Predicting patient responses to a chemical substance
In an aspect, a data processing system includes a computer-readable memory comprising computer-executable instructions, and at least one processor configured to execute executable logic including at least one artificial neural network trained to predict one or more responses to a chemical substance by identifying one or more discrete biological tissue components in a biological image. When the at least one processor is executing the computer-executable instructions, the at least one processor is configured to carry out operations including: receiving spatially arranged image data representing a biological image of a patient; generating spatially arranged image tile data representing a plurality of image tiles; processing the spatially arranged image tile data through one or more data structures storing one or more portions of executable logic included in the artificial neural network to predict one or more responses of a patient.
1. A data processing system, comprising:
a computer-readable memory comprising computer-executable instructions; and
at least one processor configured to execute executable logic including an artificial neural network trained to predict a response to a chemical substance by identifying one or more discrete biological tissue components in an input biological image, wherein when the at least one processor is executing the computer-executable instructions, the at least one processor is configured to carry out operations comprising:
receiving image data representing a biological image of a patient;
processing the image data using the artificial neural network and in accordance with trained values of a set of artificial neural network parameters to generate a response score that defines a predicted response of the patient to the chemical substance, wherein the processing comprises, at each of a plurality of time steps in a sequence of time steps:
obtaining current location data for the time step that defines a location of a current patch of the biological image;
processing the current patch of the biological image using a feature extraction module of the artificial neural network to generate a feature representation of the current patch of the biological image that characterizes locations of discrete biological tissue components in the current patch of the biological image;
processing the feature representation of the current patch of the biological image using a location module of the artificial neural network to generate a next location data for a next time step that defines a location of a next patch of the biological image to be processed at a next time step; and
providing the next location data for processing at the next time step in the sequence of time steps; and
wherein the processing further comprises generating the response score that defines the predicted response of the patient to the chemical substance based on the feature representations of the patches of the biological image.
2. The data processing system of claim 1 , wherein the biological image comprises an immunohistochemistry image.
3. The data processing system of claim 1 , wherein the artificial neural network comprises a deep recurrent attention model.
4. The data processing system of claim 1 , the operations further comprising:
generating preprocessed spatially arranged image data representing a preprocessed biological image;
wherein generating preprocessed spatially arranged image data comprises identifying one or more pixels of the biological image representing one or more locations of biological tissue and color normalizing one or more locations of biological tissue.
5. The data processing system of claim 1 , wherein the predicted response of the patient to the chemical substance defines a predicted reduction in a size of a tumor.
6. The data processing system of claim 1 , wherein the feature extraction module and the location module of the artificial neural network have been jointly trained to optimize a reinforcement learning objective function.
7. The data processing system of claim 1 , wherein the chemical substance comprises a drug for treating a medical condition, wherein the patient has the medical condition.
8. A method performed by at least one processor executing executable logic including at an artificial neural network trained to predict a response to a chemical substance by identifying one or more discrete biological tissue components in an input biological image, the method comprising:
receiving image data representing a biological image of a patient;
processing the image data using the artificial neural network and in accordance with trained values of a set of artificial neural network parameters to generate a response score that defines a predicted response of the patient to the chemical substance, wherein the processing comprises, at each of a plurality of time steps in a sequence of time steps:
obtaining current location data for the time step that defines a location of a current patch of the biological image;
processing the current patch of the biological image using a feature extraction module of the artificial neural network to generate a feature representation of the current patch of the biological image that characterizes locations of discrete biological tissue components in the current patch of the biological image;
processing the feature representation of the current patch of the biological image using a location module of the artificial neural network to generate a next location data for a next time step that defines a location of a next patch of the biological image to be processed at a next time step; and
providing the next location data for processing at the next time step in the sequence of time steps; and
wherein the processing further comprises generating the response score that defines the predicted response of the patient to the chemical substance based on the feature representations of the patches of the biological image.
9. The method of claim 8 , wherein the biological image comprises an immunohistochemistry image.
10. The method of claim 8 , wherein the artificial neural network comprises a deep recurrent attention model.
11. The method of claim 8 , further comprising:
generating preprocessed spatially arranged image data representing a preprocessed biological image;
wherein generating preprocessed spatially arranged image data comprises identifying one or more pixels of the biological image representing one or more locations of biological tissue and color normalizing one or more locations of biological tissue.
12. The method of claim 8 , wherein the predicted response of the patient to the chemical substance defines a predicted reduction in a size of a tumor.
13. The method of claim 8 , wherein the feature extraction module and the location module of the artificial neural network have been jointly trained to optimize a reinforcement learning objective function.
14. The method of claim 8 , wherein the chemical substance comprises a drug for treating a medical condition, wherein the patient has the medical condition.
15. One or more non-transitory computer storage media storing instructions that when executed by one or more computers cause the one or more computers to perform operations comprising:
receiving image data representing a biological image of a patient;
processing the image data using an artificial neural network and in accordance with trained values of a set of artificial neural network parameters to generate a response score that defines a predicted response of the patient to a chemical substance, wherein the processing comprises, at each of a plurality of time steps in a sequence of time steps:
obtaining current location data for the time step that defines a location of a current patch of the biological image;
processing the current patch of the biological image using a feature extraction module of the artificial neural network to generate a feature representation of the current patch of the biological image that characterizes locations of discrete biological tissue components in the current patch of the biological image;
processing the feature representation of the current patch of the biological image using a location module of the artificial neural network to generate a next location data for a next time step that defines a location of a next patch of the biological image to be processed at a next time step; and
providing the next location data for processing at the next time step in the sequence of time steps; and
wherein the processing further comprises generating the response score that defines the predicted response of the patient to the chemical substance based on the feature representations of the patches of the biological image.
16. The non-transitory computer storage media of claim 15 , wherein the biological image comprises an immunohistochemistry image.
17. The non-transitory computer storage media of claim 15 , wherein the artificial neural network comprises a deep recurrent attention model.
18. The non-transitory computer storage media of claim 15 , wherein the operations further comprise:
generating preprocessed spatially arranged image data representing a preprocessed biological image;
wherein generating preprocessed spatially arranged image data comprises identifying one or more pixels of the biological image representing one or more locations of biological tissue and color normalizing one or more locations of biological tissue.
19. The non-transitory computer storage media of claim 15 , wherein the predicted response of the patient to the chemical substance defines a predicted reduction in a size of a tumor.
20. The non-transitory computer storage media of claim 15 , wherein the feature extraction module and the location module of the artificial neural network have been jointly trained to optimize a reinforcement learning objective function.