System and method for iterative classification using neurophysiological signals

1. A method of training an image classification neural network, the method comprising:

presenting a first plurality of images to an observer as a visual stimulus, while collecting neurophysiological signals from a brain of said observer;

processing said neurophysiological signals to identify a neurophysiological event indicative of a detection of a target by said observer in at least one image of said first plurality of images;

training the image classification neural network to identify the target in the image, based on said identification of said neurophysiological event; and

storing said trained image classification neural network in a non-transitory computer-readable storage medium;

wherein the method comprises tiling an input image into a plurality of image tiles, wherein said first plurality of images comprises a portion of said plurality of image tiles.

2. The method according to claim 1 , further comprising applying unsupervised clustering to a second plurality of images, and selecting said first plurality of images from said second plurality of images based on said unsupervised clustering.

3. The method according to claim 2 , wherein said second plurality of images comprises said plurality of image tiles.

4. The method according to claim 1 , comprising randomly selecting said first plurality of images from a second plurality of images.

5. The method according to claim 1 , comprising receiving a reference image including said target, and selecting said first plurality of images from a second plurality of images responsively to said reference image.

6. The method according to claim 1 , further comprising:

applying said trained image classification neural network to said second plurality of images to detect therein candidate images suspected as being occupied by said target, wherein said second plurality of images comprises at least one image of said first plurality of images;

re-defining said second plurality of images, wherein at least one image of said redefined second plurality of images is a candidate image as detected by said trained image classification neural network; and

repeating said presentation, said collection and processing of said neurophysiological signals, and said training for at least one image of said redefined first plurality of images, thereby iteratively training the image classification neural network.

7. The method according to claim 6 , further comprising tiling an input image into a plurality of image tiles, wherein said second plurality of images comprises said plurality of image tiles, and wherein said first plurality of images comprises a portion of said plurality of image tiles.

8. The method according to claim 7 , wherein said re-defining said second plurality of images, comprises re-tiling said input image into a plurality of images, wherein at least one image of said retiled input image comprises said candidate image.

9. The method according to claim 6 , further comprising applying unsupervised clustering to said second plurality of images, and selecting said first plurality of images from said second plurality of images based on said unsupervised clustering.

10. The method according to claim 6 , comprising randomly selecting first plurality of images from said second plurality of images.

11. The method according to claim 1 , comprising receiving a reference image including the target, and selecting first plurality of images from said second plurality of images responsively to said reference image.

12. The method according to claim 1 , wherein said image classification neural network is a convolutional neural network.

13. The method according to claim 1 , wherein said image classification neural network comprises a first neural subnetwork configured for receiving and processing said neurophysiological data, a second neural subnetwork configured for receiving and processing said second plurality of images, and a shared subnetwork having a neural network layer receiving and combining outputs from both said first neural subnetwork and said second neural subnetwork.

14. The method according to claim 13 , wherein said image classification neural network is a convolutional neural network and at least one of said first and said second neural subnetworks is a convolutional neural subnetwork.

15. The method according to claim 13 , wherein said image classification neural network comprises a first separate output layer for said first neural subnetwork outputting a first score, and second separate output layer for said second neural subnetwork outputting a second score, and wherein the method comprises combining said first score with said second score to a combined score, labeling said image with said combined score, and using said label in at least one iteration of said training.

16. The method according to claim 15 , wherein said combined score is a weighted sum of said first and said second score.

17. The method according to claim 1 , wherein said image classification neural network comprises an autoencoder subnetwork for unsupervised feature learning.

18. The method according to claim 17 , wherein said autoencoder subnetwork is used for selecting said first plurality of images.

19. The method according to claim 17 , wherein said autoencoder subnetwork is used only for selecting said first plurality of images, but not for said training based on said identification of said neurophysiological event.

20. The method according to claim 1 , further comprising scoring said neurophysiological event using said neurophysiological signals, wherein said training is based at least in part on said score.

21. The method according to claim 20 , further comprising using said score for determining a level of similarity of said target to an object in an image observed by said observer.

22. The method according to claim 1 , further comprising presenting to the observer a stimulus describing said target prior to said presentation of said first plurality of images, wherein said stimulus comprises at least one stimulus selected from the group consisting of a visual stimulus, a textual stimulus, an auditory stimulus and an olfactory stimulus.

23. The method according to claim 1 , further comprising presenting to the observer a stimulus complementary to said target prior to said presentation of said first plurality of images, wherein said stimulus comprises at least one stimulus selected from the group consisting of a visual stimulus, a textual stimulus, an auditory stimulus and an olfactory stimulus.

24. A system for training an image classification neural network, comprising:

a neurophysiological signal collection system, configured for collecting neurophysiological signals from a brain of an observer; and

a data processor, communicating with said neurophysiological signal collection system and being configured for executing the method according to claim 1 .

