US 12,243,298 B2
System, method and apparatus for obtaining sensitive and specific predictions from deep neural networks
Matthew Toews, Montréal (CA); and James Lee, Pointe-Claire (CA)
Assigned to LITTLE ANGEL MEDICAL INC., Pointe-Claire (CA)
Filed by LITTLE ANGEL MEDICAL INC., Pointe-Claire (CA)
Filed on Mar. 1, 2022, as Appl. No. 17/684,002.
Claims priority of provisional application 63/161,926, filed on Mar. 16, 2021.
Prior Publication US 2022/0301297 A1, Sep. 22, 2022
Int. Cl. G06V 10/82 (2022.01)
CPC G06V 10/82 (2022.01) 18 Claims
OG exemplary drawing
 
17. An apparatus for encoding activations generated by a deep neural network from an image, comprising:
an acquisition device configured to acquire image data samples;
a memory configured to store other activation vectors associated with label and non-image data;
a processor configured to:
generate, using a deep neural network, for each image data sample, an activation layer comprising a set of activation vectors;
encode the set of activation vectors of the generated activation layer;
perform activation vector correspondence between at least one of the encoded activation vectors and at least one of the other activation vectors based on activation vector distance and conditioned by non-image data; and
predict at least one classification for each image data sample based on the activation vector correspondence.