SYSTEMS AND METHODS FOR PROCESSING ELECTRONIC IMAGES TO INFER BIOMARKERS
Systems and methods are disclosed for receiving a target electronic image corresponding to a target specimen, the target specimen comprising a tissue sample of a patient, applying a machine learning system to the target electronic image to identify a region of interest of the target specimen and determine an expression level of, category of, and/or presence of a biomarker in the region of interest, the biomarker comprising at least one from among an epithelial growth factor receptor (EGFR) biomarker and/or a DNA mismatch repair (MMR) deficiency biomarker, the machine learning system having been generated by processing a plurality of training images to predict whether a region of interest is present in the target electronic image, the training images comprising images of human tissue and/or images that are algorithmically generated, and outputting the determined expression level of, category of, and/or presence of the biomarker in the region of interest.
1 - 20 . (canceled)
21 . A method, comprising:
segmenting, by one or more processors, an image into a plurality of tiles;
grouping, by the one or more processors, the plurality of tiles into at least one bag of tiles;
inputting, by the one or more processors, the at least one bag of tiles into a machine-learning model trained to generate a prediction of an image class label based on the at least one bag of tiles, the machine-learning model including:
a first portion trained to generate one or more feature maps based on the at least one bag of tiles; and
a second portion trained to generate the prediction of the image class label based at least in part on the one or more feature maps; and
outputting, by the one or more processors, the prediction of the image class label.
22 . The method of claim 21 , wherein the image comprises only one whole-slide image (WSI).
23 . The method of claim 21 , further comprising receiving, by the one or more processors, the image, wherein the image comprises an image of a tissue sample.
24 . The method of claim 21 , wherein each tile of the plurality of tiles comprises a plurality of pixels corresponding to one or more regions of the image.
25 . The method of claim 21 , wherein the image comprises a histological stain image, a fluorescence in situ hybridization (FISH) image, an immunofluorescence (IF) image, or a hematoxylin and eosin (H&E) image.
26 . The method of claim 21 , wherein:
the first portion comprises one or more convolutional layers; and
and the second portion comprises an output layer.
27 . The method of claim 26 , wherein the machine-learning model further comprises a pooling layer and a fully connected layer.
28 . The method of claim 21 , wherein the machine-learning model comprises one or more convolutional neural networks (CNNs), a multiple-instance learning (MIL) machine-learning model, or a multiple-instance learning convolutional neural network (MILCNN) machine-learning model.
29 . The method of claim 21 , wherein the machine-learning model was trained by:
receiving, by the one or more processors, a training image;
segmenting, by the one or more processors, the training image into a second plurality of tiles;
grouping, by the one or more processors, the second plurality of tiles into at least one second bag of tiles; and
inputting, by the one or more processors, the at least one second bag of tiles into the machine-learning model to generate a prediction of a second image class label based on the at least one second bag of tiles;
wherein:
the first portion is trained to generate one or more second feature maps based on the at least one second bag of tiles; and
the second portion is trained to generate the prediction of the second image class label for the training image based at least in part on the one or more second feature maps.
30 . The method of claim 29 , wherein each tile of the second plurality of tiles comprises a plurality of pixels corresponding to one or more regions of the training image.
31 . The method of claim 29 , wherein:
the first portion comprises one or more convolutional layers; and the second portion comprises an output layer.
32 . The method of claim 29 , wherein segmenting the training image into at least one second bag of tiles comprises randomly sampling one or more tiles of pixels of the at least one second bag of tiles.
33 . The method of claim 21 , wherein the image class label comprises an indication of a genetic biomarker of a tissue sample captured in the image.
34 . A method of treating subject with cancer, comprising:
characterizing a tissue sample comprising the cancer from the subject as having a genetic biomarker according to the method of claim 21 ; and
displaying one or more treatment for the cancer based on the tissue sample having the genetic biomarker.
35 . A system including one or more computing devices, comprising:
one or more non-transitory computer-readable storage media including instructions;
and one or more processors coupled to the one or more storage media, the one or more processors configured to execute the instructions to:
segment an image into a plurality of tiles;
group the plurality of tiles into at least one bag of tiles; and
input the at least one bag of tiles into a machine-learning model trained to generate a prediction of an image class label based on the at least one bag of tiles, the machine-learning model including:
a first portion trained to generate one or more feature maps based on the at least one bag of tiles; and
a second portion trained to generate the prediction of the image class label based at least in part on the one or more feature maps; and
output the prediction of the image class label.
36 . The system of claim 35 , wherein the image comprises only one whole-slide image (WSI).
37 . The system of claim 35 , further comprising receiving, by the one or more processors, the image, wherein the image comprises an image of a tissue sample.
38 . The system of claim 35 , wherein each tile of the plurality of tiles comprises a plurality of pixels corresponding to one or more regions of the image.
39 . The system of claim 35 , wherein the image comprises a histological stain image, a fluorescence in situ hybridization (FISH) image, an immunofluorescence (IF) image, or a hematoxylin and eosin (H&E) image.
40 . A method, comprising:
receiving, by one or more processors, a training image;
segmenting, by the one or more processors, the training image into a plurality of tiles; grouping, by the one or more processors, the plurality of tiles into at least one bag of tiles;
training a first portion to generate one or more feature maps based on the at least one bag of tiles;
training a second portion to generate a prediction of an image class label for the training image based at least in part on the one or more feature maps.