25. A method of classifying an image, comprising:

executing the method according to claim 1 to train the image classification neural network; and

applying said trained image classification neural network to the image to determine an existence of said target in the image based on a score generated by an output layer of said trained image classification neural network.

26. A method of classifying an image, comprising:

applying the method according to claim 1 to the image;

applying said trained image classification neural network to the image to determine whether the image is suspected as being occupied by a target;

presenting the image to the observer as a visual stimulus, while collecting neurophysiological signals from a brain of said observer;

determining an existence of said target in the image based, at least in part, on said identification of said neurophysiological event.

27. A method of training an image classification neural network, the method comprising:

presenting a first plurality of images to an observer as a visual stimulus, while collecting neurophysiological signals from a brain of said observer;

processing said neurophysiological signals to identify a neurophysiological event indicative of a detection of a target by said observer in at least one image of said first plurality of images;

training the image classification neural network to identify the target in the image, based on said identification of said neurophysiological event; and

storing said trained image classification neural network in a non-transitory computer-readable storage medium;

wherein the method comprises:

applying said trained image classification neural network to a second plurality of images to detect therein candidate images suspected as being occupied by said target, wherein said second plurality of images comprises at least one image of said first plurality of images;

re-defining said second plurality of images, wherein at least one image of said redefined second plurality of images is a candidate image as detected by said trained image classification neural network; and

repeating said presentation, said collection and processing of said neurophysiological signals, and said training for at least one image of said redefined first plurality of images, thereby iteratively training the image classification neural network.

28. A method of training an image classification neural network, the method comprising:

presenting a first plurality of images to an observer as a visual stimulus, while collecting neurophysiological signals from a brain of said observer;

processing said neurophysiological signals to identify a neurophysiological event indicative of a detection of a target by said observer in at least one image of said first plurality of images;

training the image classification neural network to identify the target in the image, based on said identification of said neurophysiological event; and

storing said trained image classification neural network in a non-transitory computer-readable storage medium;

wherein said image classification neural network comprises a first neural subnetwork configured for receiving and processing said neurophysiological data, a second neural subnetwork configured for receiving and processing said second plurality of images, and a shared subnetwork having a neural network layer receiving and combining outputs from both said first neural subnetwork and said second neural subnetwork;

wherein said image classification neural network comprises a first separate output layer for said first neural subnetwork outputting a first score, and second separate output layer for said second neural subnetwork outputting a second score; and

wherein the method comprises combining said first score with said second score to a combined score, labeling said image with said combined score, and using said label in at least one iteration of said training.

29. A method of training an image classification neural network, the method comprising:

presenting a first plurality of images to an observer as a visual stimulus, while collecting neurophysiological signals from a brain of said observer;

processing said neurophysiological signals to identify a neurophysiological event indicative of a detection of a target by said observer in at least one image of said first plurality of images;

training the image classification neural network to identify the target in the image, based on said identification of said neurophysiological event; and

storing said trained image classification neural network in a non-transitory computer-readable storage medium;

wherein said image classification neural network comprises an autoencoder subnetwork for unsupervised feature learning; and

wherein said autoencoder subnetwork is used only for selecting said first plurality of images, but not for said training based on said identification of said neurophysiological event.

30. A method of classifying an image, comprising:

training an image classification neural network to provide a trained image classification neural network; and

applying said trained image classification neural network to the image to determine an existence of said target in the image based on a score generated by an output layer of said trained image classification neural network;

wherein training comprises:

presenting a first plurality of images to an observer as a visual stimulus, while collecting neurophysiological signals from a brain of said observer;

processing said neurophysiological signals to identify a neurophysiological event indicative of a detection of a target by said observer in at least one image of said first plurality of images;

training the image classification neural network to identify the target in the image, based on said identification of said neurophysiological event; and

storing said trained image classification neural network in a non-transitory computer-readable storage medium.

31. A method of classifying an image, comprising:

training an image classification neural network to provide a trained image classification neural network;

applying said trained image classification neural network to the image to determine whether the image is suspected as being occupied by a target;

presenting the image to the observer as a visual stimulus, while collecting neurophysiological signals from a brain of said observer;

determining an existence of said target in the image based, at least in part, on said identification of said neurophysiological event;

wherein training comprises:

presenting a first plurality of images to an observer as a visual stimulus, while collecting neurophysiological signals from a brain of said observer;

processing said neurophysiological signals to identify a neurophysiological event indicative of a detection of a target by said observer in at least one image of said first plurality of images;

training the image classification neural network to identify the target in the image, based on said identification of said neurophysiological event; and

storing said trained image classification neural network in a non-transitory computer-readable storage medium.