Methods and systems for providing guidance for adjusting an object based on similarity

1. A computer-implemented method for providing guidance for adjusting a target, the method comprising:

receiving, by a neural network, a reference image;

receiving, by the neural network, a target image, the target image being related to a position of a target;

determining a similarity metric based at least in part on information associated with the reference image and information associated with the target image by the neural network;

generating a target attention map corresponding to the target image based at least in part on the similarity metric;

outputting the target image and the target attention map; and

providing a guidance for adjusting the position of the target based at least in part on the target image and the target attention map;

wherein the target attention map includes one or more high response regions that contributes substantially to the similarity metric;

wherein the one or more high response regions correspond to one or more similar regions between the target image and the reference image when the one or more high response regions constitutes less than a coverage threshold of the target image; and

wherein the one or more high response regions correspond to one or more dissimilar regions between the target image and the reference image when the one or more high response regions constitutes more than the coverage threshold of the target image.

2. The computer-implemented method of claim 1 , further comprising:

processing the target attention map; and

overlaying the processed target attention map onto the target image.

3. The computer-implemented method of claim 1 , further comprising:

generating a target feature map corresponding to the target image by the neural network;

wherein the generating a target attention map corresponding to the target image includes generating the target attention map corresponding to the target image based at least in part on the similarity metric and the target feature map.

4. The computer-implemented method of claim 3 , further comprising:

generating a reference feature map corresponding to the reference image by the neural network; and

generating a reference attention map corresponding to the reference image based at least in part on the similarity metric and the reference feature map.

5. The computer-implemented method of claim 4 , further comprising:

processing the reference attention map; and

overlaying the processed reference attention map onto the reference image.

6. The computer-implemented method of claim 4 , further comprising:

generating a reference feature vector based at least in part on the reference feature map; and

generating a target feature vector based at least in part on the target feature map;

wherein the determining a similarity metric based at least in part on information associated with the reference image and information associated with the target image includes determining the similarity metric based at least in part on the reference feature vector and the target feature vector by the neural network.

7. The computer-implemented method of claim 6 , wherein:

the determining the similarity metric based at least in part on the reference feature vector and the target feature vector includes determining a Euclidean distance between the reference feature vector and the target feature vector; and

the similarity metric is a similarity score.

8. The computer-implemented method of claim 1 , wherein the receiving a reference image into a neural network includes:

receiving a protocol; and

selecting the reference image based at least in part on the protocol.

9. The computer-implemented method of claim 8 , wherein the receiving a target image into the neural network includes:

acquiring the target image based at least in part on the protocol.

10. A system for providing guidance for adjusting a target for imaging, the system comprising:

an image acquisition apparatus configured to acquire a target image, the target image being related to a position of a target;

an image processing apparatus configured to:

receive, by a neural network, a reference image;

receive, by the neural network, the target image;

determine a similarity metric based at least in part on information associated with the reference image and information associated with the target image by the neural network;

generate a target attention map corresponding to the target image based at least in part on the similarity metric; and

output the target image and the target attention map; and

a display apparatus configured to:

receive the target image and the target attention map; and

provide a guidance for adjusting the position of the target based at least in part on the target image and the target attention map;

wherein the target attention map includes one or more high response regions that contributes substantially to the similarity metric;

wherein the one or more high response regions correspond to one or more similar regions between the target image and the reference image when the one or more high response regions constitutes less than a coverage threshold of the target image; and

wherein the one or more high response regions correspond to one or more dissimilar regions between the target image and the reference image when the one or more high response regions constitutes more than the coverage threshold of the target image.

