Systems and methods for deconvolving cell types in histology slide images, using super-resolution spatial transcriptomics data
A computer-implemented method, computing system and computer-readable medium include receiving training data and training a machine learning model to generate a cell expression map. A computer-implemented method, computing system and computer-readable medium includes receiving a histology image and a cell segmentation map and processing them using a trained machine learning model.
1 . A computer-implemented method for training a machine learning model to classify cells in a histology image, the method comprising:
receiving, via one or more processors, training data; and
training a machine learning model, using the training data, to generate a cell expression map corresponding to a cropped portion of a super-resolved gene expression heat map,
wherein the cell expression map includes a plurality of cell contours, each including respective aggregated gene expression values.
2 . The computer-implemented method of claim 1 , wherein the training data includes at least one of (i) one or more cell segmentation maps, (ii) one or more spatial transcriptomics data sets, and (iii) one or more pathologists' annotations.
3 . The computer-implemented method of claim 2 , wherein at least one of the cell segmentation maps corresponds to a respective one of the spatial transcriptomics data sets and to a respective one of the pathologists' annotations.
4 . The computer-implemented method of claim 1 , further comprising:
generating an image mask corresponding to the training data; and
providing the image mask to the machine learning model with the training data, to enable the model to output features enabling model outputs to be matched to model inputs.
5 . The computer-implemented method of claim 1 , further comprising:
training the machine learning model to determine the cropped portion of the super-resolved gene expression heat map by linking coordinates of the super-resolved gene expression heat map to coordinates of images in the training data.
6 . A computer-implemented method for classifying cells in an histology image, the method comprising:
receiving, via one or more processors, the histology image and a cell segmentation map;
processing the histology image and the cell segmentation map using a trained machine learning model to generate a cell expression map including a plurality of cell contours, each including a respective predicted aggregated gene expression,
wherein the machine learning model is trained using training data; and
for each of the cell contours, processing the cell contour to generate a respective predicted cell type using a model trained to predict cell types based on aggregated gene expression information.
7 . The computer-implemented method of claim 6 , wherein the aggregated gene expression information includes an abundance value or expression level for each gene in a plurality of genes.
8 . The computer-implemented method of claim 6 , wherein processing the cell contour to generate the respective cell type includes generating a gene-by-gene comatrix, wherein one dimension of the comatrix represents one individual cell and another dimension represents an expression value for a particular gene for a particular cell.
9 . The computer-implemented method of claim 6 , further comprising:
processing each cell contour's gene expression profile using a trained cell type prediction machine learning model to determine a type of the cell.
10 . The computer-implemented method of claim 9 , wherein the trained cell type prediction machine learning model is a logistic regression machine learning model.
11 . A computing system for training a machine learning model to classify cells in a histology image, comprising:
one or more processors; and
one or more memories having stored thereon computer-executable instructions that, when executed by the one or more processors, cause the computing system to:
receive, via one or more processors, training data; and
train a machine learning model, using the training data, to generate a cell expression map corresponding to a cropped portion of a super-resolved gene expression heat map,
wherein the cell expression map includes a plurality of cell contours, each including respective aggregated gene expression values.
12 . The computing system of claim 11 , wherein the training data includes at least one of (i) one or more cell segmentation maps, (ii) one or more spatial transcriptomics data sets, and (iii) one or more pathologists' annotations.
13 . The computing system of claim 12 , wherein at least one of the cell segmentation maps corresponds to a respective one of the spatial transcriptomics data sets and to a respective one of the pathologists' annotations.
14 . The computing system of claim 11 , the instructions having stored thereon further instructions that, when executed by the one or more processors, cause the computing system to:
generate an image mask corresponding to the training data; and
provide the image mask to the machine learning model with the training data, to enable the model to output features enabling the model output to be matched to model inputs.
15 . The computing system of claim 11 , the instructions having stored thereon further instructions that, when executed by the one or more processors, cause the computing system to:
train the machine learning model to determine the cropped portion of the super-resolved gene expression heat map by linking coordinates of the super-resolved gene expression map to coordinates of images in the training data.
16 . A computing system for training a machine learning model to classify cells in a histology image, comprising:
one or more processors; and
one or more memories having stored thereon computer-executable instructions that, when executed by the one or more processors, cause the computing system to:
receive, via one or more processors, the histology image and a cell segmentation map;
process the histology image and the cell segmentation map using a trained machine learning model to generate a cell expression map including a plurality of cell contours, each including a respective predicted aggregated gene expression,
wherein the machine learning model is trained using training data; and
for each of the cell contours, process the cell contour to generate a respective predicted cell type using a model trained to predict cell types based on aggregated gene expression information.
17 . The computing system of claim 16 , wherein the aggregated gene expression information includes an abundance value or expression level for each gene in a plurality of genes.
18 . The computing system of claim 16 , the instructions having stored thereon further instructions that, when executed by the one or more processors, cause the computing system to:
generate a gene-by-gene comatrix, wherein one dimension of the comatrix represents one individual cell and another dimension represents an expression value for a particular gene for a particular cell.
19 . The computing system of claim 16 , the instructions having stored thereon further instructions that, when executed by the one or more processors, cause the computing system to:
process each cell contour's gene expression profile using a trained cell type prediction machine learning model to determine a type of the cell.
20 . The computing system of claim 19 , wherein the trained cell type prediction machine learning model is a logistic regression machine learning model.
21 . A non-transitory computer-readable medium having stored thereon computer-executable instructions that, when executed by one or more processors, cause a computer to:
receive, via one or more processors, training data; and
train a machine learning model, using the training data, to generate a cell expression map corresponding to a cropped portion of a super-resolved gene expression heat map,
wherein the cell expression map includes a plurality of cell contours, each including respective aggregated gene expression values.
22 . A non-transitory computer-readable medium having stored thereon computer-executable instructions that, when executed by one or more processors, cause a computer to:
receive, via one or more processors, a histology image and a cell segmentation map;
process the histology image and the cell segmentation map using a trained machine learning model to generate a cell expression map including a plurality of cell contours, each including a respective predicted aggregated gene expression,
wherein the machine learning model is trained using training data; and
for each of the cell contours, process the cell contour to generate a respective predicted cell type using a model trained to predict cell types based on aggregated gene expression information.