Systems and methods to train a cell object detector

1. A computer-implemented method to train a machine-learning model to detect a particular type of cell object in a candidate electronic image, the machine-learning model including a first neural network and a second neural network, the method comprising:

obtaining training data that includes a first set of cell images and corresponding first labels, wherein the first labels identify one or more types of objects in the first set of cell images including the particular type of cell object and a second set of cell images and corresponding second labels, wherein the second labels identify one or more look-alike objects that are similar in visual appearance to the particular type of cell object;

training the first neural network to detect one or more cell images that include candidate objects, wherein the candidate objects include objects of the particular type or look-alike objects, wherein the training comprises:

providing the first set of cell images and the second set of cell images as input to the first neural network;

obtaining, as output of the first neural network, one or more output labels for each image in the first set and the second set;

comparing the output labels for each image with the corresponding labels in the first set or the second set; and

adjusting one or more parameters of the first neural network based on the comparing; and

training the second neural network to identify an object of the particular type of cell object by providing the one or more cell images that include the candidate objects detected by the first neural network as input to the second neural network and the labels associated with the one or more cell images, wherein after the training the second neural network can distinguish between objects of the particular type and look-alike objects.

2. The computer-implemented method of claim 1 , further comprising:

obtaining, as output of the second neural network, an indication of whether the particular type of cell object is present in a cell image.

3. The method of claim 1 , wherein the second set of cell images and corresponding second labels further identify one or more borderline objects similar in visual appearance to the particular type of cell object.

4. The method of claim 1 , wherein the second set of cell images and corresponding second labels further identify one or more of normal objects or background objects.

5. The method of claim 1 , wherein the particular type of object includes a mitotic figure.

6. The method of claim 1 , wherein the particular type of object includes an epithelial region.

7. The method of claim 1 , wherein training the second neural network further includes:

receiving an initial outline of a potential object within an image in the first set of cell images;

moving a center of a sample window to each pixel within the initial outline; and

at each pixel location within the initial outline:

obtaining a sample of image data corresponding to the sample window;

performing one or more transformations on the sample of image data to generate one or more training images; and

providing the training images to the second neural network as training data.

8. A non-transitory computer-readable medium with instructions stored thereon that, when executed by one or more computers, cause the one or more computers to perform operations to determine whether a particular type of cell object is present in a cell image, the operations comprising:

providing the cell image as input to a trained machine-learning model, wherein the trained machine-learning model includes a first neural network and a second neural network, and is trained by:

obtaining training data that includes a first set of cell images and corresponding first labels, wherein the first labels identify one or more types of objects in the first set of cell images including the particular type of cell object and a second set of cell images and corresponding second labels, wherein the second labels identify a look-alike object that is similar in visual appearance to the particular type of cell object;

training the first neural network to detect one or more cell images that include candidate objects, wherein the candidate objects include objects of the particular type of cell object or a look-alike object type, wherein the training comprises:

providing the first set of cell images and the second set of cell images as input to the first neural network;

obtaining, as output of the first neural network, one or more output labels for each image in the first set and the second set;

comparing the output labels for each image with the corresponding labels in the first set or the second set; and

adjusting one or more parameters of the first neural network based on the comparing; and

training the second neural network to identify an object of the particular type of cell object by providing the one or more cell images that include candidate objects detected by the first neural network as input to the second neural network and the labels associated with the one or more cell images, wherein after the training the second neural network can distinguish between objects of the particular type and look-alike objects; and

obtaining, as output of the second neural network, an indication of whether the particular type of cell object is present in the cell image.

9. The non-transitory computer-readable medium of claim 8 , wherein training the second neural network further includes:

receiving an initial outline of a potential object within an image in the first set of cell images;

moving a center of a sample window to each pixel within the initial outline;

at each pixel location within the initial outline:

obtaining a sample of image data corresponding to the sample window;

performing one or more transformations on the sample of image data to generate one or more training images; and

providing the training images to the second neural network as training data.

10. The non-transitory computer-readable medium of claim 8 , wherein the second set of cell images and corresponding second labels further identify one or more borderline objects similar in visual appearance to the particular type of cell object.

11. The non-transitory computer-readable medium of claim 8 , wherein the second set of cell images and corresponding second labels further identify one or more of normal objects or background objects.

12. The non-transitory computer-readable medium of claim 8 , wherein the particular type of object includes a mitotic figure.

13. The non-transitory computer-readable medium of claim 8 , wherein the particular type of object includes an epithelial region.

14. A system, comprising:

one or more processors coupled to a computer readable memory having stored thereon software instructions that, when executed by the one or more processors, cause the one or more processors to perform or control performance of operations including:

providing a cell image as input to a trained machine-learning model, wherein the trained machine-learning model includes a first neural network and a second neural network, and is trained by:

obtaining training data that includes a first set of cell images and corresponding first labels, wherein the first labels identify one or more types of objects in the first set of cell images including a particular type of cell object and a second set of cell images and corresponding second labels, wherein the second labels identify a look-alike object that is similar in visual appearance to the particular type of cell object;

training the first neural network to detect one or more cell images that include candidate objects, wherein the candidate objects include objects of the particular type of cell object or a look-alike object type, wherein the training comprises:

providing the first set of cell images and the second set of cell images as input to the first neural network;

obtaining, as output of the first neural network, one or more output labels for each image in the first set and the second set;

comparing the output labels for each image with the corresponding labels in the first set or the second set; and

adjusting one or more parameters of the first neural network based on the comparing; and

training the second neural network to identify an object of the particular type of cell object by providing the one or more cell images that include candidate objects detected by the first neural network as input to the second neural network and the labels associated with the one or more cell images, wherein after the training the second neural network can distinguish between objects of the particular type and look-alike objects; and

obtaining, as output of the second neural network, an indication of whether the particular type of cell object is present in a cell image.

15. The system of claim 14 , wherein training the second neural network further includes:

receiving an initial outline of a potential object within an image in the first set of cell images;

moving a center of a sample window to each pixel within the initial outline; and

at each pixel location within the initial outline:

obtaining a sample of image data corresponding to the sample window;

performing one or more transformations on the sample of image data to generate one or more training images; and

providing the training images to the second neural network as training data.

16. The system of claim 14 , wherein the second set of cell images and corresponding second labels further identify one or more borderline objects similar in visual appearance to the particular type of cell object.

17. The system of claim 14 , wherein the second set of cell images and corresponding second labels further identify one or more of normal objects or background objects.

18. The system of claim 14 , wherein the particular type of object includes a mitotic figure.

19. The system of claim 14 , wherein the particular type of object includes an epithelial region.

20. The system of claim 14 , wherein the second set of cell images and corresponding second labels further identify one or more normal objects or background objects.