11. A computer-implemented method for training a neural network, the method comprising:

receiving, by a neural network, a first input image;

receiving, by the neural network, a second input image;

determining a first similarity metric based at least in part on information associated with the first input image and information associated with the second input image by the neural network;

generating a first attention map corresponding to the first input image based at least in part on the first similarity metric;

generating a second attention map corresponding to the second input image based at least in part on the first similarity metric;

modifying the first input image based at least in part on the first attention map to generate a first modified image;

modifying the second input image based at least in part on the second attention map to generate a second modified image;

determining a second similarity metric based at least in part on information associated with the first modified image and information associated with the second modified image by the neural network; and

changing one or more parameters of the neural network based at least in part on the first similarity metric and the second similarity metric;

wherein the changing one or more parameters of the neural network includes:

increasing the first similarity metric; and

decreasing the second similarity metric;

wherein the first attention map includes one or more high response regions that contributes substantially to the first similarity metric;

wherein the one or more high response regions correspond to one or more similar regions between the first input image and the second input image when the one or more high response regions constitutes less than a coverage threshold of the first input image; and

wherein the one or more high response regions correspond to one or more dissimilar regions between the first input image and the second input image when the one or more high response regions constitutes more than the coverage threshold of the first input image.

12. The computer-implemented method of claim 11 , further comprising:

generating a first input feature map corresponding to the first input image by the neural network; and

generating a second input feature map corresponding to the second input image by the neural network;

wherein the generating a first attention map corresponding to the first input image includes generating the first attention map corresponding to the first input image based at least in part on the first similarity metric and the first feature map;

wherein the generating a second attention map corresponding to the second input image includes generating the second attention map corresponding to the second input image based at least in part on the first similarity metric and the second feature map.

13. The computer-implemented method of claim 12 wherein:

the generating the first attention map corresponding to the first input image based at least in part on the first similarity metric and the first input feature map includes determining first one or more derivatives of the first similarity metric with respect to the first input feature map; and

the generating the second attention map corresponding to the second input image based at least in part on the first similarity metric and the second input feature map includes determining second one or more derivatives of the first similarity metric with respect to the second input feature map.

14. The computer-implemented method of claim 12 , further comprising:

generating a first input feature vector based at least in part on the first input feature map; and

generating a second input feature vector based at least in part on the second input feature map;

wherein the determining a first similarity metric based at least in part on information associated with the first input image and information associated with the second input image includes determining the first similarity metric based at least in part on the first input feature vector and the second input feature vector by the neural network.

15. The computer-implemented method of claim 14 , wherein:

the determining the first similarity metric based at least in part on the first input feature vector and the second input feature vector includes determining a Euclidean distance between the first input feature vector and the second input feature vector; and

the first similarity metric is a similarity score.

16. The computer-implemented method of claim 11 , wherein:

the modifying the first input image based at least in part on the first attention map includes modifying the first input image based at least in part on one or more high response regions of the first attention map; and

the modifying the second input image based at least in part on the second attention map includes modifying the second input image based at least in part on one or more high response regions of the second attention map.

17. The computer-implemented method of claim 11 , further comprising:

generating a first modified feature map corresponding to the first modified image by the neural network; and

generating a second modified feature map corresponding to the second modified image by the neural network.

18. The computer-implemented method of claim 17 , further comprising:

generating a first modified feature vector corresponding to the first modified feature map; and

generating a second modified feature vector corresponding to the second modified feature map;

wherein the determining a second similarity metric includes determining the second similarity metric based at least in part on the first modified feature vector and the second modified feature vector by the neural network.

19. The computer-implemented method of claim 11 , wherein:

the increasing the first similarity metric includes maximizing the first similarity metric; and

the decreasing the second similarity metric includes minimizing the second similarity metric.

20. The computer-implemented method of claim 11 , further comprising:

receiving, by the neural network, a first validation image;

receiving, by the neural network, a second validation image;

determining a validation-similarity metric based at least in part on information associated with the first validation image and information associated with the second validation image by the neural network;

determining whether the validation-similarity metric satisfies one or more predetermined thresholds;

if the validation-similarity metric satisfies the one or more predetermined thresholds, determining the neural network to be ready for use; and

if the validation-similarity metric does not satisfy the one or more predetermined thresholds, determining the neural network to be not ready for use.