3D segmentation with exponential logarithmic loss for highly unbalanced object sizes
View Patent ↗3D segmentation with exponential logarithmic loss for highly unbalanced object sizes is provided. In various embodiments, an artificial neural network is trained to label an anatomical feature in medical imagery by: i) providing at least one medical image to the artificial neural network; ii) determining from the artificial neural network a predicted segmentation for the at least one medical image; iii) comparing the predicted segmentation to ground truth segmentation, and computing therefrom a loss function, the loss function having an exponential-logarithmic term; and iv) updating the artificial neural network based on the loss function.
1. A method comprising:
training an artificial neural network to label at least one anatomical feature in medical imagery by:
i) providing at least one medical image to the artificial neural network;
ii) determining from the artificial neural network a predicted segmentation for the at least one medical image;
iii) comparing the predicted segmentation to ground truth segmentation, and computing therefrom a loss function, the loss function having an exponential-logarithmic term, in which a logarithmic expression is raised to an exponent; and
iv) updating the artificial neural network based on the loss function.
2. The method of claim 1 , further comprising:
repeating said training for a predetermined number of cycles.
3. The method of claim 1 , wherein the artificial neural network comprises a convolutional neural network having one or more synaptic weights, and updating the artificial neural network comprises updating the one or more synaptic weights.
4. The method of claim 1 , wherein computing the loss function comprises computing a logarithmic Dice loss.
5. The method of claim 1 , further comprising:
applying the artificial neural network to an input image;
determining from the artificial neural network a segmentation of the input image.
6. The method of claim 1 , wherein said updating comprises back-propagation.
7. The method of claim 1 , wherein the artificial neural network comprises a plurality of convolutional blocks, each comprising skip connection.
8. The method of claim 1 , wherein the artificial neural network comprises deep supervision.
9. The method of claim 1 , wherein the exponential-logarithmic term has an exponent between zero and one.
10. A system comprising:
a computing node comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor of the computing node to cause the processor to train an artificial neural network to label at least one anatomical feature in medical imagery by:
i) providing at least one medical image to the artificial neural network;
ii) determining from the artificial neural network a predicted segmentation for the at least one medical image;
iii) comparing the predicted segmentation to ground truth segmentation, and computing therefrom a loss function, the loss function having an exponential-logarithmic term, in which a logarithmic expression is raised to an exponent; and
iv) updating the artificial neural network based on the loss function.
11. The system of claim 10 , the method further comprising:
repeating said training for a predetermined number of cycles.
12. The system of claim 10 , wherein the artificial neural network comprises a convolutional neural network having one or more synaptic weights, and updating the artificial neural network comprises updating the one or more synaptic weights.
13. The system of claim 10 , wherein computing the loss function comprises computing a logarithmic Dice loss.
14. The system of claim 10 , the method further comprising:
applying the artificial neural network to an input image;
determining from the artificial neural network a segmentation of the input image.
15. The system of claim 10 , wherein said updating comprises back-propagation.
16. The system of claim 10 , wherein the artificial neural network comprises a plurality of convolutional blocks, each comprising skip connection.
17. The system of claim 10 , wherein the artificial neural network comprises deep supervision.
18. The system of claim 10 , wherein the exponential-logarithmic term has an exponent between zero and one.
19. A computer program product for labeling anatomical features in medical imagery, the computer program product comprising a computer readable storage medium having program instructions embodied therewith, the program instructions executable by a processor to cause the processor to perform a method comprising:
training an artificial neural network to label at least one anatomical feature in medical imagery by:
i) providing at least one medical image to the artificial neural network;
ii) determining from the artificial neural network a predicted segmentation for the at least one medical image;
iii) comparing the predicted segmentation to ground truth segmentation, and computing therefrom a loss function, the loss function having an exponential-logarithmic term, in which a logarithmic expression is raised to an exponent; and
iv) updating the artificial neural network based on the loss function.
20. The computer program product of claim 19 , the method further comprising:
applying the artificial neural network to an input image;
determining from the artificial neural network a segmentation of the input image.