Few-shot learning based image recognition of whole slide image at tissue level

An apparatus for generating at least one shape of a region of interest in a digital image, the apparatus comprising:

• a non-transitory, computer readable memory storing software instructions;
• and
• a processor coupled with the computer readable memory, and, upon execution of the software instructions, is configurable to:
• obtain access to a digital tissue image of a biological sample;
• tile the digital tissue image into a collection of image patches;
• obtain a plurality of features from each patch in the collection of image patches, the plurality of features defining a patch feature vector in a multidimensional feature space including the plurality of features as dimensions;
• determine a user selection of a user selected subset of patches in the collection of image patches, wherein the user selection further includes a selection, prior to selection of the subset of patches in the collection of image patches, of whether the user selected subset of patches are positive or negative patches;
• classify, by applying a trained classifier to patch vectors of other patches in the collection of patches, the other patches as belonging or not belonging to a same class of interest as the user selected subset of patches;
• and
• identify one or more regions of interest based at least in part on the results of the classifying.

A computer implemented method of generating training data to train a machine learning system, the method comprising:

• obtaining, by an image processing engine, access to a digital tissue image of a biological sample;
• tiling, by the image processing engine, the digital tissue image into a collection of image patches;
• obtaining, by the image processing engine, a plurality of features from each patch in the collection of image patches, the plurality of features defining a patch feature vector in a multidimensional feature space including the plurality of features as dimensions;
• determining, by the image processing engine, a user selection of a user selected subset of patches in the collection of image patches, wherein the user selection further includes a selection, prior to selection of the subset of patches in the collection of image patches, of whether the user selected subset of patches are positive or negative patches;
• classifying, by applying a trained generic classifier to patch vectors of other patches in the collection of patches, the other patches as belonging or not belonging to a same class of interest as the user selected subset of patches;
• and
• generating, by the image processing engine, training data for the machine learning system based on the classifying of the user-selected and the other patches in the collection of patches.