying, by a post-processing component executing within the multi-modal classification and segmentation engine, the biomedical image as an identified modality and an identified viewpoint based on the plurality of class label probabilities;
segmenting, by the multi-modal classification and segmentation engine, the biomedical image based on the plurality of segmentation label probabilities; and
outputting, by the multi-modal classification and segmentation engine, the classified and segmented biomedical image.
2. The method of claim 1 , wherein the set of training images comprise a plurality of biomedical images covering a plurality of modalities, viewpoints, and segmentations.
3. The method of claim 1 , wherein the neural network comprises a convolutional neural network.
4. The method of claim 3 , wherein the convolutional neural network comprises a U-net with multiple heads.
5. The method of claim 4 , wherein the convolutional neural network comprises a concatenating path and an expanding path, wherein the convolutional neural network comprises the classification head at the end of the concatenating path and the segmentation head at the end of the expanding path.
6. The method of claim 1 , wherein the plurality of class labels comprises class labels for a plurality of modalities.
7. The method of claim 6 , wherein the plurality of modalities comprise X-ray, magnetic resonance imaging (MRI), computerized tomography (CT), and ultrasound.
8. The method of claim 7 , wherein the plurality of class labels comprise chest X-ray (CXR), short axis MRI (SAX), two-channel MRI (2Ch), four-channel MRI (4Ch), CT, brightness modulation ultrasound (Bmod), and Doppler ultrasound (Dop).
9. The method of claim 1 , wherein the plurality of segmentation labels comprise lung, heart, myocardium, left ventricle, right ventricle, left atrium, right atrium, and Doppler waveform.
10. The method of claim 1 , wherein the neural network generates an image map for each segmentation label, wherein the image map comprises for each pixel a probability that the pixel is part of an organ or structure corresponding to the segmentation, wherein segmenting the biomedical image comprises comparing each probability to a threshold.
11. A computer program product comprising a non-transitory computer readable medium having a computer readable program stored therein, wherein the computer readable program, when executed on a computing device, causes the computing device to implement a multi-modal classification and segmentation engine for anatomical segmentation identifying modes and viewpoints in biomedical images, wherein the computer readable program causes the computing device to:
train a single neural network to perform simultaneous classification and segmentation using a set of training images, wherein the neural network comprises a classification head and a segmentation head, wherein the classification head generates a plurality of class label probabilities for a plurality of class labels, and wherein the segmentation head generates a plurality of segmentation label probabilities for a plurality of segmentation labels;
provide, by the multi-modal classification and segmentation engine, a biomedical image as the input image to the neural network;
output, by the neural network, the plurality of class label probabilities and the plurality of segmentation label probabilities based on the input image;
classify, by a post-processing component executing within the multi-modal classification and segmentation engine, the biomedical image as an identified modality and an identified viewpoint based on the plurality of class label probabilities;
segment, by the multi-modal classification and segmentation engine, the biomedical image based on the plurality of segmentation label probabilities; and
output, by the multi-modal classification and segmentation engine, the classified and segmented biomedical image.
12. The computer program product of claim 11 , wherein the set of training images comprise a plurality of biomedical images covering a plurality of modalities, viewpoints, and segmentations.
13. The computer program product of claim 11 , wherein the neural network comprises a convolutional neural network.
14. The computer program product of claim 13 , wherein the convolutional neural network comprises a U-net with multiple heads.
15. The computer program product of claim 14 , wherein the convolutional neural network comprises a concatenating path and an expanding path, wherein the convolutional neural network comprises the classification head at the end of the concatenating path and the segmentation head at the end of the expanding path.
16. The computer program product of claim 11 , wherein the plurality of class labels comprises class labels for a plurality of modalities.
17. The computer program product of claim 16 , wherein the plurality of modalities comprises X-ray, magnetic resonance imaging (MRI), computerized tomography (CT), and ultrasound.
18. The computer program product of claim 17 , wherein the plurality of class labels comprise chest X-ray (CXR), short axis MRI (SAX), two-channel MRI (2Ch), four-channel MRI (4Ch), CT, brightness modulation ultrasound (Bmod), and Doppler ultrasound (Dop).
19. The computer program product of claim 11 , wherein the plurality of segmentation labels comprise lung, heart, myocardium, left ventricle, right ventricle, left atrium, right atrium, and Doppler waveform.
20. An apparatus, comprising:
a processor; and
a memory coupled to the processor, wherein the memory comprises instructions which, when executed by the processor, cause the processor to implement a multi-modal classification and segmentation engine for anatomical segmentation identifying modes and viewpoints in biomedical images, wherein the instructions cause the processor to:
train a single neural network to perform simultaneous classification and segmentation using a set of training images, wherein the neural network comprises a classification head and a segmentation head, Wherein the classification head generates a plurality of class label probabilities for a plurality of class labels, and wherein the segmentation head generates a plurality of segmentation label probabilities for a plurality of segmentation labels;
provide, by the multi-modal classification and segmentation engine, a biomedical image as the input image to the neural network;
output, by the neural network, the plurality of class label probabilities and the plurality of segmentation label probabilities based on the input image;
classify, by a post-processing component executing within the multi-modal classification and segmentation engine, the biomedical image as an identified modality and an identified viewpoint based on the plurality of class label probabilities;
segment, by the multi-modal classification and segmentation engine, the biomedical image based on the plurality of segmentation label probabilities; and
output, by the multi-modal classification and segmentation engine, the classified and segmented biomedical image.